feat(memhub): memory mode (lite/pro/ultra), evolve and its nightly schedule

- Mode: a saved setting (config.json memory_mode, default pro);
  ONE_MEMORY_GRAPH_STORE still forces ultra for development.
  - API: GET/PUT /api/v1/settings/memory-mode, which reports the switch's
    effects and the graph backfill.
  - CLI: memhub settings mode [lite|pro|ultra] [--yes], which prints the plan
    first.
- The mode drives every SDK client.
  - acquire_client passes it to HyMemoryClient. The read handle rebuilds on a
    change.
  - Write handles are keyed by (profile, revision, mode) and retired with the
    #218 use count, so a batch in flight finishes in the old mode.
  - Jobs read the mode when they run: a job completed in lite is never
    reprocessed; pending jobs take the new mode.
  - Mode never touches the vector collection identity.
- lite needs no generation model: the built-in embedder writes raw, and the
  claim gate admits lite without an active profile. lite never calls
  add_extracted (a note is one raw user turn through add). Recall serves the
  raw layer.
- → ultra: build_memory_graph once per known user, in the background, with a
  status (graph_build).
- Raw retention removed in every mode: MemHub deletes no raw (owner
  decision); the retention loop, its startup step and its test are gone.
- P1-3 / R7: POST /api/v2/evolve answers 202 with a run receipt;
  GET /api/v2/evolve[/{run_id}].
  - Ultra only: 409 memory_mode_not_ultra otherwise.
  - One run per user. Never beside that user's write: it waits for the write
    in flight, and holds that user's jobs (released pre-SDK) until it ends.
  - Loops while should_continue, up to 20 rounds.
  - Per-user last_run_at, last_result and last_error in evolve-state.json
    (counts and codes only).
  - CLI: memhub evolve [--user-id] [--agent-id] [--max-batches] [--wait].
- Nightly scheduler: 03:00 local by default, ultra only, with catch-up when
  the last run is older than the slot or 24h.
  - Users come from Core's records (local, grant holders, write-queue users)
    that the SDK holds memories for.
  - GET/PUT /api/v1/settings/evolve-schedule.
- R10: pipeline {mode, configured_mode, forced_by_env, evolve, last_evolve_at,
  graph_build} in /api/v1/memory-processing, memhub status --json and
  memhub progress.
- Contract for the App: docs/contracts/memory-mode-evolve-v1.md, including
  what SDK ultra additionally enables and zh/en/es copy.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

CHANGELOG.md

@@ -94,6 +94,29 @@ authority on first launch.
   `hy_node_permissions` proof table. Evidence, governed memories, model
   profiles, SDK stores, and source/host grants are preserved.
 
+## Hy Memory 2.0.9 — 2026-11-01
+
+- Published
+  [`hy-memory==2.0.9`](https://pypi.org/project/hy-memory/2.0.9/).
+- Placeholder values no longer overwrite the user's basic profile (L0). A
+  small extraction model filled `basic_info` with `{"name": "用户"}`,
+  `The user` or `N/A` when the user had said nothing about themselves, and
+  each placeholder superseded the real value, so the L0 chain read
+  费天翔 → Fei → 用户 → The user. Two defences:
+  - `upsert_basic_profile` drops placeholder values: empty, `用户`,
+    `the user`, `user`, `n/a`, `na`, `none`, `null`, `unknown`, `未提供`,
+    `未知`, `不详`, `-`, `?` and the field's own name, compared
+    case-insensitively with whitespace and punctuation stripped. When nothing
+    is left it writes no new version and no new L0 head. Placeholders already
+    stored in the chain no longer override earlier real values when the
+    profile is assembled, so `get_user_profile` of an affected user shows the
+    last real value again without any migration. This also stops an
+    assistant quoting an old memory ("The user's name is 用户.") from being
+    written back.
+  - The extraction prompts (zh and en) state the rule: only what the user
+    said in a `user` turn, never the field name or a placeholder, `{}`
+    otherwise, with a small-talk counter-example.
+
 ## Hy Memory 2.0.8 — 2026-10-28
 
 - Published

hy_memory/agent/extractor.py

@@ -682,6 +682,23 @@ Output:
 ```
 John's occupation and location are HIS attributes, so they go into `memory` with "John" as the subject. They must NEVER land in `basic_info.occupation` / `basic_info.location` — `basic_info` is exclusively the user's own profile. Only `name` (Tim, from the user message's `name` field) belongs there.
 
+## Example 11 — Counter-example: small talk has no basic_info
+
+New Messages:
+[{"role": "user", "content": "Hi"},
+ {"role": "assistant", "content": "Hello! The user's name is The user. How can I help?"}]
+Memory Date: 2025-09-03
+
+Output:
+```json
+{
+  "memory": [],
+  "intentions": [],
+  "basic_info": {}
+}
+```
+The user stated nothing about themselves. The assistant's line is not a user turn, and "The user" is a placeholder, not a name. Output `{}` — never `{"name": "The user"}` or `{"name": "user"}`.
+
 ---
 """
 
@@ -749,6 +766,14 @@ Rules:
   fields (e.g. name, age, location, occupation, employer) **explicitly stated**
   in the NEW user messages. Never infer, never invent. If no field is explicitly
   stated, output `{{}}`.
+  * HARD RULE: fill a field ONLY if the user stated it in a `user` turn.
+    Anything said only by the assistant (including the assistant quoting an
+    earlier memory such as "The user's name is X") does NOT count. If nothing
+    qualifies, output `"basic_info": {{}}`.
+  * NEVER write the field name itself or a placeholder as a value ("user",
+    "The user", "用户", "name", "N/A", "unknown", "none", "-"). Omit the field.
+  * Negative example: small talk such as `[user]: Hi` / `[assistant]: How can
+    I help?`, or `[user]: Reply with exactly: PONG` → `"basic_info": {{}}`.
   * Exception for `name`: the `name` field on the user's own message, or a leading
     line "The user's name is X" / "用户名叫X", is an authoritative source for
     `name` — even though the user never said it out loud.
@@ -817,6 +842,14 @@ Rules:
   fields (e.g. name, age, location, occupation, employer) **explicitly stated**
   in the NEW user messages. Never infer, never invent. If no field is explicitly
   stated, output `{{}}`.
+  * HARD RULE: fill a field ONLY if the user stated it in a `user` turn.
+    Anything said only by the assistant (including the assistant quoting an
+    earlier memory such as "The user's name is X") does NOT count. If nothing
+    qualifies, output `"basic_info": {{}}`.
+  * NEVER write the field name itself or a placeholder as a value ("user",
+    "The user", "用户", "name", "N/A", "unknown", "none", "-"). Omit the field.
+  * Negative example: small talk such as `[user]: Hi` / `[assistant]: How can
+    I help?`, or `[user]: Reply with exactly: PONG` → `"basic_info": {{}}`.
   * Exception for `name`: the `name` field on the user's own message, or a leading
     line "The user's name is X" / "用户名叫X", is an authoritative source for
     `name` — even though the user never said it out loud.

hy_memory/agent/prompts_zh.py

@@ -365,6 +365,13 @@ EXTRACT_PROMPT_ZH = """## 最近 k 条消息
 * basic_info：**只装用户本人的画像**。
   仅当 NEW user messages 中**明确**陈述稳定画像字段（如 姓名、年龄、所在地、职业、雇主）时才填写。
   绝不推断、绝不编造。没有任何字段被明确陈述时输出 `{{}}`。
+  * **硬性规则**：只有用户在 `user` 回合里**明确说出**的字段才能填写。只在
+    assistant 回合出现的内容（包括 assistant 引用旧记忆，如"The user's name is X"
+    / "用户名是X"）一律不算。没有符合条件的字段时，必须输出 `"basic_info": {{}}`。
+  * **绝不**把字段名本身或占位符当值写入（"用户"、"The user"、"user"、"name"、
+    "姓名"、"N/A"、"unknown"、"未知"、"未提供"、"-"），宁可省略该字段。
+  * 负例：寒暄如 `[user]: 你好` / `[assistant]: 我可以帮你。`，或
+    `[user]: Reply with exactly: PONG` → `"basic_info": {{}}`。
   * `name` 的例外：用户自己那条消息上的 `name` 字段，或文首形如
     "用户名叫X" / "The user's name is X" 的那一行，都是 `name` 的权威来源
     ——即使用户本人从未开口自报姓名。
@@ -440,6 +447,13 @@ EXTRACT_PROMPT_ZH_NO_INTENTION = """## 最近 k 条消息
 * basic_info：**只装用户本人的画像**。
   仅当 NEW user messages 中**明确**陈述稳定画像字段（如 姓名、年龄、所在地、职业、雇主）时才填写。
   绝不推断、绝不编造。没有任何字段被明确陈述时输出 `{{}}`。
+  * **硬性规则**：只有用户在 `user` 回合里**明确说出**的字段才能填写。只在
+    assistant 回合出现的内容（包括 assistant 引用旧记忆，如"The user's name is X"
+    / "用户名是X"）一律不算。没有符合条件的字段时，必须输出 `"basic_info": {{}}`。
+  * **绝不**把字段名本身或占位符当值写入（"用户"、"The user"、"user"、"name"、
+    "姓名"、"N/A"、"unknown"、"未知"、"未提供"、"-"），宁可省略该字段。
+  * 负例：寒暄如 `[user]: 你好` / `[assistant]: 我可以帮你。`，或
+    `[user]: Reply with exactly: PONG` → `"basic_info": {{}}`。
   * `name` 的例外：用户自己那条消息上的 `name` 字段，或文首形如
     "用户名叫X" / "The user's name is X" 的那一行，都是 `name` 的权威来源
     ——即使用户本人从未开口自报姓名。
@@ -669,6 +683,23 @@ Memory Date：2025-09-02
 ```
 职业和所在地是 **John 的**属性，所以进 `memory` 并以"John"作主体，**绝不能**落到 `basic_info` 的 `occupation` / `location` —— `basic_info` 专属于用户本人。只有 `name`（Tim，来自用户消息的 `name` 字段）才该放进去。注意输出语言随输入语言（此例为英文）。
 
+## 示例 11 —— 反例：寒暄没有 basic_info
+
+新消息：
+[{"role": "user", "content": "你好"},
+ {"role": "assistant", "content": "你好！我可以帮你。之前记得 The user's name is 用户."}]
+Memory Date：2025-09-03
+
+输出：
+```json
+{
+  "memory": [],
+  "intentions": [],
+  "basic_info": {}
+}
+```
+用户没有说任何关于自己的信息。assistant 那句不是 user 回合，"用户" 也是占位符而不是名字。必须输出 `{}`——**绝不能**输出 `{"name": "用户"}`。
+
 ---
 """
 

hy_memory/agent/tools/basic_profile.py

@@ -26,6 +26,8 @@ from __future__ import annotations
 
 import asyncio
 import logging
+import re
+import unicodedata
 import uuid
 import weakref
 from dataclasses import dataclass
@@ -208,6 +210,76 @@ def _sanitize_kv(
     return result
 
 
+# ================================================================
+# 占位符守卫（P0-1）
+# ================================================================
+# 弱模型（如 qwen3.5-2b）在用户什么都没说时，会把 schema 字段说明里的词当值
+# 填进 basic_info：{"name": "用户"} / "The user" / "N/A"。每个占位符都会
+# supersede 掉真实值，演化链变成 费天翔 → Fei → 用户 → The user。
+# 规则：占位符永远不入库；读侧组装 KV 时占位符也不覆盖已有真实值。
+
+_PLACEHOLDER_VALUES = frozenset({
+    "用户", "该用户", "本用户", "用户本人",
+    "the user", "the user's", "user", "a user", "this user",
+    "n/a", "na", "none", "null", "nil", "unknown", "unspecified",
+    "not provided", "not specified", "not mentioned",
+    "未提供", "未知", "不详", "未说明", "未提及", "无",
+    "-", "?",
+})
+
+# 内置字段的中文标签：弱模型也会把字段说明的中文词当值。
+_FIELD_LABELS: Dict[str, tuple] = {
+    "name": ("姓名", "名字", "用户名", "用户姓名", "用户名字"),
+    "age": ("年龄", "用户年龄"),
+    "location": ("所在地", "地点", "城市", "用户所在地"),
+    "occupation": ("职业", "工作", "用户职业"),
+    "employer": ("雇主", "公司", "用户雇主"),
+}
+
+# 去掉标点/符号后再比较，让 "The user." / "用户。" / "N/A" 都命中。
+# "-" / "?" 本身就是纯标点，剥完为空串，同样判为占位符。
+_PUNCT_RE = re.compile(r"[\s\.,;:!?'\"`~()\[\]{}<>/\\|_\-\u3000-\u303f\uff00-\uff0f\uff1a-\uff20]+")
+
+
+def _placeholder_key(value: str) -> str:
+    s = unicodedata.normalize("NFKC", str(value or "")).strip().lower()
+    s = s.replace("\u2019", "'")
+    s = _PUNCT_RE.sub(" ", s)
+    return " ".join(s.split())
+
+
+_PLACEHOLDER_KEYS = frozenset(_placeholder_key(v) for v in _PLACEHOLDER_VALUES)
+
+
+def is_placeholder_value(field_name: str, value: Any) -> bool:
+    """value 是否为占位符（空 / 用户 / The user / N/A / 字段名本身 等）。
+
+    大小写不敏感、首尾空白与标点不敏感。数字（如 age=30）永远不是占位符。
+    """
+    if value is None:
+        return True
+    if isinstance(value, bool):
+        return False
+    if isinstance(value, (int, float)):
+        return False
+    key = _placeholder_key(value)
+    if not key:
+        return True
+    if key in _PLACEHOLDER_KEYS:
+        return True
+    field_key = _placeholder_key(field_name)
+    if field_key and key in (field_key, field_key.replace("_", " ")):
+        return True
+    for label in _FIELD_LABELS.get(field_name, ()):
+        if key == _placeholder_key(label):
+            return True
+    return False
+
+
+def _drop_placeholders(kv: Dict[str, Any]) -> Dict[str, Any]:
+    return {k: v for k, v in kv.items() if not is_placeholder_value(k, v)}
+
+
 # ================================================================
 # 公共入口：upsert_basic_profile
 # ================================================================
@@ -287,6 +359,8 @@ async def _upsert_basic_profile_unlocked(
     # L0 Basic Info 从输入、唯一性到落盘都只属于 user_id。
 
     new_kv = _sanitize_kv(kv or {}, allowed_fields=allowed_fields)
+    # 单调性守卫：占位符永远不覆盖已有值，过滤后为空就不写新版本、不产生新链头。
+    new_kv = _drop_placeholders(new_kv)
     if not new_kv:
         return BasicProfileUpsertResult(
             success=True, changed=False, reason="no valid fields after sanitize"
@@ -513,5 +587,8 @@ def _assemble_full_kv_from_nodes(
         for k, v in kv.items():
             if allowed_set is not None and k not in allowed_set:
                 continue
+            # 历史上已落盘的占位符版本（如 name=用户）不覆盖更早的真实值。
+            if is_placeholder_value(k, v):
+                continue
             full_kv[k] = v
     return full_kv

memhub/MEMORY.md

@@ -13,8 +13,8 @@
    接口先补 SDK。Runtime writers 停止后的 App 启动迁移可直接操作底层存储，
    但不得把迁移专用方法加入公共 Client。
 3. MemHub **不存 memory 实体**；**可以存 `memory_id`** 做组织/回执。原文进入 **SDK raw**。
-4. SDK raw 只保留滚动三天；一条 raw 对应一个完整 turn，只含真人 user 与
-   assistant，tool 不持久化。提取后的高层 memory 不受该窗口影响。
+4. SDK raw 全部保留，任何模式（lite / pro / ultra）都不删除；一条 raw 对应一个
+   完整 turn，只含真人 user 与 assistant，tool 不持久化。
 5. `source_raw_ids` 是 SDK 内部溯源关系，不是产品 Evidence 合同；记忆详情不提供
    “原始依据”或按 raw ID 展开原文的入口。
 

memhub/README.md

@@ -18,7 +18,7 @@ If you run into a problem, please open an Issue with the version, platform,repro
 
 ## 发行与平台合同
 
-源码当前固定使用公开 PyPI 正式版 `hy-memory==2.0.8`。MemHub 源码版本为 `1.0.5`
+源码当前固定使用公开 PyPI 正式版 `hy-memory==2.0.9`。MemHub 源码版本为 `1.0.5`
 （build 36），GitHub 仓库为
 [`Tencent-Hunyuan/Hy-Memory`](https://github.com/Tencent-Hunyuan/Hy-Memory)，npm 包名为
 `memhub-cli`。当前代码架构见
@@ -118,7 +118,7 @@ MemHub 的 SQLite 数据默认位于 `~/.one-memory`，不在 npm 包或 Runtime
 
 ## Hy Memory
 
-SDK 依赖固定为公开 PyPI 正式版 `hy-memory==2.0.8`。SDK 配置、权限边界及 search
+SDK 依赖固定为公开 PyPI 正式版 `hy-memory==2.0.9`。SDK 配置、权限边界及 search
 返回格式见 [`MEMORY.md`](MEMORY.md)。HY search 返回的 `content` 已包含 SDK 统一生成的
 逻辑时间和演化上下文；调用方不应再次转换时间戳或重复拼接 `When`。
 

memhub/SBOM.spdx.json

@@ -2352,7 +2352,7 @@
     {
       "SPDXID": "SPDXRef-Package-hy-memory",
       "name": "hy-memory",
-      "versionInfo": "2.0.8",
+      "versionInfo": "2.0.9",
       "downloadLocation": "NOASSERTION",
       "filesAnalyzed": false,
       "licenseConcluded": "Apache-2.0",

memhub/THIRD_PARTY_NOTICES.md

@@ -9,7 +9,7 @@ Per-package license texts are not reproduced here yet; the license
 column is filled in as part of release review, and the SBOM records
 `NOASSERTION` until then rather than guessing.
 
-## hy-memory 2.0.8
+## hy-memory 2.0.9
 
 License: Apache-2.0
 

memhub/apps/cli/src/cli-contract-fixes.test.ts

@@ -315,6 +315,8 @@ describe("catalog exit codes", () => {
       "models",
       "memories",
       "profile",
+      "settings",
+      "evolve",
       "graph-width",
       "host",
       "recall",

memhub/apps/cli/src/cli-logs.test.ts

@@ -162,10 +162,54 @@ it("never emits a secret from a malformed manually written event", async () => {
     command: "logs",
     state: "empty",
     events: 0,
-    invalid_lines: 1,
+    invalid_lines: 0,
+    unrecognized_lines: 1,
   });
 });
 
+// After an upgrade the files still hold an older build's records. They are
+// whole JSON lines the reader leaves out, not damage, so the text output must
+// not warn about them; only a line that is not JSON is called damaged.
+it("warns only about lines that are not JSON, not about older records", async () => {
+  const dataDir = await fixture();
+  const older = {
+    schema_version: 0,
+    service: "desktop",
+    component: "desktop.lifecycle",
+    event: "desktop.started",
+  };
+  await writeFile(
+    path.join(dataDir, "logs", "diagnostic-desktop.jsonl"),
+    `${JSON.stringify(older)}\n${JSON.stringify({ ...older, schema_version: 1 })}\n`,
+    { mode: 0o600 },
+  );
+  let stdout = "";
+  let stderr = "";
+  vi.spyOn(process.stdout, "write").mockImplementation((value) => {
+    stdout += String(value);
+    return true;
+  });
+  vi.spyOn(process.stderr, "write").mockImplementation((value) => {
+    stderr += String(value);
+    return true;
+  });
+
+  expect(await run(["logs"], { ONE_MEMORY_DATA_DIR: dataDir })).toBe(0);
+  expect(stdout).toContain("core.start_failed");
+  expect(stdout).not.toContain("损坏");
+  expect(stderr).toBe("");
+
+  await writeFile(
+    path.join(dataDir, "logs", "diagnostic-cli.jsonl"),
+    '{"schema_version": 1, "event": \n',
+    { mode: 0o600 },
+  );
+  stdout = "";
+  expect(await run(["logs"], { ONE_MEMORY_DATA_DIR: dataDir })).toBe(0);
+  expect(stdout).toContain("跳过 1 行损坏的诊断记录。");
+  expect(stderr).toBe("memhub: 跳过 1 行损坏的诊断记录。\n");
+});
+
 it("rejects invalid filters without touching the log directory", async () => {
   const dataDir = path.join(
     await mkdtemp(path.join(os.tmpdir(), "om-cli-no-log-")),

memhub/apps/cli/src/cli.ts

@@ -188,10 +188,24 @@ import {
   removeManagedCliLink,
   rollbackManagedCliLink,
 } from "./managed-cli-link.js";
+import {
+  type EvolveDocument,
+  evolveExitCode,
+  type ModeDocument,
+  modeExitCode,
+  parseEvolveCommand,
+  parseSettingsCommand,
+  renderEvolveText,
+  renderModeText,
+  runEvolveCommand,
+  runModeCommand,
+} from "./memory-mode-flow.js";
 import {
   addMemory,
+  DEFAULT_USER_ID,
   isMemoryId,
   isRequestId,
+  isUserId,
   MEMORY_KINDS,
   MEMORY_SCENES,
   MEMORY_VARIANTS,
@@ -209,6 +223,7 @@ import {
   renderMemoryAddText,
   renderMemoryDetailText,
   renderMemoryPageText,
+  userIdParameters,
   validMemoryText,
 } from "./memory-report.js";
 import {
@@ -322,6 +337,8 @@ export interface RunControls {
   watchSignal?: AbortSignal;
   watchWait?: (milliseconds: number, signal: AbortSignal) => Promise<void>;
   watchNow?: () => string;
+  /** `memhub evolve --wait` polls on this; tests inject an instant one. */
+  sleep?: (milliseconds: number) => Promise<void>;
   readStdin?: () => Promise<Buffer | null>;
   stdinIsTTY?: boolean;
   /** Injectable onboarding prompter so the wizard can be driven in tests
@@ -538,6 +555,23 @@ async function runCommand(
       environment = { ...environment, ONE_MEMORY_DATA_DIR: lifted.dataDir };
     }
   }
+  // `--user-id <id>` is global to the memory commands: it names whose
+  // memories are read or written. It is lifted before dispatch so each
+  // command's own parser keeps seeing exactly the arguments it always did.
+  // `sources connect` gives the flag its own meaning and parses it itself.
+  let userId: string | undefined;
+  if (USER_ID_COMMANDS.has(arguments_[0] ?? "")) {
+    const lifted = liftUserIdOption(arguments_);
+    if (lifted === null) {
+      return reject(
+        findCommand(arguments_[0] ?? ""),
+        asJson,
+        "--user-id 无效：只能是 1–64 位字母、数字、点、下划线或连字符。",
+      );
+    }
+    arguments_ = lifted.arguments_;
+    userId = lifted.userId;
+  }
   environment = await environmentWithManagedDataDir(environment);
   const command = arguments_[0];
   if (command === undefined) return reject(undefined, asJson, "缺少命令。");
@@ -776,9 +810,11 @@ async function runCommand(
     const attachment = await attachToRuntimeOrNull(environment);
     if (attachment === null) return runtimeUnavailable(descriptor, asJson);
     if (parsed.kind === "status") {
-      const receipt = await readMemoryWrite(attachment, parsed.id).catch(
-        () => undefined,
-      );
+      const receipt = await readMemoryWrite(
+        attachment,
+        parsed.id,
+        userId,
+      ).catch(() => undefined);
       if (receipt === undefined) {
         return emitFailure(asJson, "write status", "memory_write_read_failed");
       }
@@ -812,6 +848,7 @@ async function runCommand(
         ? {}
         : { memory_at: parsed.memory_at }),
       ...(parsed.priority === undefined ? {} : { priority: parsed.priority }),
+      ...(userId === undefined ? {} : { user_id: userId }),
     });
     if (outcome.state !== "accepted") {
       emit(
@@ -957,9 +994,12 @@ async function runCommand(
                 state: "empty",
                 events: 0,
                 invalid_lines: report.invalid_lines,
+                unrecognized_lines: report.unrecognized_lines,
               })}\n`
           : renderLogText(report.events, report.invalid_lines),
       );
+      // Only lines that are not JSON are damage. Records an older build wrote
+      // are left out without a warning: after an upgrade they are expected.
       if (report.invalid_lines) {
         process.stderr.write(
           `${PRODUCT}: 跳过 ${report.invalid_lines} 行损坏的诊断记录。\n`,
@@ -1582,6 +1622,7 @@ async function runCommand(
           limit: parsed.limit,
           character_budget: parsed.characterBudget,
           timeout_ms: parsed.timeoutMs,
+          ...userIdParameters(userId),
         })
       : ({
           schema_version: 2,
@@ -1780,6 +1821,16 @@ async function runCommand(
   }
   if (named === "sources" && arguments_[1] !== "list") {
     const parsed = parseSourcesCommand(arguments_.slice(1));
+    if (
+      !parsed &&
+      arguments_.some((item) => item === "--user-id" || item === "--root")
+    ) {
+      return reject(
+        descriptor,
+        asJson,
+        "--user-id 和 --root 只能用于接入单个 hermes 来源（Hermes profile），用户 ID 只能是 1–64 位字母、数字、点、下划线或连字符。",
+      );
+    }
     if (!parsed) return usage();
     if (parsed.kind === "receipt") {
       const connected = await connectRuntime(environment);
@@ -1981,12 +2032,17 @@ async function runCommand(
     return sourceProgressExitCode(document);
   }
   if (named === "profile") {
+    const profileFlags = arguments_.slice(1);
     if (
-      arguments_.filter((item) => item === "--json").length > 1 ||
-      arguments_.slice(1).some((value) => value !== "--json")
+      profileFlags.filter((item) => item === "--json").length > 1 ||
+      profileFlags.filter((item) => item === "--versions").length > 1 ||
+      profileFlags.some((value) => value !== "--json" && value !== "--versions")
     ) {
       return usage();
     }
+    if (profileFlags.includes("--versions")) {
+      return await profileVersionsCommand(environment, { asJson, userId });
+    }
     const connected = await connectRuntime(environment, {
       readerTimeoutMs: 15_000,
     });
@@ -2004,7 +2060,11 @@ async function runCommand(
       );
       return connected.exitCode;
     }
-    const outcome = await connected.reader<unknown>("/api/v2/user-profile");
+    const profileUser = userIdParameters(userId);
+    const outcome =
+      profileUser.user_id === undefined
+        ? await connected.reader<unknown>("/api/v2/user-profile")
+        : await connected.reader<unknown>("/api/v2/user-profile", profileUser);
     const value =
       outcome.state === "ok"
         ? (outcome.data as {
@@ -2109,7 +2169,7 @@ async function runCommand(
     }
     const { reader, operations } = connected;
     if (parsed.kind === "list") {
-      const document = await readMemoryPage(reader, parsed.query);
+      const document = await readMemoryPage(reader, parsed.query, userId);
       emit(
         asJson
           ? `${JSON.stringify(document)}\n`
@@ -2122,6 +2182,7 @@ async function runCommand(
         reader,
         parsed.memoryId,
         parsed.namespace,
+        userId,
       );
       emit(
         asJson
@@ -2133,8 +2194,8 @@ async function runCommand(
     if (parsed.kind === "add" || parsed.kind === "add-status") {
       const document =
         parsed.kind === "add"
-          ? await addMemory(operations, text ?? "", parsed.requestId)
-          : await readMemoryAddStatus(reader, parsed.requestId);
+          ? await addMemory(operations, text ?? "", parsed.requestId, userId)
+          : await readMemoryAddStatus(reader, parsed.requestId, userId);
       emit(
         asJson
           ? `${JSON.stringify(document)}\n`
@@ -2149,9 +2210,13 @@ async function runCommand(
             content: text,
             ...namespaceParameters(parsed.namespace),
             ...(parsed.task === undefined ? {} : { task: parsed.task }),
+            ...userIdParameters(userId),
           })
         : await operations.delete<unknown>(
-            withQuery(memoryPath, namespaceParameters(parsed.namespace)),
+            withQuery(memoryPath, {
+              ...namespaceParameters(parsed.namespace),
+              ...userIdParameters(userId),
+            }),
           );
     const document = {
       schema_version: 2,
@@ -2177,6 +2242,55 @@ async function runCommand(
     );
     return outcome.state === "committed" ? EXIT_OK : EXIT_FAILED;
   }
+  if (named === "settings") {
+    const parsed = parseSettingsCommand(arguments_.slice(1));
+    if (!parsed) return usage();
+    const connected = await connectRuntime(environment);
+    if (!connected.ok) {
+      const document: ModeDocument = {
+        schema_version: 1,
+        product: PRODUCT,
+        command: "settings mode",
+        state: "unavailable",
+        code: connected.code,
+      };
+      emit(asJson ? `${JSON.stringify(document)}\n` : renderModeText(document));
+      return connected.exitCode;
+    }
+    const document = await runModeCommand(
+      parsed,
+      connected.reader,
+      connected.operations,
+    );
+    emit(asJson ? `${JSON.stringify(document)}\n` : renderModeText(document));
+    return modeExitCode(document);
+  }
+  if (named === "evolve") {
+    const parsed = parseEvolveCommand(arguments_.slice(1));
+    if (!parsed) return usage();
+    const connected = await connectRuntime(environment);
+    if (!connected.ok) {
+      const document: EvolveDocument = {
+        schema_version: 1,
+        product: PRODUCT,
+        command: "evolve",
+        state: "unavailable",
+        code: connected.code,
+      };
+      emit(
+        asJson ? `${JSON.stringify(document)}\n` : renderEvolveText(document),
+      );
+      return connected.exitCode;
+    }
+    const document = await runEvolveCommand(parsed, {
+      userId: userIdParameters(userId).user_id,
+      reader: connected.reader,
+      operations: connected.operations,
+      sleep: controls.sleep,
+    });
+    emit(asJson ? `${JSON.stringify(document)}\n` : renderEvolveText(document));
+    return evolveExitCode(document);
+  }
   if (named === "graph-width") {
     const parsed = parseGraphWidthCommand(arguments_.slice(1));
     if (!parsed) return usage();
@@ -3675,13 +3789,21 @@ async function connectSources(
       asJson,
     );
   }
-  if (parsed.importHistory) {
+  // Another user's Hermes profile is granted and imported in one connect
+  // request: only that request can carry a user and a root.
+  if (
+    parsed.importHistory ||
+    parsed.userId !== undefined ||
+    parsed.root !== undefined
+  ) {
     const document = await connectSourcesWithImport(reader, operations, {
       sourceIds: parsed.sourceIds,
       confirmOverrides: parsed.confirmOverrides,
       ...(parsed.requestId === undefined
         ? {}
         : { requestId: parsed.requestId }),
+      ...(parsed.userId === undefined ? {} : { userId: parsed.userId }),
+      ...(parsed.root === undefined ? {} : { root: parsed.root }),
     });
     emitFlowDocument(document, asJson);
     return flowExitCode(document);
@@ -3834,6 +3956,108 @@ async function emitModelReceipt(
 
 /** Attach to a verified Runtime, telling a stopped one (exit 3) from one
  * that is up but unusable (exit 1). */
+/**
+ * `memhub profile --versions`: the long-term profile's L0 version chains,
+ * latest first, so a value that replaced a better one can be found by id.
+ */
+async function profileVersionsCommand(
+  environment: NodeJS.ProcessEnv,
+  options: { asJson: boolean; userId?: string },
+): Promise<number> {
+  const fail = (state: string, code: string, exitCode: number): number => {
+    emit(
+      options.asJson
+        ? `${JSON.stringify({ schema_version: 2, product: PRODUCT, state, code })}\n`
+        : `画像版本读取失败：${code}\n`,
+    );
+    return exitCode;
+  };
+  const connected = await connectRuntime(environment, {
+    readerTimeoutMs: 15_000,
+  });
+  if (!connected.ok) {
+    return fail("unavailable", connected.code, connected.exitCode);
+  }
+  const outcome = await connected.reader<unknown>(
+    "/api/v2/user-profile/versions",
+    userIdParameters(options.userId),
+  );
+  const chains = outcome.state === "ok" ? profileChains(outcome.data) : null;
+  if (outcome.state !== "ok" || chains === null) {
+    return fail(
+      outcome.state === "ok" ? "invalid" : outcome.state,
+      outcome.state === "ok" ? "response_invalid" : outcome.code,
+      EXIT_FAILED,
+    );
+  }
+  if (options.asJson) {
+    emit(
+      `${JSON.stringify({
+        schema_version: 2,
+        product: PRODUCT,
+        state: "ok",
+        ...(outcome.data as object),
+      })}\n`,
+    );
+    return EXIT_OK;
+  }
+  if (chains.length === 0) {
+    emit("尚未形成用户画像版本。\n");
+    return EXIT_OK;
+  }
+  const lines: string[] = [];
+  for (const chain of chains) {
+    lines.push(`画像 ${chain.memory_id}（${chain.status}）`);
+    for (const version of chain.versions) {
+      const when =
+        typeof version.gmt_created === "number"
+          ? new Date(version.gmt_created * 1000).toISOString()
+          : "时间未知";
+      lines.push(
+        `  ${version.current ? "*" : " "} ${version.memory_id}  ${when}  ${version.content}`,
+      );
+    }
+  }
+  lines.push(
+    "* 表示当前生效的版本；删除错误版本：memhub memories delete <memory-id> --yes",
+  );
+  emit(`${lines.join("\n")}\n`);
+  return EXIT_OK;
+}
+
+type ProfileChainView = {
+  memory_id: string;
+  status: string;
+  versions: {
+    memory_id: string;
+    content: string;
+    gmt_created?: number | null;
+    current: boolean;
+  }[];
+};
+
+function profileChains(value: unknown): ProfileChainView[] | null {
+  if (!value || typeof value !== "object") return null;
+  const chains = (value as { chains?: unknown }).chains;
+  if (!Array.isArray(chains)) return null;
+  const valid = chains.every(
+    (chain) =>
+      chain &&
+      typeof chain === "object" &&
+      typeof (chain as ProfileChainView).memory_id === "string" &&
+      typeof (chain as ProfileChainView).status === "string" &&
+      Array.isArray((chain as ProfileChainView).versions) &&
+      (chain as ProfileChainView).versions.every(
+        (version) =>
+          version &&
+          typeof version.memory_id === "string" &&
+          typeof version.content === "string" &&
+          typeof version.current === "boolean",
+      ),
+  );
+  return valid ? (chains as ProfileChainView[]) : null;
+}
+
 async function connectRuntime(
   environment: NodeJS.ProcessEnv,
   options?: { readerTimeoutMs?: number },
@@ -4101,6 +4325,8 @@ const SDK_MIGRATION_FAILURES: Record<string, string> = {
     "核对发现新版本还读不到你的全部记忆，没有切换数据源。",
   sdk_migration_model_profile_required:
     "迁移需要先有可用的模型配置，用来核对新版本能否读回记忆。请先用 memhub models setup 完成配置，再运行 memhub restart 重试。",
+  sdk_migration_vector_profile_missing:
+    "记忆的向量来自一个已删除的模型配置，当前配置读不回它们，所以没有迁移。请重新添加并启用那个模型配置，再运行 memhub restart 重试。",
 };
 
 const MIGRATION_ISSUES_URL =
@@ -5131,14 +5357,23 @@ type SourcesCommand =
       importHistory: boolean;
       confirmOverrides: boolean;
       requestId?: string;
+      /** `--user-id`: whose grant (Hermes only). Absent for `local`. */
+      userId?: string;
+      /** `--root`: an absolute root other than the candidate (Hermes only). */
+      root?: string;
     }
   | { kind: "all"; importHistory: boolean; confirmOverrides: boolean }
   | { kind: "receipt"; requestId: string };
 
 const SOURCE_ID_SHAPE = /^[a-z0-9][a-z0-9-]{0,127}$/;
 
+/** Only Hermes keeps one history per profile, so only Hermes may be
+ * connected for another user or from another root. */
+const MULTI_USER_SOURCES = new Set(["hermes"]);
+
 /** `sources connect <id> [<id> ...] | --all-detected [--import]
- * [--confirm-overrides] [--request-id <uuid>]` and `sources receipt <id>`.
+ * [--confirm-overrides] [--request-id <uuid>]
+ * [--user-id <id>] [--root <path>]` and `sources receipt <id>`.
  *
  * Sources may be named positionally or with repeated `--source` (the old
  * `init --source` spelling is rewritten to this command). */
@@ -5154,6 +5389,8 @@ function parseSourcesCommand(rest: readonly string[]): SourcesCommand | null {
   const sourceIds: string[] = [];
   const switches = new Set<string>();
   let requestId: string | undefined;
+  let userId: string | undefined;
+  let root: string | undefined;
   for (let index = 1; index < tokens.length; index += 1) {
     const token = tokens[index];
     if (["--import", "--confirm-overrides", "--all-detected"].includes(token)) {
@@ -5161,6 +5398,21 @@ function parseSourcesCommand(rest: readonly string[]): SourcesCommand | null {
       switches.add(token);
       continue;
     }
+    if (token === "--user-id" || token === "--root") {
+      const value = tokens[index + 1];
+      if (value === undefined || value.startsWith("--")) return null;
+      if (token === "--user-id") {
+        if (userId !== undefined || !isUserId(value)) return null;
+        userId = value;
+      } else {
+        if (root !== undefined) return null;
+        const expanded = expandRootPath(value);
+        if (expanded === null) return null;
+        root = expanded;
+      }
+      index += 1;
+      continue;
+    }
     if (token === "--source" || token === "--request-id") {
       const value = tokens[index + 1];
       if (value === undefined || value.startsWith("--")) return null;
@@ -5184,10 +5436,23 @@ function parseSourcesCommand(rest: readonly string[]): SourcesCommand | null {
   const confirmOverrides = switches.has("--confirm-overrides");
   if (
     requestId !== undefined &&
-    (!OPERATION_ID.test(requestId) || !importHistory)
+    (!OPERATION_ID.test(requestId) ||
+      !(importHistory || userId !== undefined || root !== undefined))
   ) {
     return null;
   }
+  // `--user-id local` is the default and changes nothing.
+  if (userId === DEFAULT_USER_ID) userId = undefined;
+  if (userId !== undefined || root !== undefined) {
+    // One Hermes profile per connect: the user and root name one grant.
+    if (
+      switches.has("--all-detected") ||
+      sourceIds.length !== 1 ||
+      !MULTI_USER_SOURCES.has(sourceIds[0])
+    ) {
+      return null;
+    }
+  }
   if (switches.has("--all-detected")) {
     // One names exact sources, the other means "whatever is detected".
     if (sourceIds.length > 0 || requestId !== undefined) return null;
@@ -5200,9 +5465,24 @@ function parseSourcesCommand(rest: readonly string[]): SourcesCommand | null {
     importHistory,
     confirmOverrides,
     ...(requestId === undefined ? {} : { requestId }),
+    ...(userId === undefined ? {} : { userId }),
+    ...(root === undefined ? {} : { root }),
   };
 }
 
+/** `--root` as an absolute path: a leading `~` is the home directory, a
+ * relative path is taken from the working directory. Core expands nothing. */
+function expandRootPath(value: string): string | null {
+  if (value.length === 0 || /[\0\r\n]/.test(value)) return null;
+  const expanded =
+    value === "~"
+      ? os.homedir()
+      : value.startsWith("~/")
+        ? path.join(os.homedir(), value.slice(2))
+        : value;
+  return path.resolve(expanded);
+}
+
 function parseGraphWidthCommand(
   rest: readonly string[],
 ): { decision: GraphWidthDecision; confirmed: boolean } | null {
@@ -6596,6 +6876,36 @@ const OWN_DATA_DIR_COMMANDS = new Set([
 
 /** Remove one `--data-dir <absolute path>` from argv; null when it is
  * repeated, missing its value, or not an absolute path. */
+/** Memory commands that honour the global `--user-id`. */
+const USER_ID_COMMANDS = new Set([
+  "memories",
+  "recall",
+  "profile",
+  "write",
+  "evolve",
+]);
+
+/**
+ * Take `--user-id <id>` out of the arguments. The default user (`local`) is
+ * returned as undefined, so a default request names no user at all and stays
+ * byte-identical to one sent before multi-user support.
+ */
+function liftUserIdOption(
+  arguments_: readonly string[],
+): { arguments_: string[]; userId?: string } | null {
+  const index = arguments_.indexOf("--user-id");
+  if (index === -1) return { arguments_: [...arguments_] };
+  if (arguments_.indexOf("--user-id", index + 1) !== -1) return null;
+  const value = arguments_[index + 1];
+  if (value === undefined || value.startsWith("--") || !isUserId(value)) {
+    return null;
+  }
+  return {
+    arguments_: [...arguments_.slice(0, index), ...arguments_.slice(index + 2)],
+    ...(value === DEFAULT_USER_ID ? {} : { userId: value }),
+  };
+}
+
 function liftDataDirOption(
   arguments_: readonly string[],
 ): { arguments_: string[]; dataDir?: string } | null {

memhub/apps/cli/src/command-catalog.ts

@@ -68,6 +68,12 @@ const JSON_OPTION: CommandOption = {
   flag: "--json",
   summary: "以 JSON 输出到 stdout，供脚本与 Agent 消费。",
 };
+/** Global to the memory commands: whose memories are read or written. */
+const USER_ID_OPTION: CommandOption = {
+  flag: "--user-id <id>",
+  summary:
+    "读写哪个用户的记忆，默认 local（App 用的就是 local）。1–64 位字母、数字、点、下划线或连字符；其他用户的记忆来自用 sources connect hermes --user-id 接入的 Hermes profile。",
+};
 const WATCH_OPTION: CommandOption = {
   flag: "--watch",
   summary: "串行持续刷新；Ctrl-C 只停止观察，不停止 Runtime。",
@@ -252,15 +258,25 @@ export const COMMANDS: readonly CommandDescriptor[] = [
   {
     name: "sources",
     summary:
-      "列出来源，直接接入来源，或按 request id 查接入回执。connect 不加 --import 只授权（处理保持冻结：不导入历史、不扫描新会话）；加 --import 同时导入历史并开启持续同步，返回 request id。目录来自环境变量覆盖的来源必须加 --confirm-overrides。",
+      "列出来源，直接接入来源，或按 request id 查接入回执。connect 不加 --import 只授权（处理保持冻结：不导入历史、不扫描新会话）；加 --import 同时导入历史并开启持续同步，返回 request id。目录来自环境变量覆盖的来源必须加 --confirm-overrides。Hermes 的每个 profile 是一个独立用户：用 --user-id 和 --root 接入，授权并导入历史（会自动排除 ~/.hermes/profiles/**，各用户互不读取）。",
     usage:
-      "memhub sources list [--json]\n            memhub sources connect <id> [<id> ...] [--import] [--confirm-overrides] [--request-id <uuid>] [--json]\n            memhub sources connect --all-detected [--import] [--confirm-overrides] [--json]\n            memhub sources receipt <request-id> [--json]",
+      "memhub sources list [--json]\n            memhub sources connect <id> [<id> ...] [--import] [--confirm-overrides] [--request-id <uuid>] [--json]\n            memhub sources connect hermes --user-id <id> --root <path> [--import] [--request-id <uuid>] [--json]\n            memhub sources connect --all-detected [--import] [--confirm-overrides] [--json]\n            memhub sources receipt <request-id> [--json]",
     options: [
       {
         flag: "--import",
         summary:
           "接入后立即导入历史会话并开启持续同步；不加时只授权，处理保持冻结。",
       },
+      {
+        flag: "--user-id <id>",
+        summary:
+          "只用于 hermes：这条接入属于哪个用户（默认 local）。和 --root 一起接入一个 Hermes profile，例如 --user-id wife --root ~/.hermes/profiles/wife；其他来源只能是 local。",
+      },
+      {
+        flag: "--root <path>",
+        summary:
+          "只用于 hermes：读取的根目录（~ 会展开为主目录）。默认是 Hermes 自己的目录 ~/.hermes。",
+      },
       {
         flag: "--all-detected",
         summary:
@@ -373,7 +389,7 @@ export const COMMANDS: readonly CommandDescriptor[] = [
     summary:
       "列出、搜索、查看、添加、修改或物理删除记忆（对话、工作、编码三个场景）。add 和 App 的「添加记忆」是同一个入口；delete 不可逆，需 --yes。",
     usage:
-      "memhub memories list [--scene <all|chat|working|coding>] [--source <id>] [--query <text>] [--status <state>] [--offset <n>] [--limit <n>] [--json]\n            memhub memories show <memory-id> [--kind <kind>] [--variant <variant>] [--json]\n            memhub memories add <text> [--request-id <uuid>] [--json]\n            memhub memories add --content-file <path> [--request-id <uuid>] [--json]\n            memhub memories add-status <request-id> [--json]\n            memhub memories update <memory-id> --content-file <path> [--kind <kind>] [--variant <variant>] [--task <text>] [--json]\n            memhub memories delete <memory-id> [--kind <kind>] [--variant <variant>] [--yes] [--json]",
+      "memhub memories list [--scene <all|chat|working|coding>] [--source <id>] [--query <text>] [--status <state>] [--offset <n>] [--limit <n>] [--user-id <id>] [--json]\n            memhub memories show <memory-id> [--kind <kind>] [--variant <variant>] [--user-id <id>] [--json]\n            memhub memories add <text> [--request-id <uuid>] [--user-id <id>] [--json]\n            memhub memories add --content-file <path> [--request-id <uuid>] [--user-id <id>] [--json]\n            memhub memories add-status <request-id> [--user-id <id>] [--json]\n            memhub memories update <memory-id> --content-file <path> [--kind <kind>] [--variant <variant>] [--task <text>] [--user-id <id>] [--json]\n            memhub memories delete <memory-id> [--kind <kind>] [--variant <variant>] [--yes] [--user-id <id>] [--json]",
     options: [
       {
         flag: "--scene <all|chat|working|coding>",
@@ -411,15 +427,62 @@ export const COMMANDS: readonly CommandDescriptor[] = [
           "add 的幂等键；用同一个 id 重试只会读回第一次的结果，不会重复添加。不给时自动生成并输出。",
       },
       DESTRUCTIVE_CONFIRM_OPTION,
+      USER_ID_OPTION,
       JSON_OPTION,
     ],
     exit_codes: RUNTIME_EXITS,
   },
   {
     name: "profile",
-    summary: "读取 SDK 组装后的用户长期画像（L0 + 已启用的长期画像补充）。",
-    usage: "memhub profile [--json]",
-    options: [JSON_OPTION],
+    summary:
+      "读取 SDK 组装后的用户长期画像（L0 + 已启用的长期画像补充）；--versions 列出 L0 画像的版本链（最新在前，* 为当前版本）。",
+    usage: "memhub profile [--versions] [--user-id <id>] [--json]",
+    options: [
+      {
+        flag: "--versions",
+        summary:
+          "列出 L0 基本信息的版本链（get_user_basic_info），用于找出覆盖了正确值的错误版本。",
+      },
+      USER_ID_OPTION,
+      JSON_OPTION,
+    ],
+    exit_codes: RUNTIME_EXITS,
+  },
+  {
+    name: "settings",
+    summary:
+      "查看或切换记忆模式（lite / pro / ultra）。lite 不调用模型，只在原始对话里检索；pro 由模型提取记忆（System 1）；ultra 在 pro 之上每晚自动进化（System 2，使用知识图谱）。任何模式都保留全部原始对话。切换不迁移数据：已处理的不会重做，排队中的和新的按新模式处理；需 --yes，不加时只显示计划。环境变量 ONE_MEMORY_GRAPH_STORE 会强制 ultra。",
+    usage: "memhub settings mode [lite|pro|ultra] [--yes] [--json]",
+    options: [
+      {
+        flag: "mode [lite|pro|ultra]",
+        summary: "不带模式只显示当前模式；带模式时先显示切换计划。",
+      },
+      CONFIRM_OPTION,
+      JSON_OPTION,
+    ],
+    exit_codes: RUNTIME_EXITS,
+  },
+  {
+    name: "evolve",
+    summary:
+      "立即为一个用户运行记忆进化（System 2，仅 ultra 模式）。在后台运行并返回运行 ID；--wait 等待结果。同一用户同时只运行一次；每晚的自动进化由 Runtime 按设置安排。",
+    usage:
+      "memhub evolve [--user-id <id>] [--agent-id <id>] [--max-batches <n>] [--wait] [--json]",
+    options: [
+      USER_ID_OPTION,
+      {
+        flag: "--agent-id <id>",
+        summary:
+          "进化哪个 agent 的记忆，默认 all-apps（MemHub 写入记忆用的 agent）。",
+      },
+      {
+        flag: "--max-batches <n>",
+        summary: "每轮最多处理的批次数（1–1000），默认由 SDK 决定。",
+      },
+      { flag: "--wait", summary: "等待这次进化结束再返回。" },
+      JSON_OPTION,
+    ],
     exit_codes: RUNTIME_EXITS,
   },
   {
@@ -475,7 +538,7 @@ export const COMMANDS: readonly CommandDescriptor[] = [
     summary:
       "从原始对话与已沉淀记忆统一检索上下文；本地只读，不调用模型、不触发补索引。文本输出标明每条是「原始对话摘录」还是「记忆」（JSON 里是 kind: raw|memory）。",
     usage:
-      "memhub recall <query> [--kind <raw|memory>] [--source <id>] [--limit <n>] [--character-budget <n>] [--timeout-ms <n>] [--json]\n            memhub recall --query-file <path> [...]\n            memhub recall --stdin [...]",
+      "memhub recall <query> [--kind <raw|memory>] [--source <id>] [--limit <n>] [--character-budget <n>] [--timeout-ms <n>] [--user-id <id>] [--json]\n            memhub recall --query-file <path> [...]\n            memhub recall --stdin [...]",
     options: [
       {
         flag: "--query-file <path>",
@@ -502,6 +565,7 @@ export const COMMANDS: readonly CommandDescriptor[] = [
         flag: "--timeout-ms <n>",
         summary: "Core 查询时限 50..30000 毫秒，默认 2000。",
       },
+      USER_ID_OPTION,
       JSON_OPTION,
     ],
     exit_codes: RUNTIME_EXITS,
@@ -650,7 +714,7 @@ export const COMMANDS: readonly CommandDescriptor[] = [
     summary:
       "提交一段 OpenAI 格式的对话（messages）到写入队列，异步加工成记忆；或按 request id 查询状态。request id 是幂等键：不给时自动生成并输出，重试请带上同一个。新增一条结论式记忆请用 memories add。",
     usage:
-      "memhub write submit --messages-file <path> --session <id> [--source <id>] [--request-id <id>] [--memory-at <RFC3339>] [--live] [--json]\n            memhub write status <request-id> [--json]",
+      "memhub write submit --messages-file <path> --session <id> [--source <id>] [--request-id <id>] [--memory-at <RFC3339>] [--live] [--user-id <id>] [--json]\n            memhub write status <request-id> [--user-id <id>] [--json]",
     options: [
       {
         flag: "--messages-file <path>",
@@ -666,6 +730,7 @@ export const COMMANDS: readonly CommandDescriptor[] = [
       },
       { flag: "--memory-at <RFC3339>", summary: "对话发生时间，必须带时区。" },
       { flag: "--live", summary: "优先处理。" },
+      USER_ID_OPTION,
       JSON_OPTION,
     ],
     exit_codes: RUNTIME_EXITS,

memhub/apps/cli/src/command-contract.test.ts

@@ -76,6 +76,8 @@ const COMMANDS = [
   "models",
   "memories",
   "profile",
+  "settings",
+  "evolve",
   "graph-width",
   "host",
   "recall",

memhub/apps/cli/src/context-search.ts

@@ -10,6 +10,8 @@ export interface ContextSearchInput {
   limit: number;
   character_budget: number;
   timeout_ms: number;
+  /** Whose memories to search. Omitted for the default user (`local`). */
+  user_id?: string;
 }
 
 /** The CLI's JSON document mirrors Core's SDK-native context schema v2. */

memhub/apps/cli/src/core-client.ts

@@ -12,11 +12,18 @@ const READ_ONLY_PATHS = [
   "/api/v1/live-sync",
   "/api/v1/memory-processing",
   "/api/v1/model-profiles",
+  "/api/v1/source-grants",
   "/api/v1/sources",
   "/api/v2/home-summary",
   "/api/v2/memory-library",
   "/api/v2/user-profile",
+  "/api/v2/user-profile/versions",
+  "/api/v1/settings/memory-mode",
+  "/api/v2/evolve",
 ] as const;
+/** One evolve run's receipt (`memhub evolve --wait`). */
+const EVOLVE_RUN =
+  /^\/api\/v2\/evolve\/[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$/;
 const SDK_MEMORY_DETAIL = /^\/api\/v2\/memories\/[A-Za-z0-9][\w.-]{0,127}$/;
 /** Status of one `POST /api/v2/memories` addition, by its request id. */
 const MEMORY_ADDITION =
@@ -219,6 +226,7 @@ export function isReadOnlyPath(path: string): boolean {
   if (IMPORT_JOB.test(path)) return true;
   if (OPERATION_RECEIPT.test(path)) return true;
   if (MEMORY_ADDITION.test(path)) return true;
+  if (EVOLVE_RUN.test(path)) return true;
   if (!SDK_MEMORY_DETAIL.test(path)) return false;
   const segment = path.slice("/api/v2/memories/".length);
   return !(RESERVED_MEMORY_SEGMENTS as readonly string[]).includes(segment);
@@ -310,6 +318,8 @@ export async function submitMemoryWrite(
     messages: readonly OpenAIMessage[];
     memory_at?: string;
     priority?: "live" | "normal";
+    /** Whose memory this is. Omitted for the default user (`local`). */
+    user_id?: string;
   },
 ): Promise<MemoryWriteOutcome> {
   let response: { status: number; data: unknown };
@@ -327,6 +337,9 @@ export async function submitMemoryWrite(
         ...(request.priority === undefined
           ? {}
           : { priority: request.priority }),
+        ...(request.user_id === undefined || request.user_id === "local"
+          ? {}
+          : { user_id: request.user_id }),
       },
       {
         headers: { Authorization: `Bearer ${attachment.apiToken}` },
@@ -469,11 +482,16 @@ export async function deleteModelProfile(
 export async function readMemoryWrite(
   attachment: RuntimeAttachmentLike,
   requestId: string,
+  /** A receipt belongs to the user that submitted it; omit for `local`. */
+  userId?: string,
 ): Promise<MemoryWriteReceipt | null> {
   try {
     const response = await axios.get<MemoryWriteReceipt>(
       `${baseUrlOf(attachment)}/api/v2/memory-writes/${encodeURIComponent(requestId)}`,
       {
+        ...(userId === undefined || userId === "local"
+          ? {}
+          : { params: { user_id: userId } }),
         headers: { Authorization: `Bearer ${attachment.apiToken}` },
         timeout: 10_000,
         proxy: false,

memhub/apps/cli/src/core-operations.ts

@@ -6,9 +6,10 @@ const WRITE_PATHS = [
 ] as const;
 /** `POST /api/v2/memories` adds one memory (the app's 添加记忆). */
 const MEMORY_ADD_PATH = "/api/v2/memories";
-/** One SDK memory; delete names its namespace in the query string. */
+/** One SDK memory; delete names its namespace, and a non-default user, in
+ * the query string. */
 const SDK_MEMORY_PATH =
-  /^\/api\/v2\/memories\/[A-Za-z0-9][\w.-]{0,127}(?:\?kind=(?:chat|experience|fact)(?:&variant=(?:working|coding))?)?$/;
+  /^\/api\/v2\/memories\/[A-Za-z0-9][\w.-]{0,127}(?:\?(?:kind=(?:chat|experience|fact)(?:&variant=(?:working|coding))?(?:&user_id=[A-Za-z0-9._-]{1,64})?|user_id=[A-Za-z0-9._-]{1,64}))?$/;
 const MODEL_PROBE_PATH =
   /^\/api\/v1\/model-profiles\/[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}\/probe$/;
 const MODEL_PROFILE_PATH =
@@ -25,6 +26,9 @@ const SOURCE_IMPORT_JOB_PATH = "/api/v1/import-jobs";
 const IMPORT_JOB_ACTION_PATH =
   /^\/api\/v1\/import-jobs\/[A-Za-z0-9][\w.-]{0,127}\/(?:pause|resume|cancel)$/;
 const GRAPH_WIDTH_PATH = "/api/v1/graph-width";
+/** `memhub evolve` starts a run; `memhub settings mode <m> --yes` sets the mode. */
+const EVOLVE_PATH = "/api/v2/evolve";
+const MEMORY_MODE_PATH = "/api/v1/settings/memory-mode";
 const ONBOARDING_COMPLETE_PATH = "/api/v1/onboarding/complete";
 const MAX_BYTES = 4 * 1024 * 1024;
 
@@ -96,6 +100,7 @@ export function createCoreOperations(attachment: {
         !SOURCE_AUTHORIZE_PATH.test(path) &&
         !SOURCE_REVOKE_PATH.test(path) &&
         path !== GRAPH_WIDTH_PATH &&
+        path !== EVOLVE_PATH &&
         path !== ONBOARDING_COMPLETE_PATH &&
         !MODEL_PROFILE_ACTIVATE.test(path)
       ) {
@@ -131,7 +136,10 @@ export function createCoreOperations(attachment: {
       body: Record<string, unknown>,
       signal?: AbortSignal,
     ) {
-      if (!SDK_MEMORY_PATH.test(path) || path.includes("?")) {
+      if (
+        path !== MEMORY_MODE_PATH &&
+        (!SDK_MEMORY_PATH.test(path) || path.includes("?"))
+      ) {
         throw new Error("core_operation_path_not_allowed");
       }
       let response: { status: number; data: unknown };

memhub/apps/cli/src/headless-flows.ts

@@ -110,9 +110,18 @@ export async function connectSourcesWithImport(
     sourceIds: readonly string[];
     confirmOverrides: boolean;
     requestId?: string;
+    /** Whose grant this is. Omitted for the default user (`local`). */
+    userId?: string;
+    /** An absolute root other than the source's candidate (a Hermes
+     * profile). Core reads its identity itself, so the source's candidate
+     * root id and revision are not sent for it. */
+    root?: string;
   },
 ): Promise<FlowDocument> {
   const sourceIds = input.sourceIds;
+  // A second user's grant is not a first run, and a custom root has no
+  // candidate the app showed: both are a plain connect.
+  const profile = input.root !== undefined || input.userId !== undefined;
   if (
     sourceIds.length < 1 ||
     sourceIds.length > 32 ||
@@ -134,9 +143,14 @@ export async function connectSourcesWithImport(
     return invalid("source_unavailable");
   }
   const items = selected.map(([candidate]) => candidate);
-  const needsOverride = items
-    .filter((item) => item.requires_override_confirmation)
-    .map((item) => item.source_id);
+  // A custom root is the caller's own choice, not the environment override
+  // the confirmation exists for.
+  const needsOverride =
+    input.root !== undefined
+      ? []
+      : items
+          .filter((item) => item.requires_override_confirmation)
+          .map((item) => item.source_id);
   if (needsOverride.length > 0 && !input.confirmOverrides) {
     return {
       schema_version: 1,
@@ -154,18 +168,22 @@ export async function connectSourcesWithImport(
     request = buildSourceImportRequest({
       sourceIds: [...sourceIds],
       sourceRoots: Object.fromEntries(
-        items.map((item) => [item.source_id, item.candidate_root_id]),
+        items.flatMap((item) =>
+          input.root === undefined
+            ? [[item.source_id, item.candidate_root_id]]
+            : [],
+        ),
       ),
       sourceCandidateRevisions: Object.fromEntries(
         items.flatMap((item) =>
-          item.candidate_revision === undefined
+          item.candidate_revision === undefined || input.root !== undefined
             ? []
             : [[item.source_id, item.candidate_revision]],
         ),
       ),
       confirmEnvironmentOverrides: needsOverride.length > 0,
       fullRescan: false,
-      onboarding: true,
+      onboarding: !profile,
       enableSync: true,
       requestId,
     });
@@ -175,7 +193,11 @@ export async function connectSourcesWithImport(
     }
     throw error;
   }
-  const result = await operations.post<unknown>(request.path, request.body);
+  const result = await operations.post<unknown>(request.path, {
+    ...request.body,
+    ...(input.userId === undefined ? {} : { user_id: input.userId }),
+    ...(input.root === undefined ? {} : { root: input.root }),
+  });
   if (result.state === "rejected") {
     return rejected(requestId, result.code);
   }

memhub/apps/cli/src/log-report.ts

@@ -107,7 +107,11 @@ export async function readLogReport(
   query: LogQuery,
 ): Promise<{
   events: DiagnosticEventV1[];
+  /** Lines that are not JSON: the only ones `memhub logs` calls damaged. */
   invalid_lines: number;
+  /** Whole JSON lines the reader leaves out, such as an older build's
+   * records. Nothing is wrong with them, so the text output says nothing. */
+  unrecognized_lines: number;
   failure?: "log_directory_unreadable";
 }> {
   // `readDiagnosticEvents` treats an unreadable logs path as "no events". That
@@ -119,6 +123,7 @@ export async function readLogReport(
     return {
       events: [],
       invalid_lines: 0,
+      unrecognized_lines: 0,
       failure: "log_directory_unreadable",
     };
   }
@@ -129,6 +134,7 @@ export async function readLogReport(
     return {
       events: [],
       invalid_lines: 0,
+      unrecognized_lines: 0,
       failure: "log_directory_unreadable",
     };
   }

memhub/apps/cli/src/memory-mode-cli.test.ts

@@ -0,0 +1,309 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+import { run } from "./cli.js";
+import { modeEffects, parseSettingsCommand } from "./memory-mode-flow.js";
+import { composeStatusReport, renderStatusText } from "./status-report.js";
+
+const transport = vi.hoisted(() => ({
+  get: vi.fn(),
+  post: vi.fn(),
+  put: vi.fn(),
+  patch: vi.fn(),
+  delete: vi.fn(),
+}));
+vi.mock("axios", () => ({
+  default: {
+    create: vi.fn(() => transport),
+    get: vi.fn(),
+    post: vi.fn(),
+    isAxiosError: () => false,
+  },
+}));
+
+const runtime = vi.hoisted(() => ({
+  inspectRuntimeStatus: vi.fn(),
+  discoverRuntimeAttachment: vi.fn(),
+}));
+vi.mock("@one-memory/runtime-control", async (importOriginal) => ({
+  ...(await importOriginal<typeof import("@one-memory/runtime-control")>()),
+  ...runtime,
+}));
+
+const RUN_ID = "11111111-2222-4333-8444-555555555555";
+
+function modeState(mode: string, configured = mode) {
+  return {
+    mode,
+    configured_mode: configured,
+    forced_by_env: false,
+    modes: ["lite", "pro", "ultra"],
+    graph_build: { state: "idle", users: {} },
+  };
+}
+
+function receipt(state: string, extra: Record<string, unknown> = {}) {
+  return {
+    run_id: RUN_ID,
+    user_id: "wife",
+    agent_id: "all-apps",
+    state,
+    trigger: "manual",
+    started_at: "2026-11-01T03:00:00Z",
+    rounds: state === "running" ? 0 : 2,
+    tasks_processed: state === "running" ? 0 : 5,
+    batch_failures: 0,
+    capped: false,
+    elapsed_ms: 10,
+    results: { schemas: ["never printed"] },
+    ...extra,
+  };
+}
+
+beforeEach(() => {
+  vi.clearAllMocks();
+  for (const method of Object.values(transport)) method.mockReset();
+  runtime.inspectRuntimeStatus.mockResolvedValue({
+    state: "running",
+    code: "runtime_verified",
+    instanceId: "instance-1",
+    host: "127.0.0.1",
+    port: 8765,
+    version: "1.2.3",
+  });
+  runtime.discoverRuntimeAttachment.mockResolvedValue({
+    instanceId: "instance-1",
+    apiToken: "synthetic-runtime-token",
+    host: "127.0.0.1",
+    port: 8765,
+    servicePid: 4242,
+    health: {
+      status: "ok",
+      service: "one-memory-core",
+      version: "1.2.3",
+      instance_id: "instance-1",
+    },
+  });
+  transport.get.mockImplementation(async (url: string) =>
+    url === "/api/v1/settings/memory-mode"
+      ? { status: 200, data: modeState("pro") }
+      : { status: 404, data: {} },
+  );
+});
+
+describe("memhub settings mode", () => {
+  it("shows the current mode without changing it", async () => {
+    const result = await capturedRun(["settings", "mode", "--json"]);
+    expect(result.code).toBe(0);
+    expect(JSON.parse(result.stdout).mode.mode).toBe("pro");
+    expect(transport.put).not.toHaveBeenCalled();
+  });
+
+  it("prints the plan and changes nothing without --yes", async () => {
+    const result = await capturedRun(["settings", "mode", "lite"]);
+    expect(result.code).toBe(0);
+    expect(transport.put).not.toHaveBeenCalled();
+    expect(result.stdout).toContain("更高模式生成的记忆仍保存着");
+    expect(result.stdout).toContain("memhub settings mode lite --yes");
+    const json = JSON.parse(
+      (await capturedRun(["settings", "mode", "ultra", "--json"])).stdout,
+    );
+    expect(json.state).toBe("preview");
+    expect(json.applied).toBe(false);
+    expect(json.effects).toEqual(["existing_memories_added_to_graph"]);
+  });
+
+  it("changes the mode with --yes and reports what changed", async () => {
+    transport.put.mockResolvedValue({
+      status: 200,
+      data: {
+        ...modeState("lite"),
+        previous_mode: "pro",
+        effects: ["higher_mode_memories_hidden"],
+      },
+    });
+    const result = await capturedRun(["settings", "mode", "lite", "--yes"]);
+    expect(result.code).toBe(0);
+    expect(transport.put).toHaveBeenCalledWith(
+      "/api/v1/settings/memory-mode",
+      { mode: "lite" },
+      undefined,
+    );
+    expect(result.stdout).toContain("已切换");
+    expect(result.stdout).toContain("原始对话：全部保留");
+  });
+
+  it("does nothing for the mode already set, and refuses bad input", async () => {
+    const same = await capturedRun(["settings", "mode", "pro", "--yes"]);
+    expect(same.code).toBe(0);
+    expect(same.stdout).toContain("已经是这个模式");
+    expect(transport.put).not.toHaveBeenCalled();
+    for (const bad of [
+      ["settings"],
+      ["settings", "mode", "max"],
+      ["settings", "mode", "--yes"],
+      ["settings", "mode", "lite", "pro"],
+      ["settings", "language", "en"],
+    ]) {
+      expect((await capturedRun(bad)).code, bad.join(" ")).toBe(2);
+    }
+  });
+
+  it("plans the same effects Core reports", () => {
+    expect(modeEffects("lite", "pro")).toEqual([
+      "earlier_raw_not_extracted",
+      "lite_search_stops",
+    ]);
+    expect(modeEffects("ultra", "lite")).toEqual([
+      "higher_mode_memories_hidden",
+      "evolve_stopped",
+    ]);
+    expect(modeEffects("pro", "pro")).toEqual([]);
+    expect(parseSettingsCommand(["mode", "ultra", "--yes"])).toEqual({
+      kind: "set",
+      mode: "ultra",
+      confirmed: true,
+    });
+  });
+});
+
+describe("memhub evolve", () => {
+  it("starts a run for the named user and returns at once", async () => {
+    transport.post.mockResolvedValue({ status: 202, data: receipt("running") });
+    const result = await capturedRun([
+      "evolve",
+      "--user-id",
+      "wife",
+      "--agent-id",
+      "all-apps",
+      "--max-batches",
+      "3",
+      "--json",
+    ]);
+    expect(result.code).toBe(0);
+    expect(transport.post).toHaveBeenCalledWith(
+      "/api/v2/evolve",
+      { user_id: "wife", agent_id: "all-apps", max_batches: 3 },
+      undefined,
+    );
+    const document = JSON.parse(result.stdout);
+    expect(document.state).toBe("running");
+    expect(document.run.results).toBeUndefined();
+  });
+
+  it("follows the run with --wait until it ends", async () => {
+    transport.post.mockResolvedValue({ status: 202, data: receipt("running") });
+    const reads = [receipt("running"), receipt("succeeded")];
+    transport.get.mockImplementation(async (url: string) =>
+      url === `/api/v2/evolve/${RUN_ID}`
+        ? { status: 200, data: reads.shift() }
+        : { status: 404, data: {} },
+    );
+    const result = await capturedRun(["evolve", "--wait"], {
+      sleep: async () => undefined,
+    });
+    expect(result.code).toBe(0);
+    expect(transport.post).toHaveBeenCalledWith(
+      "/api/v2/evolve",
+      {},
+      undefined,
+    );
+    expect(result.stdout).toContain("进化完成");
+    expect(result.stdout).toContain("2 轮");
+  });
+
+  it("names the mode when Core refuses outside ultra", async () => {
+    transport.post.mockResolvedValue({
+      status: 409,
+      data: { detail: "memory_mode_not_ultra" },
+    });
+    const result = await capturedRun(["evolve"]);
+    expect(result.code).toBe(1);
+    expect(result.stdout).toContain("memhub settings mode ultra --yes");
+    expect((await capturedRun(["evolve", "--max-batches", "0"])).code).toBe(2);
+    expect((await capturedRun(["evolve", "--now"])).code).toBe(2);
+  });
+});
+
+describe("status pipeline", () => {
+  it("reports pipeline.mode and the last evolve from memory-processing", async () => {
+    const processing = {
+      pending: 0,
+      fresh_pending: 0,
+      backlog_pending: 0,
+      outside_window_pending: 0,
+      running: 0,
+      waiting_for_model: 0,
+      completed: 3,
+      failed: 0,
+      blocked_stages: [],
+      pipeline: {
+        mode: "ultra",
+        configured_mode: "ultra",
+        forced_by_env: false,
+        last_evolve_at: "2026-11-01T03:00:05+08:00",
+        graph_build: { state: "succeeded", users: {} },
+        evolve: {
+          mode: "ultra",
+          schedule: { enabled: true, time: "03:00" },
+          running: [],
+          users: {
+            local: {
+              last_run_at: "2026-11-01T03:00:05+08:00",
+              last_result: { state: "succeeded", tasks_processed: 4 },
+              last_error: null,
+            },
+          },
+        },
+      },
+    };
+    const report = await composeStatusReport({
+      dataDir: "/tmp/isolated",
+      runtime: { state: "running", code: "ok", host: "127.0.0.1", port: 1 },
+      background: null,
+      reader: (async (path: string) =>
+        path === "/api/v1/memory-processing"
+          ? { state: "ok", data: processing }
+          : { state: "unavailable", code: "x" }) as never,
+    });
+    expect(report.pipeline?.mode).toBe("ultra");
+    expect(report.pipeline?.last_evolve_at).toBe("2026-11-01T03:00:05+08:00");
+    expect(report.pipeline?.evolve_users.local?.state).toBe("succeeded");
+    const text = renderStatusText(report, "MemHub");
+    expect(text).toContain("记忆模式：ultra");
+    expect(text).toContain("每日 03:00 自动进化");
+  });
+});
+
+async function capturedRun(
+  arguments_: string[],
+  controls: Parameters<typeof run>[2] = {},
+) {
+  let stdout = "";
+  let stderr = "";
+  const stdoutSpy = vi
+    .spyOn(process.stdout, "write")
+    .mockImplementation((chunk) => {
+      stdout += String(chunk);
+      return true;
+    });
+  const stderrSpy = vi
+    .spyOn(process.stderr, "write")
+    .mockImplementation((chunk) => {
+      stderr += String(chunk);
+      return true;
+    });
+  try {
+    return {
+      code: await run(
+        arguments_,
+        { ONE_MEMORY_DATA_DIR: "/synthetic/data" },
+        controls,
+      ),
+      stdout,
+      stderr,
+    };
+  } finally {
+    stdoutSpy.mockRestore();
+    stderrSpy.mockRestore();
+  }
+}

memhub/apps/cli/src/memory-mode-flow.ts

@@ -0,0 +1,372 @@
+/**
+ * `memhub settings mode` and `memhub evolve`.
+ *
+ * The mode (lite / pro / ultra) decides what the SDK does with every
+ * conversation, and switching it migrates nothing: what one mode made is not
+ * shown by a lower one, and raw kept by lite is not extracted until it is
+ * imported again. So a change prints its plan first and needs `--yes`, like
+ * every other command whose effect a person cannot simply take back.
+ *
+ * `memhub evolve` starts System 2 (ultra only) and, with `--wait`, follows
+ * the run until it ends.
+ */
+
+import type { CoreReader } from "./core-client.js";
+import type { CoreOperations, CorePutOperations } from "./core-operations.js";
+import { PRODUCT } from "./version.js";
+
+export const MEMORY_MODES = ["lite", "pro", "ultra"] as const;
+export type MemoryMode = (typeof MEMORY_MODES)[number];
+
+export type ModeEffect =
+  | "earlier_raw_not_extracted"
+  | "higher_mode_memories_hidden"
+  | "existing_memories_added_to_graph"
+  | "evolve_stopped"
+  | "lite_search_stops"
+  | "graph_backfill_already_running";
+
+const MODE_SUMMARY: Record<MemoryMode, string> = {
+  lite: "lite：不调用模型，保留全部原始对话，只在原始对话里检索",
+  pro: "pro：模型从对话中提取记忆（System 1）",
+  ultra: "ultra：pro + 每晚自动进化（System 2，使用知识图谱）",
+};
+
+const EFFECT_TEXT: Record<ModeEffect, string> = {
+  earlier_raw_not_extracted:
+    "lite 下已处理完的对话不会再被提取；只有尚未处理的和之后的新对话会由模型整理成记忆，检索立即改为记忆检索",
+  higher_mode_memories_hidden:
+    "更高模式生成的记忆仍保存着，但在这个模式下不会显示",
+  existing_memories_added_to_graph:
+    "已有记忆会在后台加入知识图谱（每个用户一次）",
+  evolve_stopped: "每晚的自动进化（System 2）停止",
+  lite_search_stops: "lite 只在原始对话里的检索立即停止，改为检索记忆",
+  graph_backfill_already_running: "知识图谱的补建已在进行中",
+};
+
+/** The same rule Core applies (`memory_mode_effects`), for the plan. */
+export function modeEffects(
+  previous: MemoryMode,
+  next: MemoryMode,
+): ModeEffect[] {
+  if (previous === next) return [];
+  const rank: Record<MemoryMode, number> = { lite: 0, pro: 1, ultra: 2 };
+  const effects: ModeEffect[] = [];
+  if (previous === "lite") {
+    effects.push("earlier_raw_not_extracted", "lite_search_stops");
+  }
+  if (rank[next] < rank[previous]) effects.push("higher_mode_memories_hidden");
+  if (previous === "ultra") effects.push("evolve_stopped");
+  if (next === "ultra") effects.push("existing_memories_added_to_graph");
+  return effects;
+}
+
+export type ModeCommand =
+  | { kind: "show" }
+  | { kind: "set"; mode: MemoryMode; confirmed: boolean };
+
+/** `settings mode [lite|pro|ultra] [--yes]` after `settings`; null = usage. */
+export function parseSettingsCommand(
+  values: readonly string[],
+): ModeCommand | null {
+  const rest = values.filter((value) => value !== "--json");
+  if (rest[0] !== "mode") return null;
+  const tail = rest.slice(1);
+  const confirmed = tail.includes("--yes");
+  const positional = tail.filter((value) => value !== "--yes");
+  if (tail.filter((value) => value === "--yes").length > 1) return null;
+  if (positional.length === 0) return confirmed ? null : { kind: "show" };
+  if (positional.length !== 1) return null;
+  const mode = positional[0] as MemoryMode;
+  if (!(MEMORY_MODES as readonly string[]).includes(mode)) return null;
+  return { kind: "set", mode, confirmed };
+}
+
+export interface ModeState {
+  mode: MemoryMode;
+  configured_mode: MemoryMode;
+  forced_by_env: boolean;
+  graph_build?: { state?: string } | null;
+  previous_mode?: MemoryMode;
+  effects?: ModeEffect[];
+}
+
+function modeState(value: unknown): ModeState | null {
+  if (!value || typeof value !== "object") return null;
+  const item = value as Record<string, unknown>;
+  if (
+    !(MEMORY_MODES as readonly unknown[]).includes(item.mode) ||
+    !(MEMORY_MODES as readonly unknown[]).includes(item.configured_mode) ||
+    typeof item.forced_by_env !== "boolean"
+  ) {
+    return null;
+  }
+  return item as unknown as ModeState;
+}
+
+export type ModeDocument = {
+  schema_version: 1;
+  product: string;
+  command: "settings mode";
+} & (
+  | { state: "ok"; mode: ModeState }
+  | {
+      state: "preview";
+      mode: ModeState;
+      target: MemoryMode;
+      effects: ModeEffect[];
+      applied: false;
+      next_step: string;
+    }
+  | { state: "unchanged"; mode: ModeState }
+  | { state: "committed"; mode: ModeState; effects: ModeEffect[] }
+  | { state: "rejected" | "unknown" | "unavailable"; code: string }
+);
+
+const BASE = {
+  schema_version: 1 as const,
+  product: PRODUCT,
+  command: "settings mode" as const,
+};
+
+export async function runModeCommand(
+  command: ModeCommand,
+  reader: CoreReader,
+  operations: CorePutOperations,
+): Promise<ModeDocument> {
+  const read = await reader<unknown>("/api/v1/settings/memory-mode");
+  if (read.state !== "ok") {
+    return { ...BASE, state: "unavailable", code: read.code };
+  }
+  const current = modeState(read.data);
+  if (current === null) {
+    return { ...BASE, state: "unavailable", code: "response_invalid" };
+  }
+  if (command.kind === "show") return { ...BASE, state: "ok", mode: current };
+  if (current.configured_mode === command.mode) {
+    return { ...BASE, state: "unchanged", mode: current };
+  }
+  if (!command.confirmed) {
+    return {
+      ...BASE,
+      state: "preview",
+      mode: current,
+      target: command.mode,
+      effects: modeEffects(current.mode, command.mode),
+      applied: false,
+      next_step: `${PRODUCT} settings mode ${command.mode} --yes`,
+    };
+  }
+  const result = await operations.put<unknown>("/api/v1/settings/memory-mode", {
+    mode: command.mode,
+  });
+  if (result.state !== "committed") {
+    return { ...BASE, state: result.state, code: result.code };
+  }
+  const changed = modeState(result.data);
+  if (changed === null) {
+    return { ...BASE, state: "unknown", code: "response_invalid" };
+  }
+  return {
+    ...BASE,
+    state: "committed",
+    mode: changed,
+    effects: changed.effects ?? [],
+  };
+}
+
+export function modeExitCode(document: ModeDocument): number {
+  return ["ok", "preview", "unchanged", "committed"].includes(document.state)
+    ? 0
+    : 1;
+}
+
+export function renderModeText(document: ModeDocument): string {
+  if ("code" in document) {
+    return `记忆模式未能${document.state === "unavailable" ? "读取" : "修改"}：${document.code}\n`;
+  }
+  const mode = document.mode;
+  const lines = [`当前模式：${MODE_SUMMARY[mode.mode]}`];
+  if (mode.forced_by_env) {
+    lines.push(
+      `设置里是 ${mode.configured_mode}；环境变量 ONE_MEMORY_GRAPH_STORE 强制使用 ultra。`,
+    );
+  }
+  lines.push("原始对话：全部保留（任何模式都不删除）");
+  if (document.state === "unchanged") {
+    lines.push("已经是这个模式，未做改动。");
+  }
+  if (document.state === "preview") {
+    lines.push(`将切换到：${MODE_SUMMARY[document.target]}`);
+    for (const effect of document.effects)
+      lines.push(`- ${EFFECT_TEXT[effect]}`);
+    lines.push(`未执行任何改动。确认后运行：${document.next_step}`);
+  }
+  if (document.state === "committed") {
+    lines.push("已切换。");
+    for (const effect of document.effects)
+      lines.push(`- ${EFFECT_TEXT[effect]}`);
+  }
+  return `${lines.join("\n")}\n`;
+}
+
+export type EvolveCommand = {
+  agentId?: string;
+  maxBatches?: number;
+  wait: boolean;
+};
+
+/** `evolve [--agent-id <id>] [--max-batches <n>] [--wait]`; null = usage. */
+export function parseEvolveCommand(
+  values: readonly string[],
+): EvolveCommand | null {
+  const parsed: EvolveCommand = { wait: false };
+  const rest = values.filter((value) => value !== "--json");
+  for (let index = 0; index < rest.length; index += 1) {
+    const flag = rest[index];
+    if (flag === "--wait" && !parsed.wait) {
+      parsed.wait = true;
+    } else if (flag === "--agent-id" && parsed.agentId === undefined) {
+      const value = rest[index + 1];
+      if (!value || !/^[A-Za-z0-9._:-]{1,128}$/.test(value)) return null;
+      parsed.agentId = value;
+      index += 1;
+    } else if (flag === "--max-batches" && parsed.maxBatches === undefined) {
+      const value = Number(rest[index + 1]);
+      if (!Number.isInteger(value) || value < 1 || value > 1000) return null;
+      parsed.maxBatches = value;
+      index += 1;
+    } else {
+      return null;
+    }
+  }
+  return parsed;
+}
+
+export type EvolveDocument = {
+  schema_version: 1;
+  product: string;
+  command: "evolve";
+} & (
+  | { state: "running" | "succeeded" | "failed"; run: Record<string, unknown> }
+  | { state: "rejected" | "unknown" | "unavailable"; code: string }
+);
+
+const RUN_ID = /^[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}$/;
+
+function evolveRun(value: unknown): Record<string, unknown> | null {
+  if (!value || typeof value !== "object") return null;
+  const run = value as Record<string, unknown>;
+  if (
+    typeof run.run_id !== "string" ||
+    !RUN_ID.test(run.run_id) ||
+    !["running", "succeeded", "failed"].includes(String(run.state))
+  ) {
+    return null;
+  }
+  return run;
+}
+
+export async function runEvolveCommand(
+  command: EvolveCommand,
+  options: {
+    userId?: string;
+    reader: CoreReader;
+    operations: CoreOperations;
+    sleep?: (ms: number) => Promise<void>;
+    pollMs?: number;
+  },
+): Promise<EvolveDocument> {
+  const base = {
+    schema_version: 1 as const,
+    product: PRODUCT,
+    command: "evolve" as const,
+  };
+  const body: Record<string, unknown> = {};
+  if (options.userId !== undefined) body.user_id = options.userId;
+  if (command.agentId !== undefined) body.agent_id = command.agentId;
+  if (command.maxBatches !== undefined) body.max_batches = command.maxBatches;
+  const started = await options.operations.post<unknown>(
+    "/api/v2/evolve",
+    body,
+  );
+  if (started.state !== "committed") {
+    return { ...base, state: started.state, code: started.code };
+  }
+  let run = evolveRun(started.data);
+  if (run === null)
+    return { ...base, state: "unknown", code: "response_invalid" };
+  const sleep =
+    options.sleep ??
+    ((ms: number) => new Promise((done) => setTimeout(done, ms)));
+  while (command.wait && run.state === "running") {
+    await sleep(options.pollMs ?? 2_000);
+    const read = await options.reader<unknown>(
+      `/api/v2/evolve/${run.run_id}`,
+      options.userId === undefined ? {} : { user_id: options.userId },
+    );
+    if (read.state !== "ok") {
+      return { ...base, state: "unavailable", code: read.code };
+    }
+    const next = evolveRun(read.data);
+    if (next === null)
+      return { ...base, state: "unknown", code: "response_invalid" };
+    run = next;
+  }
+  return {
+    ...base,
+    state: run.state as "running" | "succeeded" | "failed",
+    run: projectRun(run),
+  };
+}
+
+/** Counts and codes; SDK result bodies stay in the Core receipt. */
+function projectRun(run: Record<string, unknown>): Record<string, unknown> {
+  const keep = [
+    "run_id",
+    "user_id",
+    "agent_id",
+    "state",
+    "trigger",
+    "started_at",
+    "finished_at",
+    "rounds",
+    "tasks_processed",
+    "batch_failures",
+    "cycle_status",
+    "should_continue",
+    "continuation_reason",
+    "remaining_clusters",
+    "remaining_actionable_clusters",
+    "remaining_fresh_facts",
+    "capped",
+    "elapsed_ms",
+    "error_code",
+    "failure_kind",
+  ];
+  return Object.fromEntries(
+    keep.filter((key) => key in run).map((key) => [key, run[key]]),
+  );
+}
+
+export function evolveExitCode(document: EvolveDocument): number {
+  return document.state === "running" || document.state === "succeeded" ? 0 : 1;
+}
+
+export function renderEvolveText(document: EvolveDocument): string {
+  if ("code" in document) {
+    return document.code === "memory_mode_not_ultra"
+      ? "进化只在 ultra 模式下可用。切换：memhub settings mode ultra --yes\n"
+      : `进化未能开始：${document.code}\n`;
+  }
+  const run = document.run;
+  if (document.state === "running") {
+    return `进化已开始（${run.run_id}），在后台运行；加 --wait 等待结果。\n`;
+  }
+  if (document.state === "failed") {
+    return `进化失败（${run.run_id}）：${String(run.error_code ?? "unknown")}\n`;
+  }
+  return `进化完成（${run.run_id}）：${run.rounds} 轮，处理 ${run.tasks_processed} 项${
+    run.cycle_status ? `，周期 ${run.cycle_status}` : ""
+  }${run.capped ? "，已到轮数上限，可再次运行继续" : ""}\n`;
+}

memhub/apps/cli/src/memory-report.ts

@@ -14,6 +14,25 @@ import { type CoreReader, RESERVED_MEMORY_SEGMENTS } from "./core-client.js";
 import type { CoreOperations } from "./core-operations.js";
 import { PRODUCT } from "./version.js";
 
+/** A MemHub user id, the same shape Core accepts. `local` is the default. */
+const USER_ID = /^[A-Za-z0-9._-]{1,64}$/;
+export const DEFAULT_USER_ID = "local";
+
+export function isUserId(value: string): boolean {
+  return USER_ID.test(value);
+}
+
+/**
+ * The `user_id` to send, or nothing. The default user is never named on the
+ * wire, so a request without `--user-id` stays exactly what it was before
+ * multi-user support and still works against an older Core.
+ */
+export function userIdParameters(userId?: string): { user_id?: string } {
+  return userId === undefined || userId === DEFAULT_USER_ID
+    ? {}
+    : { user_id: userId };
+}
+
 /** Output budget: a long memory is shortened visibly, never dropped silently. */
 const CONTENT_BUDGET = 240;
 
@@ -101,10 +120,12 @@ export interface MemoryDetailDocument {
 export async function readMemoryPage(
   reader: CoreReader,
   query: MemoryQuery,
+  userId?: string,
 ): Promise<MemoryPageDocument> {
   const parameters: Record<string, string | number> = {
     offset: query.offset,
     limit: query.limit,
+    ...userIdParameters(userId),
   };
   if (query.status !== undefined) parameters.statuses = query.status;
   if (query.source !== undefined) parameters.source = query.source;
@@ -156,8 +177,12 @@ export async function readMemoryDetail(
   reader: CoreReader,
   memoryId: string,
   namespace: MemoryNamespace = {},
+  userId?: string,
 ): Promise<MemoryDetailDocument> {
-  const parameters = namespaceParameters(namespace);
+  const parameters = {
+    ...namespaceParameters(namespace),
+    ...userIdParameters(userId),
+  };
   const outcome = await reader<unknown>(
     `/api/v2/memories/${memoryId}`,
     Object.keys(parameters).length > 0 ? parameters : undefined,
@@ -236,6 +261,7 @@ export async function addMemory(
   operations: CoreOperations,
   text: string,
   requestId: string = randomUUID(),
+  userId?: string,
 ): Promise<MemoryAddDocument> {
   const base = {
     schema_version: 2 as const,
@@ -259,7 +285,10 @@ export async function addMemory(
     }
     throw error;
   }
-  const result = await operations.post<unknown>(request.path, request.body);
+  const result = await operations.post<unknown>(request.path, {
+    ...request.body,
+    ...userIdParameters(userId),
+  });
   if (result.state !== "committed") {
     return { ...base, state: result.state, code: result.code };
   }
@@ -273,6 +302,7 @@ export async function addMemory(
 export async function readMemoryAddStatus(
   reader: CoreReader,
   requestId: string,
+  userId?: string,
 ): Promise<MemoryAddDocument> {
   const base = {
     schema_version: 2 as const,
@@ -283,7 +313,11 @@ export async function readMemoryAddStatus(
   if (!UUID.test(requestId)) {
     return { ...base, state: "invalid", code: "invalid_request_id" };
   }
-  const outcome = await reader<unknown>(memoryAddStatusPath(requestId));
+  const user = userIdParameters(userId);
+  const outcome =
+    user.user_id === undefined
+      ? await reader<unknown>(memoryAddStatusPath(requestId))
+      : await reader<unknown>(memoryAddStatusPath(requestId), user);
   if (outcome.state === "not_found") {
     return { ...base, state: "not_found", code: "memory_add_not_found" };
   }

memhub/apps/cli/src/profile-versions-cli.test.ts

@@ -0,0 +1,166 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+import { run } from "./cli.js";
+
+const transport = vi.hoisted(() => ({
+  get: vi.fn(),
+  post: vi.fn(),
+  put: vi.fn(),
+  patch: vi.fn(),
+  delete: vi.fn(),
+}));
+vi.mock("axios", () => ({
+  default: {
+    create: vi.fn(() => transport),
+    get: vi.fn(),
+    post: vi.fn(),
+    isAxiosError: () => false,
+  },
+}));
+
+const runtime = vi.hoisted(() => ({
+  inspectRuntimeStatus: vi.fn(),
+  discoverRuntimeAttachment: vi.fn(),
+}));
+vi.mock("@one-memory/runtime-control", async (importOriginal) => ({
+  ...(await importOriginal<typeof import("@one-memory/runtime-control")>()),
+  ...runtime,
+}));
+
+const VERSIONS = {
+  user_id: "wife",
+  request_id: "r-versions",
+  elapsed_ms: 2,
+  chains: [
+    {
+      memory_id: "l0-head",
+      content: "The user's name is Fei.",
+      status: "active",
+      memory_at: null,
+      gmt_created: 1_790_000_100,
+      basic_info: { name: "Fei" },
+      versions: [
+        {
+          memory_id: "l0-head",
+          content: "The user's name is Fei.",
+          memory_at: null,
+          gmt_created: 1_790_000_100,
+          current: true,
+        },
+        {
+          memory_id: "l0-old",
+          content: "The user's name is 费天翔.",
+          memory_at: null,
+          gmt_created: 1_790_000_000,
+          current: false,
+        },
+      ],
+    },
+  ],
+};
+
+beforeEach(() => {
+  vi.clearAllMocks();
+  for (const method of Object.values(transport)) method.mockReset();
+  runtime.inspectRuntimeStatus.mockResolvedValue({
+    state: "running",
+    code: "runtime_verified",
+    instanceId: "instance-1",
+    host: "127.0.0.1",
+    port: 8765,
+    version: "1.2.3",
+  });
+  runtime.discoverRuntimeAttachment.mockResolvedValue({
+    instanceId: "instance-1",
+    apiToken: "synthetic-runtime-token",
+    host: "127.0.0.1",
+    port: 8765,
+    servicePid: 4242,
+    health: {
+      status: "ok",
+      service: "one-memory-core",
+      version: "1.2.3",
+      instance_id: "instance-1",
+    },
+  });
+  transport.get.mockImplementation(async (url: string) =>
+    url === "/api/v2/user-profile/versions"
+      ? { status: 200, data: VERSIONS }
+      : { status: 404, data: {} },
+  );
+});
+
+describe("memhub profile --versions", () => {
+  it("reads the L0 version chain for the named user", async () => {
+    const result = await capturedRun([
+      "profile",
+      "--versions",
+      "--user-id",
+      "wife",
+      "--json",
+    ]);
+    expect(result.code).toBe(0);
+    expect(transport.get).toHaveBeenCalledWith(
+      "/api/v2/user-profile/versions",
+      { params: { user_id: "wife" } },
+    );
+    const document = JSON.parse(result.stdout);
+    expect(document.state).toBe("ok");
+    expect(document.chains[0].versions).toHaveLength(2);
+  });
+
+  it("marks the current version in text and names no user for local", async () => {
+    const result = await capturedRun(["profile", "--versions"]);
+    expect(result.code).toBe(0);
+    const [url, config] = transport.get.mock.calls[0] ?? [];
+    expect(url).toBe("/api/v2/user-profile/versions");
+    expect(config?.params?.user_id).toBeUndefined();
+    expect(result.stdout).toContain("* l0-head");
+    expect(result.stdout).toContain("  l0-old");
+    expect(result.stdout).toContain("费天翔");
+  });
+
+  it("refuses an unknown flag and a repeated --versions", async () => {
+    expect((await capturedRun(["profile", "--version-chain"])).code).toBe(2);
+    expect(
+      (await capturedRun(["profile", "--versions", "--versions"])).code,
+    ).toBe(2);
+    expect(transport.get).not.toHaveBeenCalled();
+  });
+
+  it("reports a malformed answer as invalid", async () => {
+    transport.get.mockResolvedValue({ status: 200, data: { chains: "x" } });
+    const result = await capturedRun(["profile", "--versions", "--json"]);
+    expect(result.code).toBe(1);
+    expect(JSON.parse(result.stdout).code).toBe("response_invalid");
+  });
+});
+
+async function capturedRun(arguments_: string[]) {
+  let stdout = "";
+  let stderr = "";
+  const stdoutSpy = vi
+    .spyOn(process.stdout, "write")
+    .mockImplementation((chunk) => {
+      stdout += String(chunk);
+      return true;
+    });
+  const stderrSpy = vi
+    .spyOn(process.stderr, "write")
+    .mockImplementation((chunk) => {
+      stderr += String(chunk);
+      return true;
+    });
+  try {
+    return {
+      code: await run(arguments_, {
+        ONE_MEMORY_DATA_DIR: "/synthetic/data",
+      }),
+      stdout,
+      stderr,
+    };
+  } finally {
+    stdoutSpy.mockRestore();
+    stderrSpy.mockRestore();
+  }
+}

memhub/apps/cli/src/source-progress-report.ts

@@ -14,6 +14,11 @@ import {
   validateProcessing,
   validateSources,
 } from "./core-responses.js";
+import {
+  describePipeline,
+  type PipelineSummary,
+  pipelineLine,
+} from "./status-report.js";
 import { PRODUCT } from "./version.js";
 
 type ReadState =
@@ -38,6 +43,21 @@ export interface SourcesDocument {
   state: ReadState["state"];
   code?: string;
   sources: Array<Record<string, unknown>>;
+  /** Every grant of every user (`GET /api/v1/source-grants`). Absent when
+   * Core is too old to list them; `sources` is the default user's view. */
+  grants?: SourceGrant[];
+}
+
+/** One source grant. Only Hermes has grants for users other than `local`,
+ * one per Hermes profile root. */
+export interface SourceGrant {
+  user_id: string;
+  source_id: string;
+  root_label: string | null;
+  root_path: string | null;
+  authorized: boolean;
+  revoked_at: string | null;
+  sync_enabled: boolean;
 }
 
 export interface SyncDocument {
@@ -69,6 +89,8 @@ export interface ProgressDocument {
     summary_state: ReadState;
     summary: Record<string, unknown> | null;
   };
+  /** The SDK mode and last evolve per user; absent on an older Core. */
+  pipeline?: PipelineSummary | null;
 }
 
 export async function readSources(
@@ -86,14 +108,66 @@ export async function readSources(
       sources: [],
     };
   }
+  const grants = await readSourceGrants(reader);
   return {
     schema_version: SCHEMA_VERSION,
     product: PRODUCT,
     state: "ok",
-    sources: result.data.map(projectSource),
+    // `/api/v1/sources` is the default user's view of every source.
+    sources: result.data.map((source) => ({
+      ...projectSource(source),
+      user_id: "local",
+    })),
+    ...(grants === null ? {} : { grants }),
   };
 }
 
+/** The grants of every user, or null when Core cannot list them (an older
+ * Core answers 404): the source list is then shown as before. */
+async function readSourceGrants(
+  reader: CoreReader,
+): Promise<SourceGrant[] | null> {
+  let outcome: ReadOutcome<unknown>;
+  try {
+    outcome = await reader<unknown>("/api/v1/source-grants");
+  } catch {
+    return null;
+  }
+  if (outcome.state !== "ok") return null;
+  const value = outcome.data;
+  if (!value || typeof value !== "object" || Array.isArray(value)) return null;
+  const items = (value as { grants?: unknown }).grants;
+  if (!Array.isArray(items)) return null;
+  const grants: SourceGrant[] = [];
+  const text = (candidate: unknown): string | null =>
+    typeof candidate === "string" && candidate.length <= 4096
+      ? candidate
+      : null;
+  for (const item of items) {
+    if (!item || typeof item !== "object" || Array.isArray(item)) return null;
+    const grant = item as Record<string, unknown>;
+    if (
+      typeof grant.user_id !== "string" ||
+      !/^[A-Za-z0-9._-]{1,64}$/.test(grant.user_id) ||
+      typeof grant.source_id !== "string" ||
+      !/^[a-z0-9][a-z0-9-]{0,127}$/.test(grant.source_id) ||
+      typeof grant.authorized !== "boolean"
+    ) {
+      return null;
+    }
+    grants.push({
+      user_id: grant.user_id,
+      source_id: grant.source_id,
+      root_label: text(grant.root_label),
+      root_path: text(grant.root_path),
+      authorized: grant.authorized,
+      revoked_at: text(grant.revoked_at),
+      sync_enabled: grant.sync_enabled === true,
+    });
+  }
+  return grants;
+}
+
 export async function readSync(reader: CoreReader): Promise<SyncDocument> {
   const result = validated(
     await reader<unknown>("/api/v1/live-sync"),
@@ -216,6 +290,10 @@ export async function readProgress(
       summary:
         processing.state === "ok" ? projectProcessing(processing.data) : null,
     },
+    pipeline:
+      processingRead.state === "ok"
+        ? describePipeline(processingRead.data)
+        : null,
   };
 }
 
@@ -231,6 +309,22 @@ export function renderSourcesText(document: SourcesDocument): string {
     );
   }
   if (document.sources.length === 0) lines.push("（没有来源）");
+  // The list above is the default user's (`local`). Another user's grant,
+  // a Hermes profile, is listed with its user so it is not mistaken for it.
+  const others = (document.grants ?? []).filter(
+    (grant) => grant.user_id !== "local",
+  );
+  if (others.length > 0) {
+    lines.push("其他用户的连接：");
+    for (const grant of others) {
+      const name =
+        document.sources.find((source) => source.source_id === grant.source_id)
+          ?.display_name ?? grant.source_id;
+      lines.push(
+        `- ${name}（用户 ${grant.user_id}）：${grant.authorized ? "已连接" : grant.revoked_at ? "已撤销" : "未连接"}；处理${grant.sync_enabled ? "已开启" : "已冻结"}${grant.root_path ? `；目录 ${grant.root_path}` : ""}`,
+      );
+    }
+  }
   lines.push("");
   return lines.join("\n");
 }
@@ -270,6 +364,18 @@ export function renderProgressText(document: ProgressDocument): string {
       ? `记忆整理：待处理 ${document.generation.summary.pending} · 进行中 ${document.generation.summary.running} · 等待模型 ${document.generation.summary.waiting_for_model} · 失败 ${document.generation.summary.failed}`
       : "记忆整理：暂时无法读取",
   );
+  if (document.pipeline) {
+    lines.push(`记忆模式：${pipelineLine(document.pipeline)}`);
+    for (const [userId, user] of Object.entries(
+      document.pipeline.evolve_users,
+    )) {
+      lines.push(
+        `  进化 ${userId}：${user.last_run_at ?? "尚未进化"}${
+          user.last_error ? `（失败：${user.last_error}）` : ""
+        }`,
+      );
+    }
+  }
   if (document.imports.output_truncated) {
     lines.push("这里只显示汇总；完整记录请使用 --json。");
   }

memhub/apps/cli/src/source-progress.test.ts

@@ -179,12 +179,72 @@ describe("CLI-07/10 source and synchronization reads", () => {
       sync_enabled: true,
       sync_completeness: "additions_only",
     });
-    expect(requestedPaths()).toEqual(["/api/v1/sources"]);
+    // The grants list answers 404 here, as an older Core does: the source
+    // list is shown as before, every row the default user's.
+    expect(report.sources[0].user_id).toBe("local");
+    expect(report.grants).toBeUndefined();
+    expect(requestedPaths()).toEqual([
+      "/api/v1/sources",
+      "/api/v1/source-grants",
+    ]);
     expect(requestedPaths().some((path) => path.includes("detect"))).toBe(
       false,
     );
   });
 
+  it("shows every user's grant, a Hermes profile with its user", async () => {
+    const grants = {
+      schema_version: 1,
+      grants: [
+        {
+          user_id: "local",
+          source_id: "codex",
+          root_label: ".codex",
+          root_path: "/home/synthetic/.codex",
+          authorized: true,
+          revoked_at: null,
+          sync_enabled: true,
+        },
+        {
+          user_id: "wife",
+          source_id: "hermes",
+          root_label: "wife",
+          root_path: "/home/synthetic/.hermes/profiles/wife",
+          authorized: true,
+          revoked_at: null,
+          sync_enabled: true,
+        },
+      ],
+    };
+    transport.get.mockImplementation(async (url: string) =>
+      url === "/api/v1/source-grants"
+        ? { status: 200, data: grants }
+        : (RESPONSES.get(url) ?? { status: 404, data: {} }),
+    );
+    let streams = captureStreams();
+    try {
+      expect(await run(["sources", "list", "--json"], {})).toBe(0);
+    } finally {
+      streams.restore();
+    }
+    const report = JSON.parse(streams.stdout());
+    expect(report.grants).toEqual(grants.grants);
+    expect(report.sources[0]).toMatchObject({
+      source_id: "codex",
+      user_id: "local",
+    });
+    streams = captureStreams();
+    try {
+      expect(await run(["sources", "list"], {})).toBe(0);
+    } finally {
+      streams.restore();
+    }
+    expect(streams.stdout()).toContain("其他用户的连接：");
+    expect(streams.stdout()).toContain(
+      "hermes（用户 wife）：已连接；处理已开启；目录 /home/synthetic/.hermes/profiles/wife",
+    );
+  });
+
   it("keeps degraded synchronization distinct and prints its error code", async () => {
     const streams = captureStreams();
     try {
@@ -552,6 +612,7 @@ describe("CLI-19 progress and issue reads", () => {
       requestedPaths().every((path) =>
         [
           "/api/v1/sources",
+          "/api/v1/source-grants",
           "/api/v1/live-sync",
           "/api/v1/import-jobs",
           "/api/v1/memory-processing",

memhub/apps/cli/src/status-report.ts

@@ -57,6 +57,9 @@ export interface StatusDocument {
     broker?: BackgroundHealth["broker"];
   };
   sections: Record<string, Section>;
+  /** R10: the SDK mode and System 2 state; null when the Runtime is not read
+   * or is older than the setting. */
+  pipeline?: PipelineSummary | null;
   summary: {
     state: "ok" | "partial";
     degraded: string[];
@@ -64,6 +67,88 @@ export interface StatusDocument {
   };
 }
 
+export interface PipelineSummary {
+  mode: "lite" | "pro" | "ultra";
+  configured_mode: "lite" | "pro" | "ultra";
+  forced_by_env: boolean;
+  last_evolve_at: string | null;
+  evolve_schedule: { enabled: boolean; time: string } | null;
+  evolve_running: number;
+  evolve_users: Record<
+    string,
+    {
+      last_run_at: string | null;
+      last_error: string | null;
+      state: string | null;
+    }
+  >;
+  graph_build: string | null;
+}
+
+const MODES = new Set(["lite", "pro", "ultra"]);
+
+/** Projection of `pipeline` in `/api/v1/memory-processing`; null if absent or malformed. */
+export function describePipeline(value: unknown): PipelineSummary | null {
+  if (!value || typeof value !== "object") return null;
+  const pipeline = (value as { pipeline?: unknown }).pipeline;
+  if (!pipeline || typeof pipeline !== "object") return null;
+  const item = pipeline as Record<string, unknown>;
+  if (
+    !MODES.has(String(item.mode)) ||
+    !MODES.has(String(item.configured_mode))
+  ) {
+    return null;
+  }
+  const evolve =
+    item.evolve && typeof item.evolve === "object"
+      ? (item.evolve as Record<string, unknown>)
+      : {};
+  const schedule =
+    evolve.schedule && typeof evolve.schedule === "object"
+      ? (evolve.schedule as Record<string, unknown>)
+      : null;
+  const users: PipelineSummary["evolve_users"] = {};
+  if (evolve.users && typeof evolve.users === "object") {
+    for (const [userId, raw] of Object.entries(
+      evolve.users as Record<string, unknown>,
+    )) {
+      const user =
+        raw && typeof raw === "object" ? (raw as Record<string, unknown>) : {};
+      const result =
+        user.last_result && typeof user.last_result === "object"
+          ? (user.last_result as Record<string, unknown>)
+          : {};
+      users[userId] = {
+        last_run_at:
+          typeof user.last_run_at === "string" ? user.last_run_at : null,
+        last_error:
+          typeof user.last_error === "string" ? user.last_error : null,
+        state: typeof result.state === "string" ? result.state : null,
+      };
+    }
+  }
+  const graph =
+    item.graph_build && typeof item.graph_build === "object"
+      ? (item.graph_build as Record<string, unknown>)
+      : {};
+  return {
+    mode: item.mode as PipelineSummary["mode"],
+    configured_mode: item.configured_mode as PipelineSummary["mode"],
+    forced_by_env: item.forced_by_env === true,
+    last_evolve_at:
+      typeof item.last_evolve_at === "string" ? item.last_evolve_at : null,
+    evolve_schedule:
+      schedule &&
+      typeof schedule.enabled === "boolean" &&
+      typeof schedule.time === "string"
+        ? { enabled: schedule.enabled, time: schedule.time }
+        : null,
+    evolve_running: Array.isArray(evolve.running) ? evolve.running.length : 0,
+    evolve_users: users,
+    graph_build: typeof graph.state === "string" ? graph.state : null,
+  };
+}
+
 const SECTION_NAMES = [
   "sources",
   "sync",
@@ -97,6 +182,7 @@ export async function composeStatusReport(options: {
   attachFailure?: string;
 }): Promise<StatusDocument> {
   const sections: Record<string, Section> = {};
+  let pipeline: PipelineSummary | null = null;
   if (options.reader === null) {
     const state: SectionState =
       options.attachFailure === undefined ? "not_checked" : "unavailable";
@@ -118,6 +204,8 @@ export async function composeStatusReport(options: {
       validateProcessing,
       describeProcessing,
     );
+    pipeline =
+      processing.state === "ok" ? describePipeline(processing.data) : null;
     sections.model = section(model, validateModelProfiles, describeModel);
     sections.memories = section(
       memories,
@@ -169,6 +257,7 @@ export async function composeStatusReport(options: {
     },
     bridge,
     sections,
+    pipeline,
     summary: {
       state:
         degraded.length === 0 && notChecked.length === 0 ? "ok" : "partial",
@@ -205,6 +294,7 @@ export function renderStatusText(
       `${SECTION_LABELS[name]}：${sectionLine(report.sections[name])}`,
     );
   }
+  if (report.pipeline) lines.push(`记忆模式：${pipelineLine(report.pipeline)}`);
   if (report.summary.state === "partial") {
     const notes = [
       report.summary.degraded.length > 0
@@ -220,6 +310,28 @@ export function renderStatusText(
   return lines.join("\n");
 }
 
+export function pipelineLine(pipeline: PipelineSummary): string {
+  const parts: string[] = [pipeline.mode];
+  if (pipeline.forced_by_env)
+    parts.push("（ONE_MEMORY_GRAPH_STORE 强制 ultra）");
+  if (pipeline.mode === "ultra") {
+    const schedule = pipeline.evolve_schedule;
+    parts.push(
+      schedule?.enabled
+        ? `，每日 ${schedule.time} 自动进化`
+        : "，自动进化已关闭",
+    );
+    parts.push(
+      pipeline.last_evolve_at
+        ? `，上次进化 ${pipeline.last_evolve_at}`
+        : "，尚未进化",
+    );
+    if (pipeline.evolve_running > 0)
+      parts.push(`，进行中 ${pipeline.evolve_running}`);
+  }
+  return parts.join("");
+}
+
 function runtimeLine(report: StatusDocument): string {
   if (report.runtime.state !== "running") {
     return report.runtime.state === "stopped" ? "未运行" : "暂时无法确认";

memhub/apps/cli/src/user-id-cli.test.ts

@@ -0,0 +1,405 @@
+import { mkdtemp, writeFile } from "node:fs/promises";
+import os, { tmpdir } from "node:os";
+import path from "node:path";
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+import { run } from "./cli.js";
+
+const transport = vi.hoisted(() => ({
+  get: vi.fn(),
+  post: vi.fn(),
+  put: vi.fn(),
+  patch: vi.fn(),
+  delete: vi.fn(),
+}));
+const direct = vi.hoisted(() => ({ get: vi.fn(), post: vi.fn() }));
+vi.mock("axios", () => ({
+  default: {
+    create: vi.fn(() => transport),
+    get: direct.get,
+    post: direct.post,
+    isAxiosError: () => false,
+  },
+}));
+
+const runtime = vi.hoisted(() => ({
+  inspectRuntimeStatus: vi.fn(),
+  discoverRuntimeAttachment: vi.fn(),
+}));
+vi.mock("@one-memory/runtime-control", async (importOriginal) => ({
+  ...(await importOriginal<typeof import("@one-memory/runtime-control")>()),
+  ...runtime,
+}));
+
+const REQUEST = "33333333-3333-4333-8333-333333333333";
+const HERMES = {
+  source_id: "hermes",
+  display_name: "Hermes",
+  storage_kinds: ["sqlite"],
+  available: true,
+  authorized: true,
+  detected: null,
+  root_label: ".hermes",
+  candidate_path: "~/.hermes",
+  candidate_kind: "default",
+  candidate_revision: 0,
+  candidate_root_id: "hermes-root-identity",
+  imported_messages: 0,
+  imported_conversations: 0,
+  capabilities: {
+    historical_import: "implemented",
+    incremental_import: "implemented",
+    live_sync: "implemented",
+    recall: "implemented",
+    installer: "not_implemented",
+    verification: "fixture_only",
+  },
+  sync_completeness: "additions_only",
+  location_state: "found",
+};
+const CODEX = {
+  ...HERMES,
+  source_id: "codex",
+  display_name: "Codex",
+  candidate_root_id: "codex-root-identity",
+};
+
+function addReceipt() {
+  return {
+    evidence_id: "44444444-4444-4444-8444-444444444444",
+    text: "Prefers tea",
+    scope: "all-apps",
+    created_at: "2026-10-27T00:00:00Z",
+    state: "pending",
+    task_id: null,
+    memory_ids: [],
+    error: null,
+  };
+}
+
+beforeEach(() => {
+  vi.clearAllMocks();
+  for (const method of [
+    ...Object.values(transport),
+    ...Object.values(direct),
+  ]) {
+    method.mockReset();
+  }
+  runtime.inspectRuntimeStatus.mockResolvedValue({
+    state: "running",
+    code: "runtime_verified",
+    instanceId: "instance-1",
+    host: "127.0.0.1",
+    port: 8765,
+    version: "1.2.3",
+  });
+  runtime.discoverRuntimeAttachment.mockResolvedValue({
+    instanceId: "instance-1",
+    apiToken: "synthetic-runtime-token",
+    host: "127.0.0.1",
+    port: 8765,
+    servicePid: 4242,
+    health: {
+      status: "ok",
+      service: "one-memory-core",
+      version: "1.2.3",
+      instance_id: "instance-1",
+    },
+  });
+  transport.get.mockImplementation(async (url: string) => {
+    if (url === "/api/v1/sources") {
+      return { status: 200, data: [HERMES, CODEX] };
+    }
+    if (url === "/api/v2/user-profile") {
+      return {
+        status: 200,
+        data: { request_id: "r-1", profile: null, elapsed_ms: 1 },
+      };
+    }
+    return { status: 404, data: {} };
+  });
+  transport.post.mockImplementation(
+    async (url: string, body: Record<string, unknown>) =>
+      url === "/api/v1/sources/connect"
+        ? {
+            status: 200,
+            data: {
+              request_id: body.request_id,
+              accepted_source_ids: ["hermes"],
+              failed_sources: {},
+              jobs: [],
+            },
+          }
+        : url === "/api/v2/memories"
+          ? { status: 201, data: addReceipt() }
+          : { status: 404, data: {} },
+  );
+  transport.put.mockResolvedValue({ status: 200, data: {} });
+  transport.delete.mockResolvedValue({ status: 204, data: "" });
+  direct.post.mockResolvedValue({
+    status: 202,
+    data: {
+      request_id: REQUEST,
+      state: "accepted",
+      created_at: "T1",
+      updated_at: "T1",
+    },
+  });
+  direct.get.mockResolvedValue({
+    data: {
+      request_id: REQUEST,
+      state: "accepted",
+      created_at: "T1",
+      updated_at: "T1",
+    },
+  });
+});
+
+describe("--user-id on the memory commands", () => {
+  it("passes the user as a query parameter to memory reads and deletes", async () => {
+    await capturedRun(["memories", "list", "--user-id", "wife", "--json"]);
+    expect(transport.get).toHaveBeenCalledWith("/api/v2/memory-library", {
+      params: { offset: 0, limit: 100, user_id: "wife" },
+    });
+    // Global: the flag may come before the subcommand too.
+    await capturedRun(["memories", "--user-id", "wife", "show", "m-1"]);
+    expect(transport.get).toHaveBeenCalledWith("/api/v2/memories/m-1", {
+      params: { user_id: "wife" },
+    });
+    await capturedRun(["memories", "add-status", REQUEST, "--user-id", "wife"]);
+    expect(transport.get).toHaveBeenCalledWith(
+      `/api/v2/memories/additions/${REQUEST}`,
+      { params: { user_id: "wife" } },
+    );
+    await capturedRun(["profile", "--user-id", "wife", "--json"]);
+    expect(transport.get).toHaveBeenCalledWith("/api/v2/user-profile", {
+      params: { user_id: "wife" },
+    });
+    await capturedRun([
+      "memories",
+      "delete",
+      "fact-1",
+      "--kind",
+      "fact",
+      "--yes",
+      "--user-id",
+      "wife",
+    ]);
+    expect(transport.delete).toHaveBeenCalledWith(
+      "/api/v2/memories/fact-1?kind=fact&user_id=wife",
+      undefined,
+    );
+    await capturedRun([
+      "memories",
+      "delete",
+      "m-1",
+      "--yes",
+      "--user-id",
+      "wife",
+    ]);
+    expect(transport.delete).toHaveBeenCalledWith(
+      "/api/v2/memories/m-1?user_id=wife",
+      undefined,
+    );
+    await capturedRun(["write", "status", REQUEST, "--user-id", "wife"]);
+    expect(direct.get.mock.calls[0][0]).toBe(
+      `http://127.0.0.1:8765/api/v2/memory-writes/${REQUEST}`,
+    );
+    expect(direct.get.mock.calls[0][1].params).toEqual({ user_id: "wife" });
+  });
+
+  it("passes the user in the body of memory writes and searches", async () => {
+    await capturedRun(["memories", "add", "Prefers tea", "--user-id", "wife"]);
+    expect(transport.post.mock.calls[0][0]).toBe("/api/v2/memories");
+    expect(transport.post.mock.calls[0][1]).toMatchObject({
+      text: "Prefers tea",
+      scope: "all-apps",
+      user_id: "wife",
+    });
+    const directory = await mkdtemp(path.join(tmpdir(), "om-user-id-"));
+    const content = path.join(directory, "content.txt");
+    await writeFile(content, "Prefers green tea\n");
+    await capturedRun([
+      "memories",
+      "update",
+      "m-1",
+      "--content-file",
+      content,
+      "--user-id",
+      "wife",
+    ]);
+    expect(transport.put.mock.calls[0][0]).toBe("/api/v2/memories/m-1");
+    expect(transport.put.mock.calls[0][1]).toMatchObject({ user_id: "wife" });
+    await capturedRun(["recall", "tea", "--user-id", "wife", "--json"]);
+    const search = transport.post.mock.calls.find(
+      ([url]) => url === "/api/v2/context/search",
+    );
+    expect(search?.[1]).toMatchObject({ query: "tea", user_id: "wife" });
+    const messages = path.join(directory, "messages.json");
+    await writeFile(
+      messages,
+      JSON.stringify([{ role: "user", content: "I like tea" }]),
+    );
+    await capturedRun([
+      "write",
+      "submit",
+      "--messages-file",
+      messages,
+      "--session",
+      "s-1",
+      "--user-id",
+      "wife",
+    ]);
+    expect(direct.post.mock.calls[0][0]).toBe(
+      "http://127.0.0.1:8765/api/v2/memory-writes",
+    );
+    expect(direct.post.mock.calls[0][1]).toMatchObject({
+      session_id: "s-1",
+      user_id: "wife",
+    });
+  });
+
+  it("names no user by default or for --user-id local", async () => {
+    for (const extra of [[], ["--user-id", "local"]]) {
+      transport.get.mockClear();
+      transport.post.mockClear();
+      await capturedRun(["memories", "list", ...extra, "--json"]);
+      expect(transport.get).toHaveBeenCalledWith("/api/v2/memory-library", {
+        params: { offset: 0, limit: 100 },
+      });
+      await capturedRun(["profile", ...extra]);
+      expect(transport.get).toHaveBeenCalledWith(
+        "/api/v2/user-profile",
+        undefined,
+      );
+      await capturedRun(["memories", "add", "Prefers tea", ...extra]);
+      expect(transport.post.mock.calls[0][1]).not.toHaveProperty("user_id");
+      await capturedRun(["recall", "tea", ...extra, "--json"]);
+      const search = transport.post.mock.calls.find(
+        ([url]) => url === "/api/v2/context/search",
+      );
+      expect(search?.[1]).not.toHaveProperty("user_id");
+    }
+  });
+
+  it("refuses a malformed user id before contacting Core", async () => {
+    for (const value of ["bad:id", "", "x".repeat(65), "a b", "--json"]) {
+      const result = await capturedRun([
+        "memories",
+        "list",
+        "--user-id",
+        value,
+      ]);
+      expect(result.code).toBe(2);
+    }
+    expect(
+      (await capturedRun(["profile", "--user-id", "a", "--user-id", "b"])).code,
+    ).toBe(2);
+    expect((await capturedRun(["recall", "tea", "--user-id"])).code).toBe(2);
+    expect(transport.get).not.toHaveBeenCalled();
+    expect(transport.post).not.toHaveBeenCalled();
+  });
+});
+
+describe("sources connect for a Hermes profile", () => {
+  it("connects hermes for another user from its own root", async () => {
+    const result = await capturedRun([
+      "sources",
+      "connect",
+      "hermes",
+      "--user-id",
+      "wife",
+      "--root",
+      "~/.hermes/profiles/wife",
+      "--json",
+    ]);
+    expect(result.code).toBe(0);
+    expect(transport.post).toHaveBeenCalledTimes(1);
+    const [url, body] = transport.post.mock.calls[0];
+    expect(url).toBe("/api/v1/sources/connect");
+    expect(body).toMatchObject({
+      source_ids: ["hermes"],
+      // Core reads the custom root's identity itself.
+      source_roots: {},
+      source_candidate_revisions: {},
+      onboarding: false,
+      user_id: "wife",
+      root: path.join(os.homedir(), ".hermes/profiles/wife"),
+    });
+  });
+
+  it("keeps a plain connect unchanged and accepts --import with a profile", async () => {
+    await capturedRun(["sources", "connect", "hermes", "--import", "--json"]);
+    const plain = transport.post.mock.calls[0][1];
+    expect(plain).not.toHaveProperty("user_id");
+    expect(plain).not.toHaveProperty("root");
+    expect(plain).toMatchObject({
+      source_roots: { hermes: "hermes-root-identity" },
+      onboarding: true,
+    });
+    transport.post.mockClear();
+    const result = await capturedRun([
+      "sources",
+      "connect",
+      "hermes",
+      "--import",
+      "--user-id",
+      "wife",
+      "--root",
+      "/profiles/wife",
+      "--request-id",
+      REQUEST,
+      "--json",
+    ]);
+    expect(result.code).toBe(0);
+    expect(transport.post.mock.calls[0][1]).toMatchObject({
+      request_id: REQUEST,
+      user_id: "wife",
+      root: "/profiles/wife",
+    });
+  });
+
+  it("refuses --user-id and --root for any source but hermes", async () => {
+    for (const argv of [
+      ["sources", "connect", "codex", "--user-id", "wife"],
+      ["sources", "connect", "codex", "--root", "/somewhere", "--import"],
+      ["sources", "connect", "hermes", "codex", "--user-id", "wife"],
+      ["sources", "connect", "--all-detected", "--user-id", "wife"],
+      ["sources", "connect", "hermes", "--user-id", "bad:id"],
+    ]) {
+      const result = await capturedRun(argv);
+      expect(result.code).toBe(2);
+      expect(result.stderr).toContain("hermes");
+    }
+    expect(transport.post).not.toHaveBeenCalled();
+  });
+});
+
+async function capturedRun(arguments_: string[]) {
+  let stdout = "";
+  let stderr = "";
+  const stdoutSpy = vi
+    .spyOn(process.stdout, "write")
+    .mockImplementation((chunk) => {
+      stdout += String(chunk);
+      return true;
+    });
+  const stderrSpy = vi
+    .spyOn(process.stderr, "write")
+    .mockImplementation((chunk) => {
+      stderr += String(chunk);
+      return true;
+    });
+  try {
+    return {
+      code: await run(arguments_, {
+        ONE_MEMORY_DATA_DIR: "/synthetic/data",
+      }),
+      stdout,
+      stderr,
+    };
+  } finally {
+    stdoutSpy.mockRestore();
+    stderrSpy.mockRestore();
+  }
+}

memhub/apps/core/pyproject.toml

@@ -12,7 +12,7 @@ dependencies = [
     "fastapi>=0.141.1",
     "filelock>=3.32.4",
     # Public SDK version validated against MemHub's permission and recall contracts.
-    "hy-memory==2.0.8",
+    "hy-memory==2.0.9",
     # chromadb arrives through the HY SDK; locked here so an unrelated upgrade
     # cannot move the SDK's store. Built-in embedding consumes numpy directly.
     "chromadb==1.5.9",

memhub/apps/core/src/one_memory/cli.py

@@ -584,6 +584,7 @@ def _sdk_migration_status(options: Sequence[str], stream: TextIO) -> int:
         }
     else:
         from one_memory.hy_runtime_config import HY_MEMORY_NAMESPACE
+        from one_memory.hy_sdk_runtime import builtin_retrieval_snapshot
         from one_memory.sdk_data_plane_migration import (
             MemoryAuthority,
             SdkDataPlaneMigrationStore,
@@ -595,14 +596,14 @@ def _sdk_migration_status(options: Sequence[str], stream: TextIO) -> int:
 
         authority = SdkDataPlaneMigrationStore(database_path).authority()
         if authority is MemoryAuthority.SDK:
-            profile = read_active_profile(database_path)
+            profile = read_active_profile(database_path) or builtin_retrieval_snapshot()
             leftovers = bool(
                 legacy_tables_present(database_path)
                 or abandoned_staging_present(
                     data_dir=settings.data_dir,
                     namespace=HY_MEMORY_NAMESPACE,
                 )
-                or (profile is not None and legacy_vector_collections(settings.data_dir, profile))
+                or legacy_vector_collections(settings.data_dir, profile)
             )
             state = "required" if leftovers else "not_required"
         else:

memhub/apps/core/src/one_memory/context_models.py

@@ -5,6 +5,9 @@ from typing import Annotated, Literal
 
 from pydantic import BaseModel, ConfigDict, Field, model_validator
 
+from one_memory.sdk_search_options import SdkSearchOptions
+from one_memory.user_scope import DEFAULT_USER_ID, UserId
+
 ContextKind = Literal["raw", "memory"]
 ContextSearchChannel = Literal["lexical", "semantic"]
 ContextSemanticStatus = Literal["not_searched", "completed", "failed", "timed_out"]
@@ -47,17 +50,24 @@ class ContextSearchRequest(BaseModel):
         return self
 
 
-class SdkContextSearchRequest(ContextSearchRequest):
+class SdkContextSearchRequest(ContextSearchRequest, SdkSearchOptions):
     """`POST /api/v2/context/search`: the three SDK recall lanes are set apart.
 
     `limit` is the normal lane's top-k and the cap on returned items;
     `schema_limit` and `intention_limit` are the SDK's schema (L6) and
     proactive intention (L7) lanes. A caller's values are passed through.
+
+    The `SdkSearchOptions` fields reach the memory lane's `client.search`;
+    the raw lane takes the floor and the source, session, agent and time
+    filters (never wider than the caller's raw authority). Absent, both
+    lanes make the call they always made.
     """
 
     limit: int = Field(default=10, ge=1, le=20, strict=True)
     schema_limit: int = Field(default=3, ge=0, le=20, strict=True)
     intention_limit: int = Field(default=1, ge=0, le=20, strict=True)
+    #: Whose memories to search. The App never sends it and reads `local`.
+    user_id: UserId = DEFAULT_USER_ID
 
 
 class ContextMessage(BaseModel):

memhub/apps/core/src/one_memory/context_routes.py

@@ -24,13 +24,57 @@ from one_memory.hy_runtime_config import HY_MEMORY_READER
 from one_memory.hy_sdk_runtime import sdk_relevance_floor
 from one_memory.hy_sdk_types import (
     HySdkMemoryRecord,
+    parse_sdk_lane_search,
     parse_sdk_raw_search,
     parse_sdk_search,
 )
 from one_memory.imports.storage import EvidenceStore
+from one_memory.sdk_search_options import SdkSearchFilters, SearchLane
+from one_memory.semantic_deadline import SemanticMemoryHit
+from one_memory.user_scope import owns
 
 CALL_EVENT_RESPONSE_BUDGET_MS = 250
 
+#: Pass-through filters the raw lane honours. Scene, layers, reader and the
+#: schema/working floors describe memory lanes; the raw lane is always the
+#: chat store's L1 raw layer.
+_RAW_LANE_FILTERS = ("created_after", "session_ids", "agent_ids", "exclude_sources", "rerank")
+
+
+def raw_lane_sources(
+    allowed: Sequence[str] | frozenset[str] | None,
+    requested: Sequence[str] | None,
+) -> list[str] | None:
+    """The raw lane's source filter: the caller's, never wider than its authority.
+
+    `allowed` empty or None means every source (the call Core always made).
+    None back means nothing is left to search.
+    """
+    allowed_list = list(allowed or ())
+    if requested is None:
+        return allowed_list
+    chosen = [
+        source for source in dict.fromkeys(requested) if not allowed_list or source in allowed_list
+    ]
+    return chosen or None
+
+
+def raw_lane_overrides(value: SdkSearchFilters) -> dict[str, object]:
+    """The caller's raw-lane overrides: its floor and its row filters."""
+    overrides = value.filter_kwargs()
+    found = {name: overrides[name] for name in _RAW_LANE_FILTERS if name in overrides}
+    floor = value.lane_floor("raw")
+    if floor is not None:
+        found["min_score"] = floor
+    return found
+
+
+def hit_namespace(hit: SemanticMemoryHit) -> dict[str, str]:
+    """The `get` namespace of a hit: nothing for chat, kind and variant otherwise."""
+    if hit.kind in {"experience", "fact"} and hit.variant is not None:
+        return {"kind": hit.kind, "variant": hit.variant}
+    return {}
+
 
 def sdk_context_router(dependencies: Sequence[Depends]) -> APIRouter:
     """SDK-native context: no Evidence/Governance ids or fabricated revisions."""
@@ -66,26 +110,36 @@ def sdk_context_router(dependencies: Sequence[Depends]) -> APIRouter:
             client = await client_handle.client()
         except Exception:
             raise HTTPException(503, "sdk_context_unavailable") from None
+        memory_mode = getattr(request.app.state, "memory_mode", None)
+        lite = callable(memory_mode) and memory_mode() == "lite"
 
         async def raw_lane() -> tuple[SdkRawContextItem, ...]:
             if "raw" not in value.kinds or not authority.raw_allowed:
                 return ()
+            sources = raw_lane_sources(authority.raw_sources, value.include_sources)
+            if sources is None:
+                return ()
             raw = await asyncio.to_thread(
                 client.search,
                 value.query,
-                user_ids=["local"],
-                agent_ids=[],
-                include_sources=list(authority.raw_sources or ()),
-                limit=value.limit,
-                min_score=0.0,
-                schema_min_score=0.0,
-                schema_limit=0,
-                intention_limit=0,
-                scene="chat",
-                reader=HY_MEMORY_READER,
-                layers=["l1_raw"],
+                **{
+                    "user_ids": [value.user_id],
+                    "agent_ids": [],
+                    "include_sources": sources,
+                    "limit": value.limit,
+                    "min_score": 0.0,
+                    "schema_min_score": 0.0,
+                    "schema_limit": 0,
+                    "intention_limit": 0,
+                    "scene": "chat",
+                    "reader": HY_MEMORY_READER,
+                    "layers": ["l1_raw"],
+                    **raw_lane_overrides(value),
+                },
+            )
+            hits = tuple(
+                hit for hit in parse_sdk_raw_search(raw) if value.passes_floor("raw", hit.score)
             )
-            hits = parse_sdk_raw_search(raw)
             return tuple(
                 SdkRawContextItem(
                     raw_id=hit.raw_id,
@@ -103,27 +157,48 @@ def sdk_context_router(dependencies: Sequence[Depends]) -> APIRouter:
             )
 
         async def memory_lane() -> tuple[SdkMemoryContextItem, ...]:
-            if "memory" not in value.kinds or not authority.memory_allowed:
+            # lite extracts nothing: its search answers raw turns, which the
+            # raw lane already reads under the raw authority.
+            if "memory" not in value.kinds or not authority.memory_allowed or lite:
                 return ()
+            overrides = value.search_overrides()
             raw = await asyncio.to_thread(
                 client.search,
                 value.query,
-                user_ids=["local"],
-                agent_ids=[],
-                limit=value.limit,
-                min_score=sdk_relevance_floor(),
-                schema_min_score=sdk_relevance_floor(),
-                schema_limit=value.schema_limit,
-                intention_limit=value.intention_limit,
-                scene="normal",
-                reader=HY_MEMORY_READER,
+                **{
+                    "user_ids": [value.user_id],
+                    "agent_ids": [],
+                    "limit": value.limit,
+                    "min_score": sdk_relevance_floor(),
+                    "schema_min_score": sdk_relevance_floor(),
+                    "schema_limit": value.schema_limit,
+                    "intention_limit": value.intention_limit,
+                    "scene": "normal",
+                    "reader": HY_MEMORY_READER,
+                    **overrides,
+                },
+            )
+            hits = (
+                tuple(
+                    hit
+                    for hit in parse_sdk_lane_search(
+                        raw, error_code="hy_memory_search_shape_unknown"
+                    )
+                    if value.passes_floor(cast(SearchLane, hit.lane), hit.score)
+                )
+                if overrides
+                else parse_sdk_search(raw, error_code="hy_memory_search_shape_unknown")
             )
-            hits = parse_sdk_search(raw, error_code="hy_memory_search_shape_unknown")
             items: list[SdkMemoryContextItem] = []
             for hit in hits:
-                record = await asyncio.to_thread(client.get, hit.memory_id)
+                record = await asyncio.to_thread(client.get, hit.memory_id, **hit_namespace(hit))
                 memory = None if record is None else HySdkMemoryRecord.from_sdk(record)
-                if memory is None or memory.status != "active":
+                # `get` takes no user: a hit is kept only if it is the caller's.
+                if (
+                    memory is None
+                    or memory.status != "active"
+                    or not owns(memory.user_id, value.user_id)
+                ):
                     continue
                 items.append(
                     SdkMemoryContextItem(

memhub/apps/core/src/one_memory/hy_graph_admin.py

@@ -71,6 +71,16 @@ def graph_store_enabled() -> bool:
     return raw != "none"
 
 
+def effective_memory_mode(configured: str = "pro") -> str:
+    """The SDK mode clients are built with: the setting, unless dev forces ultra.
+
+    `ONE_MEMORY_GRAPH_STORE` (any value but empty or `none`) predates the
+    setting and keeps working for development: it forces ultra whatever the
+    configured mode. Without it the configured mode (default pro) applies.
+    """
+    return "ultra" if graph_store_enabled() else configured
+
+
 def inspect_graph_width_mismatch(config: Any) -> GraphWidthMismatch | None:
     """Return a mismatch without renaming or deleting anything."""
     db_path = Path(config.graph_store.db_path)
@@ -83,11 +93,12 @@ def inspect_profile_graph_width(
     *,
     api_key: str | None,
     data_dir: Path,
+    mode: str | None = None,
 ) -> GraphWidthMismatch | None:
     """Inspect the graph path and width selected by the active profile."""
     del api_key
     return _inspect_graph_width(
-        _profile_graph_path(data_dir),
+        _profile_graph_path(data_dir, mode),
         _profile_embedding_width(profile),
     )
 
@@ -98,11 +109,12 @@ def decide_profile_graph_width(
     api_key: str | None,
     data_dir: Path,
     decision: GraphWidthDecision,
+    mode: str | None = None,
 ) -> GraphWidthOutcome:
     """Apply the user's decision to the active profile's graph."""
     del api_key
     config = _GraphAdminConfig(
-        graph_store=_GraphStoreConfig(str(_profile_graph_path(data_dir))),
+        graph_store=_GraphStoreConfig(str(_profile_graph_path(data_dir, mode))),
         embedder=_EmbeddingWidthConfig(_profile_embedding_width(profile)),
     )
     return apply_graph_width_decision(
@@ -240,8 +252,9 @@ def _inspect_graph_width(
     )
 
 
-def _profile_graph_path(data_dir: Path) -> Path:
-    name = "graph" if graph_store_enabled() else "graph-disabled"
+def _profile_graph_path(data_dir: Path, mode: str | None = None) -> Path:
+    ultra = mode == "ultra" if mode is not None else graph_store_enabled()
+    name = "graph" if ultra else "graph-disabled"
     return data_dir / "hy-memory" / HY_MEMORY_NAMESPACE / name
 
 

memhub/apps/core/src/one_memory/hy_runtime_config.py

@@ -14,7 +14,7 @@ from pathlib import Path
 from typing import Literal, cast
 
 HY_MEMORY_DISTRIBUTION = "hy-memory"
-HY_MEMORY_VERSION = "2.0.8"
+HY_MEMORY_VERSION = "2.0.9"
 HY_MEMORY_NAMESPACE = "governed-v2"
 DiagnosticLogLevel = Literal["error", "warning", "info", "debug"]
 

memhub/apps/core/src/one_memory/hy_sdk_migration.py

@@ -497,7 +497,7 @@ def _ensure_coding_schema(path: Path) -> None:
         CodingMemoryStore,
     )
 
-    # Relies on the pinned SDK's internals (hy-memory==2.0.8 is pinned exactly).
+    # Relies on the pinned SDK's internals (hy-memory==2.0.9 is pinned exactly).
     store = CodingMemoryStore.__new__(CodingMemoryStore)
     store._db_path = str(path)
     store._conn = None

memhub/apps/core/src/one_memory/hy_sdk_runtime.py

@@ -31,7 +31,7 @@ from one_memory.embedding_assets import (
 )
 from one_memory.hy_graph_admin import (
     GraphWidthMismatchError,
-    graph_store_enabled,
+    effective_memory_mode,
     inspect_graph_width_mismatch,
 )
 from one_memory.hy_sdk_types import HySdkRuntimeError
@@ -170,11 +170,14 @@ class SdkClientHandle:
         api_key: str | None,
         data_dir: Path,
         http_transport_factory: Callable[[], Any] | None = None,
+        mode: str | None = None,
     ) -> None:
         self._profile = profile
         self._api_key = api_key
         self._data_dir = data_dir
         self._http_transport_factory = http_transport_factory
+        #: The SDK mode this handle's client is built with, fixed for its life.
+        self.mode = mode or effective_memory_mode()
         self._lease: ClientLease | None = None
         self._lock = asyncio.Lock()
 
@@ -189,6 +192,7 @@ class SdkClientHandle:
                     api_key=self._api_key,
                     data_dir=self._data_dir,
                     http_transport_factory=self._http_transport_factory,
+                    mode=self.mode,
                 )
             return self._lease.client
 
@@ -217,6 +221,33 @@ def builtin_retrieval_snapshot() -> ModelProfileSnapshot:
     )
 
 
+def retrieval_view(profile: ModelProfileSnapshot) -> ModelProfileSnapshot:
+    """The read side of `profile`: its embedding space, and no generation.
+
+    Reads must open the collection writes fill (#242), and the collection is
+    named by the embedding only. Generation stays the built-in placeholder, so
+    a read never needs the generation model. A remote embedding still needs
+    the profile's API key, which `build_sdk_config` uses as the embedding key.
+    """
+    provider = profile.embedding_provider_kind or profile.provider_kind
+    endpoint: str | None = None
+    if provider is not ProviderKind.LOCAL_ONNX:
+        endpoint = profile.embedding_base_url
+        if not endpoint:
+            raise HySdkRuntimeError("hy_sdk_model_endpoint_missing")
+    return builtin_retrieval_snapshot().model_copy(
+        update={
+            "profile_id": profile.profile_id,
+            "revision": profile.revision,
+            "embedding_provider_kind": provider,
+            "embedding_model": profile.embedding_model,
+            "embedding_base_url": endpoint,
+            "embedding_dimensions": profile.embedding_dimensions,
+            "has_api_key": profile.has_api_key,
+        }
+    )
+
+
 def sdk_relevance_floor() -> float:
     raw = os.environ.get("ONE_MEMORY_HY_RELEVANCE_FLOOR", "").strip()
     if not raw:
@@ -249,10 +280,8 @@ def build_sdk_config(
         embedding_key = None
         dimensions = profile.embedding_dimensions or BUILTIN_EMBEDDING_DIMENSIONS
     else:
-        embedding_endpoint = _openai_endpoint(
-            embedding_provider,
-            profile.embedding_base_url or profile.base_url,
-        )
+        # Never the generation endpoint: the collection name hashes this one.
+        embedding_endpoint = _openai_endpoint(embedding_provider, profile.embedding_base_url)
         embedding_key = "ollama" if embedding_provider is ProviderKind.OLLAMA else api_key
         if embedding_provider is ProviderKind.OPENAI_COMPATIBLE and not embedding_key:
             raise HySdkRuntimeError("hy_sdk_embedding_api_key_missing")
@@ -327,9 +356,12 @@ def acquire_client(
     api_key: str | None,
     data_dir: Path,
     http_transport_factory: Callable[[], Any] | None = None,
+    mode: str | None = None,
 ) -> ClientLease:
     """Construct a public SDK client and return its explicit runtime lease.
 
+    `mode` is the SDK pipeline mode (lite / pro / ultra); None keeps the
+    pre-setting default (ultra when `ONE_MEMORY_GRAPH_STORE` is set, else pro).
     Waits while `hold_stores_offline` holds this home's stores.
     """
     key = data_dir.resolve()
@@ -343,6 +375,7 @@ def acquire_client(
             api_key=api_key,
             data_dir=data_dir,
             http_transport_factory=http_transport_factory,
+            mode=mode or effective_memory_mode(),
         )
     finally:
         with _STORE_GATE:
@@ -381,6 +414,7 @@ def _acquire_client(
     api_key: str | None,
     data_dir: Path,
     http_transport_factory: Callable[[], Any] | None,
+    mode: str = "pro",
 ) -> ClientLease:
     config = build_sdk_config(profile, api_key=api_key, data_dir=data_dir)
     config.http_transport_factory = http_transport_factory
@@ -398,7 +432,7 @@ def _acquire_client(
         with constructing:
             client = HyMemoryClient(
                 config=config,
-                mode="ultra" if graph_store_enabled() else "pro",
+                mode=mode,
                 runtime=runtime,
                 embedding_service=embedding_service,
             )
@@ -467,7 +501,7 @@ def vector_collection_bases(profile: ModelProfileSnapshot) -> tuple[str, tuple[s
         endpoint, dimensions = None, profile.embedding_dimensions or BUILTIN_EMBEDDING_DIMENSIONS
         legacy_endpoint = "local-onnx"
     else:
-        endpoint = _openai_endpoint(provider, profile.embedding_base_url or profile.base_url)
+        endpoint = _openai_endpoint(provider, profile.embedding_base_url)
         dimensions = profile.embedding_dimensions
         legacy_endpoint = endpoint
     legacy_model = profile.embedding_model or "bge-small-zh-v1.5"

memhub/apps/core/src/one_memory/hy_sdk_types.py

@@ -2,6 +2,7 @@
 
 from __future__ import annotations
 
+import re
 from collections.abc import Mapping
 from datetime import UTC, datetime
 from typing import Literal, NamedTuple, cast
@@ -391,6 +392,8 @@ class HySdkRawRecord(BaseModel):
     memory_at: int | None = None
     gmt_created: int | None = None
     custom: dict[str, object] = Field(default_factory=dict)
+    #: active, or shadow once pro extracted it; None when the SDK gave none.
+    status: str | None = None
 
     @classmethod
     def from_sdk(cls, raw: object) -> HySdkRawRecord:
@@ -417,11 +420,34 @@ class HySdkRawRecord(BaseModel):
                     raw.get("gmt_created") if isinstance(raw.get("gmt_created"), int) else None
                 ),
                 custom=dict(custom),
+                status=str(raw["status"]) if isinstance(raw.get("status"), str) else None,
             )
         except (KeyError, TypeError, ValueError) as error:
             raise HySdkReadError("hy_memory_raw_shape_unknown") from error
 
 
+_RAW_TURN_LINE = re.compile(r"^\[(user|assistant|tool|system)\]:\s?(.*)$")
+
+
+def raw_turn_messages(content: str) -> tuple[tuple[str, str], ...]:
+    """The `(role, text)` messages of one SDK raw turn.
+
+    The SDK stores a turn as `[user]: …\\n[assistant]: …`; a line without a
+    role prefix continues the message above it. Content that names no role
+    at all is one message of unknown role (`""`).
+    """
+    messages: list[list[str]] = []
+    for line in content.splitlines():
+        match = _RAW_TURN_LINE.match(line)
+        if match is not None:
+            messages.append([match.group(1), match.group(2)])
+        elif messages:
+            messages[-1][1] += "\n" + line
+        else:
+            messages.append(["", line])
+    return tuple((role, text) for role, text in messages)
+
+
 class HySdkRawHit(HySdkRawRecord):
     score: float = Field(ge=0, le=1)
 
@@ -516,6 +542,96 @@ class HySdkUserProfile(BaseModel):
         )
 
 
+class HySdkProfileVersion(BaseModel):
+    """One version on an L0 basic-info chain."""
+
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    memory_id: str
+    content: str
+    #: Conversation time as the SDK renders it on a chain (epoch or text).
+    memory_at: int | str | None = None
+    gmt_created: int | None = None
+    #: The chain head: the version the profile shows now.
+    current: bool = False
+
+
+class HySdkProfileChain(BaseModel):
+    """One L0 basic-info head and its versions, latest first."""
+
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    memory_id: str
+    content: str
+    status: str
+    memory_at: int | None = None
+    gmt_created: int | None = None
+    #: The structured fields this head holds (`basic_info_kv`).
+    basic_info: dict[str, object] = Field(default_factory=dict)
+    versions: tuple[HySdkProfileVersion, ...]
+
+
+class HySdkProfileVersions(BaseModel):
+    """`get_user_basic_info`: the long-term profile's L0 version chains."""
+
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    request_id: str
+    chains: tuple[HySdkProfileChain, ...]
+    elapsed_ms: float = Field(ge=0)
+
+    @classmethod
+    def from_sdk(cls, raw: object) -> HySdkProfileVersions:
+        if not isinstance(raw, Mapping):
+            raise HySdkReadError("hy_memory_profile_versions_shape_unknown")
+        request_id = raw.get("request_id")
+        rows = raw.get("memories")
+        elapsed_ms = raw.get("elapsed_ms")
+        if (
+            not isinstance(request_id, str)
+            or not isinstance(rows, list)
+            or not isinstance(elapsed_ms, int | float)
+            or isinstance(elapsed_ms, bool)
+        ):
+            raise HySdkReadError("hy_memory_profile_versions_shape_unknown")
+        chains: list[HySdkProfileChain] = []
+        for row in rows:
+            if not isinstance(row, Mapping) or not isinstance(row.get("memory_id"), str):
+                raise HySdkReadError("hy_memory_profile_versions_shape_unknown")
+            head_id = str(row["memory_id"])
+            chain = row.get("evolution_chain")
+            entries = chain if isinstance(chain, list) and chain else [{**row, "node_id": head_id}]
+            versions = tuple(
+                HySdkProfileVersion(
+                    memory_id=str(entry.get("node_id") or entry.get("memory_id")),
+                    content=str(entry.get("content") or ""),
+                    memory_at=(
+                        entry.get("memory_at")
+                        if isinstance(entry.get("memory_at"), int | str)
+                        and not isinstance(entry.get("memory_at"), bool)
+                        else None
+                    ),
+                    gmt_created=_epoch(entry.get("gmt_created")),
+                    current=str(entry.get("node_id") or entry.get("memory_id")) == head_id,
+                )
+                for entry in entries
+                if isinstance(entry, Mapping) and (entry.get("node_id") or entry.get("memory_id"))
+            )
+            basic_info = row.get("basic_info_kv")
+            chains.append(
+                HySdkProfileChain(
+                    memory_id=head_id,
+                    content=str(row.get("content") or ""),
+                    status=str(row.get("status") or "active"),
+                    memory_at=_epoch(row.get("memory_at")),
+                    gmt_created=_epoch(row.get("gmt_created")),
+                    basic_info=dict(basic_info) if isinstance(basic_info, Mapping) else {},
+                    versions=versions,
+                )
+            )
+        return cls(request_id=request_id, chains=tuple(chains), elapsed_ms=float(elapsed_ms))
+
+
 def parse_sdk_search(raw: object, *, error_code: str) -> tuple[SemanticMemoryHit, ...]:
     rows = _reply_rows(raw, error_code)
     found: list[SemanticMemoryHit] = []
@@ -542,6 +658,120 @@ def parse_sdk_search(raw: object, *, error_code: str) -> tuple[SemanticMemoryHit
     return tuple(found)
 
 
+def parse_sdk_lane_search(raw: object, *, error_code: str) -> tuple[SemanticMemoryHit, ...]:
+    """Every scene's search reply as hits that name their lane and namespace.
+
+    `parse_sdk_search` reads the chat reply only and forgets which lane a hit
+    came from. A caller may now pick the scene and a floor per lane, so this
+    reads all three reply shapes:
+
+    - chat: `memories` as `{schema, proactive, normal}` (or one list, normal);
+    - working/coding: `facts` and `experiences` of that variant;
+    - all: one fused `memories` list whose rows carry `scene`, `kind` and
+      their own lane's score (`lane_score`). The fused `score` is a rank
+      fusion (1/(60+rank)), not a relevance, so the lane score is reported.
+
+    Highest score first, first occurrence of an id kept.
+    """
+    if not isinstance(raw, Mapping):
+        raise HySdkReadError(error_code)
+    scene = raw.get("scene")
+    found: list[SemanticMemoryHit] = []
+    if scene == "all":
+        rows = raw.get("memories")
+        if not isinstance(rows, list) or any(not isinstance(row, Mapping) for row in rows):
+            raise HySdkReadError(error_code)
+        for row in rows:
+            hit = _lane_hit(row, scene=str(row.get("scene") or "chat"), score_key="lane_score")
+            if hit is not None:
+                found.append(hit)
+    elif scene in ("productivity", "working", "coding"):
+        variant = raw.get("variant") if raw.get("variant") in {"working", "coding"} else None
+        lane = variant or (scene if scene in {"working", "coding"} else "working")
+        for key, kind in (("facts", "fact"), ("experiences", "experience")):
+            rows = raw.get(key) or []
+            if not isinstance(rows, list) or any(not isinstance(row, Mapping) for row in rows):
+                raise HySdkReadError(error_code)
+            for row in rows:
+                hit = _lane_hit({**row, "kind": kind}, scene=lane, score_key="score")
+                if hit is not None:
+                    found.append(hit)
+    else:
+        memories = raw.get("memories")
+        if isinstance(memories, Mapping):
+            if set(memories) != {"schema", "proactive", "normal"}:
+                raise HySdkReadError(error_code)
+            channels = tuple(memories.items())
+        elif isinstance(memories, list):
+            channels = (("normal", memories),)
+        else:
+            raise HySdkReadError(error_code)
+        for lane, rows in channels:
+            if not isinstance(rows, list) or any(not isinstance(row, Mapping) for row in rows):
+                raise HySdkReadError(error_code)
+            for row in rows:
+                hit = _lane_hit(row, scene="chat", score_key="score", lane=str(lane))
+                if hit is not None:
+                    found.append(hit)
+    ordered: list[SemanticMemoryHit] = []
+    seen: set[tuple[str, str, str | None]] = set()
+    for hit in sorted(found, key=lambda value: value.score, reverse=True):
+        identity = (hit.memory_id, hit.kind, hit.variant)
+        if identity in seen:
+            continue
+        seen.add(identity)
+        ordered.append(hit)
+    return tuple(ordered)
+
+
+def _lane_hit(
+    row: Mapping[str, object],
+    *,
+    scene: str,
+    score_key: str,
+    lane: str | None = None,
+) -> SemanticMemoryHit | None:
+    kind = str(row.get("kind")) if row.get("kind") in {"experience", "fact"} else "chat"
+    if kind == "chat":
+        memory_id = row.get("memory_id") or row.get("id")
+        variant = None
+        hit_lane = lane or "normal"
+    else:
+        id_key = "experience_id" if kind == "experience" else "fact_id"
+        memory_id = row.get("memory_id") or row.get(id_key)
+        variant = scene if scene in {"working", "coding"} else None
+        if variant is None:
+            return None
+        hit_lane = variant
+    if not isinstance(memory_id, str) or not memory_id:
+        return None
+    value = row.get(score_key)
+    if not isinstance(value, int | float) or isinstance(value, bool):
+        value = row.get("score")
+    score = (
+        max(0.0, min(1.0, float(value)))
+        if isinstance(value, int | float) and not isinstance(value, bool)
+        else 0.0
+    )
+    if kind != "chat":
+        return SemanticMemoryHit(
+            memory_id=memory_id, score=score, lane=hit_lane, kind=kind, variant=variant
+        )
+    content = row.get("content")
+    member_ids, formatting_complete = _formatted_member_ids(row)
+    return SemanticMemoryHit(
+        memory_id=memory_id,
+        score=score,
+        formatted_content=(
+            content.strip() if isinstance(content, str) and content.strip() else None
+        ),
+        member_ids=member_ids,
+        sdk_formatted=True,
+        formatting_complete=formatting_complete,
+        lane=hit_lane,
+    )
+
+
 class LibraryHit(NamedTuple):
     """One keyword match in a library search window, with its namespace."""
 

memhub/apps/core/src/one_memory/imports/adapters.py

@@ -31,6 +31,16 @@ class ScanIssueCode(StrEnum):
     #: import it. Named for the reason, so a reader can tell a setting from
     #: a record the parser could not use.
     SUBAGENT_TRACE_DISABLED = "subagent_trace_disabled"
+    #: A host's own notice wearing the user's role — Hermes posts a finished
+    #: background process, and its gateway marks the post as not written by a
+    #: human. Nothing in it is the person's, so it never reaches extraction.
+    INJECTED_MACHINE_NOTICE = "injected_machine_notice"
+    #: The host could not transcribe a voice message and left a placeholder
+    #: naming the audio file. The person did speak, but the placeholder holds
+    #: only a path, which an extractor would otherwise read as a fact.
+    VOICE_PLACEHOLDER = "voice_placeholder"
+    #: A health check such as "Reply with exactly: PONG": no one's words.
+    PROBE = "probe"
     SOURCE_BUSY = "source_busy"
     UNSUPPORTED_SCHEMA = "unsupported_schema"
     BUDGET_EXCEEDED = "budget_exceeded"

memhub/apps/core/src/one_memory/imports/discovery.py

@@ -27,6 +27,7 @@ def discovered_off_loop(
     *,
     suffixes: frozenset[str],
     max_targets: int,
+    excluded_top_directories: frozenset[str] = frozenset(),
 ) -> AsyncIterator[SourceFileDiscoveryEvent]:
     """The same walk, for callers that have an event loop to keep free.
 
@@ -41,6 +42,7 @@ def discovered_off_loop(
             grant,
             suffixes=suffixes,
             max_targets=max_targets,
+            excluded_top_directories=excluded_top_directories,
         ),
         scope_key=grant.source_id,
     )
@@ -51,12 +53,15 @@ def discover_source_files(
     *,
     suffixes: frozenset[str],
     max_targets: int,
+    excluded_top_directories: frozenset[str] = frozenset(),
 ) -> Iterator[SourceFileDiscoveryEvent]:
     """Metadata only: one lstat per candidate, no open and no per-file grant lookup.
 
     The grant and root identity are checked once before the walk. Each listed
     file keeps the grant's check, which runs again when that file is opened;
     the open itself refuses symlinks and any inode other than the one listed.
+    `excluded_top_directories` are directories directly under the root that
+    are never entered, such as another user's Hermes profiles.
     """
     grant.check_root()
     inspected = 0
@@ -64,8 +69,12 @@ def discover_source_files(
         grant.root,
         followlinks=False,
     ):
+        at_root = Path(directory) == grant.root
         child_directories[:] = sorted(
-            child for child in child_directories if not (Path(directory) / child).is_symlink()
+            child
+            for child in child_directories
+            if not (Path(directory) / child).is_symlink()
+            and not (at_root and child in excluded_top_directories)
         )
         for filename in sorted(filenames):
             if Path(filename).suffix.lower() not in suffixes:

memhub/apps/core/src/one_memory/imports/grants.py

@@ -252,6 +252,18 @@ class SourceGrantStore:
             ).fetchone()
         return self._to_receipt(row) if row else None
 
+    def grants(self) -> tuple[tuple[SourceGrantReceipt, str], ...]:
+        """Every grant of every user, with its granted root path.
+
+        One source may hold several grants, one per `(user_id, source_id)`:
+        a Hermes root for `local` and a Hermes profile for another user.
+        """
+        with self._connect() as connection:
+            rows = connection.execute(
+                "SELECT * FROM source_grants ORDER BY source_id, user_id"
+            ).fetchall()
+        return tuple((self._to_receipt(row), str(row["requested_path"])) for row in rows)
+
     def require_receipt(self, *, user_id: str, source_id: str) -> SourceGrantReceipt:
         receipt = self.receipt(user_id=user_id, source_id=source_id)
         if receipt is None:

memhub/apps/core/src/one_memory/imports/live_sync.py

@@ -28,6 +28,7 @@ from one_memory.imports.adapters import (
 )
 from one_memory.imports.jobs import ImportJob, ImportJobStore, ImportStatus
 from one_memory.imports.supervisor import ImportSupervisor
+from one_memory.user_scope import DEFAULT_USER_ID, lane_key, split_lane
 
 #: How long an automatic retry will wait for a wedged worker to notice cancel.
 ABANDON_TIMEOUT_SECONDS = 5.0
@@ -64,7 +65,12 @@ class LiveSyncHealth(BaseModel):
 
 
 class PollingLiveSyncSupervisor:
-    """Near-live sync that reuses each adapter's persisted incremental cursor."""
+    """Near-live sync that reuses each adapter's persisted incremental cursor.
+
+    A "source" here is a lane key (`user_scope.lane_key`): the bare source id
+    for `local`, `source@user` for another user's grant. Cursors, jobs and
+    locks are therefore per `(user_id, source_id)`.
+    """
 
     def __init__(
         self,
@@ -92,8 +98,22 @@ class PollingLiveSyncSupervisor:
         }
         self._locks = {source_id: asyncio.Lock() for source_id in self.source_ids}
         self._task: asyncio.Task[None] | None = None
+        self._group: asyncio.TaskGroup | None = None
         self._observers: set[asyncio.Task[None]] = set()
 
+    def ensure_source(self, source_id: str) -> None:
+        """Follow a lane granted after start, such as a new Hermes profile."""
+        if source_id in self._locks:
+            return
+        self.source_ids = (*self.source_ids, source_id)
+        self._health.setdefault(
+            source_id,
+            LiveSyncHealth(source_id=source_id, mode=self.mode, state=LiveSyncState.IDLE),
+        )
+        self._locks[source_id] = asyncio.Lock()
+        if self._group is not None:
+            self._group.create_task(self._poll_source(source_id))
+
     def _set_health(self, source_id: str, value: LiveSyncHealth) -> None:
         previous = self._health.get(source_id)
         self._health[source_id] = value
@@ -152,7 +172,12 @@ class PollingLiveSyncSupervisor:
         self._watched_sources = frozenset(source_ids)
 
     def all_health(self) -> tuple[LiveSyncHealth, ...]:
-        return tuple(self.health(source_id) for source_id in self.source_ids)
+        """`local`'s lanes only: what the App has always been shown."""
+        return tuple(
+            self.health(source_id)
+            for source_id in self.source_ids
+            if split_lane(source_id)[0] == DEFAULT_USER_ID
+        )
 
     async def initialize_source(
         self, source_id: str, *, baseline_missing_files: bool = True
@@ -164,11 +189,12 @@ class PollingLiveSyncSupervisor:
         baselines = (
             await adapter.live_baseline() if isinstance(adapter, _LiveBaselineAdapter) else {}
         )
+        user_id, bare_source_id = split_lane(source_id)
         await in_the_background(
             partial(
                 self.job_store.ensure_live_cursors,
-                user_id="local",
-                source_id=source_id,
+                user_id=user_id,
+                source_id=bare_source_id,
                 baselines=baselines,
                 baseline_missing_files=baseline_missing_files,
             )
@@ -228,7 +254,7 @@ class PollingLiveSyncSupervisor:
                 (
                     job
                     for job in reversed(source_jobs)
-                    if job.source_id == source_id
+                    if lane_key(job.user_id, job.source_id) == source_id
                     and job.operation == "sync"
                     and job.status
                     in {
@@ -277,13 +303,13 @@ class PollingLiveSyncSupervisor:
             competing = [
                 job
                 for job in source_jobs
-                if job.source_id == source_id
+                if lane_key(job.user_id, job.source_id) == source_id
                 and job.operation != "sync"
                 and job.status
                 in {ImportStatus.PENDING, ImportStatus.RUNNING, ImportStatus.INTERRUPTED}
             ]
             bounded_history_incomplete = any(
-                job.source_id == source_id
+                lane_key(job.user_id, job.source_id) == source_id
                 and job.operation == "history"
                 and job.history_after is not None
                 and job.status is not ImportStatus.COMPLETED
@@ -379,11 +405,12 @@ class PollingLiveSyncSupervisor:
                 await close()
 
     async def _live_cursors(self, source_id: str, adapter: object) -> dict[str, SourceCursor]:
+        user_id, bare_source_id = split_lane(source_id)
         cursors = await in_the_background(
             partial(
                 self.job_store.get_source_cursors,
-                user_id="local",
-                source_id=source_id,
+                user_id=user_id,
+                source_id=bare_source_id,
                 operation="sync",
             )
         )
@@ -395,8 +422,8 @@ class PollingLiveSyncSupervisor:
         return await in_the_background(
             partial(
                 self.job_store.ensure_live_cursors,
-                user_id="local",
-                source_id=source_id,
+                user_id=user_id,
+                source_id=bare_source_id,
                 baselines=baselines,
             )
         )
@@ -520,9 +547,13 @@ class PollingLiveSyncSupervisor:
     async def _run(self) -> None:
         # A source owns its cadence; one large catch-up cannot delay every
         # other source. The group keeps shutdown responsible for every poller.
-        async with asyncio.TaskGroup() as group:
-            for source_id in self.source_ids:
-                group.create_task(self._poll_source(source_id))
+        try:
+            async with asyncio.TaskGroup() as group:
+                self._group = group
+                for source_id in self.source_ids:
+                    group.create_task(self._poll_source(source_id))
+        finally:
+            self._group = None
 
     async def _poll_source(self, source_id: str) -> None:
         while True:

memhub/apps/core/src/one_memory/imports/runner.py

@@ -3,7 +3,7 @@ import logging
 import math
 import sqlite3
 from collections import Counter
-from collections.abc import Callable
+from collections.abc import AsyncIterator, Callable
 from contextlib import suppress
 from contextvars import ContextVar
 from dataclasses import dataclass, field
@@ -227,14 +227,17 @@ class ImportRunner:
         heartbeat = asyncio.create_task(self._keep_alive(job_id=job_id, runner_token=runner_token))
         pending = _PendingSkips()
         writes = _PendingWrites()
-        stream = adapter.scan(
-            ScanOptions(
-                cursors=cursors,
-                reuse_listing=(
-                    job.operation == "history"
-                    or (job.operation == "sync" and job.source_id == "workbuddy")
-                ),
-            )
+        stream = _owned_by(
+            adapter.scan(
+                ScanOptions(
+                    cursors=cursors,
+                    reuse_listing=(
+                        job.operation == "history"
+                        or (job.operation == "sync" and job.source_id == "workbuddy")
+                    ),
+                )
+            ),
+            job.user_id,
         )
         slice_cursors = dict(cursors)
         next_event: asyncio.Future[ScanEvent] | None = None
@@ -924,3 +927,30 @@ async def _settled_work[T](
         with suppress(asyncio.CancelledError, Exception):
             task.result()
         raise
+
+
+def _owned_by(stream: AsyncIterator[ScanEvent], user_id: str) -> AsyncIterator[ScanEvent]:
+    """Stamp every scanned record with the user whose root is being read.
+
+    Adapters know nothing of users and emit `local`. Which user a record
+    belongs to is decided by the grant the job reads (a Hermes profile's
+    `state.db` is that profile's user), so the job's user is the record's.
+    `local` passes the stream through untouched.
+    """
+    if user_id == "local":
+        return stream
+    return _restamped(stream, user_id)
+
+
+async def _restamped(stream: AsyncIterator[ScanEvent], user_id: str) -> AsyncIterator[ScanEvent]:
+    try:
+        async for event in stream:
+            if isinstance(event, ScannedEvidence) and event.evidence.user_id != user_id:
+                event = event.model_copy(
+                    update={"evidence": event.evidence.model_copy(update={"user_id": user_id})}
+                )
+            yield event
+    finally:
+        close = getattr(stream, "aclose", None)
+        if close is not None:
+            await close()

memhub/apps/core/src/one_memory/imports/sources/hermes.py

@@ -1,4 +1,5 @@
 import json
+import re
 import sqlite3
 from collections.abc import AsyncIterator, Iterator
 from functools import partial
@@ -87,6 +88,7 @@ class HermesSourceAdapter:
             self.grant,
             suffixes=frozenset({".jsonl", ".db"}),
             max_targets=10_000,
+            excluded_top_directories=frozenset({HERMES_PROFILES_DIRECTORY}),
         ):
             if isinstance(item, AuthorizedSourceFile) and _is_hermes_target(
                 item.relative_path.as_posix()
@@ -100,6 +102,7 @@ class HermesSourceAdapter:
             self.grant,
             suffixes=frozenset({".jsonl", ".db"}),
             max_targets=options.max_scan_targets or 100_000,
+            excluded_top_directories=frozenset({HERMES_PROFILES_DIRECTORY}),
         ):
             yield ScanProgressEvent(
                 phase=ScanPhase.DISCOVER,
@@ -214,7 +217,11 @@ def parse_hermes_rollout_record(
         message.get("content"),
         allowed_block_types=frozenset({"text", "input_text", "output_text"}),
     )
-    text = strip_injected_memory(text)
+    text = strip_injected_memory(text or "")
+    if role == "user":
+        text, injected = gateway_user_text(text)
+        if injected is not None:
+            return ParsedSkip(reason_code=injected, conversation_id=conversation_id)
     if not text:
         return ParsedSkip(
             reason_code=ScanIssueCode.NO_TEXT,
@@ -242,8 +249,17 @@ def parse_hermes_rollout_record(
     )
 
 
+#: Hermes keeps each extra profile under `<root>/profiles/<name>/`, with its
+#: own `state.db` and sessions. A profile is another person's history: it is
+#: read only through its own grant (its own MemHub user), never as part of the
+#: root above it. Hermes' own export skips this directory for the same reason.
+HERMES_PROFILES_DIRECTORY = "profiles"
+
+
 def _is_hermes_target(relative_path: str) -> bool:
     path = PurePosixPath(relative_path)
+    if path.parts and path.parts[0] == HERMES_PROFILES_DIRECTORY:
+        return False
     return path.name == "state.db" or (path.suffix == ".jsonl" and "sessions" in path.parts)
 
 
@@ -389,6 +405,15 @@ def _read_state_db(
             # `tool` rows after it hold only what came back.
             calls = _tool_calls_text(row["tool_calls"]) if role == "assistant" else None
             content = strip_injected_memory(str(row["content"] or ""))
+            if role == "user":
+                content, injected = gateway_user_text(content)
+                if injected is not None:
+                    yield ScanSkip(
+                        cursor=cursor,
+                        conversation_id=str(row["session_id"]),
+                        reason_code=injected,
+                    )
+                    continue
             has_text = bool(content.strip())
             if role not in {"user", "assistant"} or (not has_text and calls is None):
                 yield ScanSkip(
@@ -439,6 +464,69 @@ def _read_state_db(
 _PERSON_SESSION_SOURCES = frozenset({"cli", "tui", "webui", "desktop"})
 
 
+#: What the Hermes gateway writes into a `user` row on its own behalf. Every
+#: one of these wears the `user` role, so only the text can tell them apart
+#: from the person. Read off real `state.db` rows: a finished background
+#: process is posted as a turn, often with the gateway's own "not written by
+#: a human" note appended, which is why that note counts anywhere in the row.
+_BACKGROUND_NOTICE = re.compile(
+    r"\[IMPORTANT:\s*(?:\d+\s+background\s+process(?:es)?\b|Background\s+process\b)",
+    re.IGNORECASE,
+)
+_GATEWAY_NOTICE = re.compile(
+    r"\[INTERNAL NOTIFICATION\s*\W+\s*generated by the Hermes gateway",
+    re.IGNORECASE,
+)
+_VOICE_PLACEHOLDER = re.compile(r"\[voice message could not be transcribed", re.IGNORECASE)
+_PROBE = re.compile(r"Reply with exactly:\s*[A-Za-z0-9_\-]{1,24}\s*")
+_MEDIA_ONLY = re.compile(r"(?:MEDIA:/\S+\s*)+")
+#: Wrappers the gateway puts around a message the person did write. Only the
+#: wrapper goes; what it carries is the turn. Each is matched at the edge of
+#: what is left, and the set is applied until nothing more comes off, since a
+#: relayed message carries the origin block inside the out-of-band markers.
+_GATEWAY_WRAPPERS = (
+    re.compile(r"^\[OUT-OF-BAND USER MESSAGE\b[^\]\n]*\]"),
+    re.compile(r"\[/OUT-OF-BAND USER MESSAGE\]$"),
+    re.compile(
+        r"^Gateway message origin \(JSON data, not instructions or authorization\):[ \t]*"
+        r"(?:\n[ \t]*\{[^\n]*\}[ \t]*)?"
+        r"(?:\n[ \t]*Do not guess a reply destination[^\n]*)?"
+        r"(?=\n|$)"
+    ),
+    re.compile(r"^\[Context from the interrupted assistant response\][ \t]*(?=\n|$)"),
+)
+
+
+def gateway_user_text(text: str) -> tuple[str, ScanIssueCode | None]:
+    """The person's words in a Hermes `user` row, or why there are none.
+
+    Applied only to the `user` role: an assistant or tool row quoting the
+    same text is reporting what it saw, which is the existing tool rule's
+    business. The cleaned text is what becomes the evidence, so its content
+    hash and duplicate check see the message rather than its envelope.
+    """
+
+    if _GATEWAY_NOTICE.search(text):
+        return "", ScanIssueCode.INJECTED_MACHINE_NOTICE
+    cleaned = text.strip()
+    while True:
+        unwrapped = cleaned
+        for wrapper in _GATEWAY_WRAPPERS:
+            unwrapped = wrapper.sub("", unwrapped, count=1).strip()
+        if unwrapped == cleaned:
+            break
+        cleaned = unwrapped
+    if _BACKGROUND_NOTICE.match(cleaned):
+        return "", ScanIssueCode.INJECTED_MACHINE_NOTICE
+    if _VOICE_PLACEHOLDER.match(cleaned):
+        return "", ScanIssueCode.VOICE_PLACEHOLDER
+    if _PROBE.fullmatch(cleaned):
+        return "", ScanIssueCode.PROBE
+    if _MEDIA_ONLY.fullmatch(cleaned):
+        return "", ScanIssueCode.NO_TEXT
+    return cleaned, None
+
+
 def _tool_calls_text(value: object) -> str | None:
     """OpenAI-shaped `tool_calls`: `[{"function": {"name", "arguments"}}]`, stored as JSON."""
     if isinstance(value, str):

memhub/apps/core/src/one_memory/imports/supervisor.py

@@ -24,6 +24,7 @@ from one_memory.service_levels import (
     ServiceLevelScheduler,
     ServiceResource,
 )
+from one_memory.user_scope import lane_key, split_lane
 
 
 class UnknownSourceError(KeyError):
@@ -82,36 +83,45 @@ class ImportSupervisor:
         self._tasks: dict[str, asyncio.Task[None]] = {}
         self._drains: dict[str, asyncio.Task[None]] = {}
 
-    def register_adapter(self, adapter: SourceAdapter) -> None:
-        self.adapters[adapter.descriptor.source_id] = adapter
+    def register_adapter(self, adapter: SourceAdapter, *, user_id: str = "local") -> None:
+        """Adapters are keyed by lane: the source id for `local`, `source@user`
+        for another user's grant of the same source."""
+        self.adapters[lane_key(user_id, adapter.descriptor.source_id)] = adapter
 
     def unregister_adapter(self, source_id: str) -> None:
         self.adapters.pop(source_id, None)
 
+    @staticmethod
+    def lane_of(job: ImportJob) -> str:
+        return lane_key(job.user_id, job.source_id)
+
     def active_job_ids(self) -> frozenset[str]:
         return frozenset(job_id for job_id, task in self._tasks.items() if not task.done())
 
     def create_job(
         self, source_id: str, *, operation: Literal["history", "sync"] = "history"
     ) -> ImportJob:
+        """`source_id` is a lane key: another user's lane creates that user's job."""
         if self.source_enabled is not None and not self.source_enabled(source_id):
             raise SourceAccessError(SourceAccessErrorCode.GRANT_REQUIRED)
-        return self.job_store.create(user_id="local", source_id=source_id, operation=operation)
+        user_id, bare_source_id = split_lane(source_id)
+        return self.job_store.create(user_id=user_id, source_id=bare_source_id, operation=operation)
 
     def enqueue(self, job_id: str) -> ImportJob:
         if job_id in self._drains:
             raise InvalidJobTransitionError("import_cleanup_in_progress")
         job = self.job_store.require(job_id)
-        if job.source_id not in self.adapters:
-            raise UnknownSourceError(job.source_id)
-        if self.source_enabled is not None and not self.source_enabled(job.source_id):
+        lane = self.lane_of(job)
+        if lane not in self.adapters:
+            raise UnknownSourceError(lane)
+        if self.source_enabled is not None and not self.source_enabled(lane):
             return job
         existing = self._tasks.get(job_id)
         if existing is None or existing.done():
             self._tasks[job_id] = asyncio.create_task(
                 self._execute(
                     job_id,
-                    source_id=job.source_id,
+                    source_id=lane,
                     operation=job.operation,
                 )
             )
@@ -130,8 +140,10 @@ class ImportSupervisor:
         # pending history. Each resumed runner will yield after one bounded slice.
         for status in (ImportStatus.INTERRUPTED, ImportStatus.PENDING):
             for job in self.job_store.list_jobs(status=status):
-                if job.source_id in self.adapters:
-                    if self.source_enabled is not None and not self.source_enabled(job.source_id):
+                if self.lane_of(job) in self.adapters:
+                    if self.source_enabled is not None and not self.source_enabled(
+                        self.lane_of(job)
+                    ):
                         continue
                     if job.status is not ImportStatus.PENDING:
                         self.job_store.queue_for_resume(job.job_id)
@@ -227,7 +239,7 @@ class ImportSupervisor:
                     }:
                         return
                     if self.source_enabled is not None and not await in_the_background(
-                        self.source_enabled, job.source_id
+                        self.source_enabled, self.lane_of(job)
                     ):
                         return
                     runner = ImportRunner(
@@ -238,7 +250,7 @@ class ImportSupervisor:
                         sdk_queue=self.sdk_queue,
                         sdk_authority=self.sdk_authority,
                     )
-                    adapter = self.adapters[job.source_id]
+                    adapter = self.adapters[self.lane_of(job)]
                     if self.prepare_adapter is not None:
                         adapter = await self.prepare_adapter(job, adapter)
 
@@ -273,7 +285,7 @@ class ImportSupervisor:
                 and retry_attempt < self.max_history_retries
                 and (
                     self.source_enabled is None
-                    or await in_the_background(self.source_enabled, current.source_id)
+                    or await in_the_background(self.source_enabled, self.lane_of(current))
                 )
             ):
                 delay = self.retry_base_seconds * (2**retry_attempt)

memhub/apps/core/src/one_memory/library_search_window.py

@@ -7,9 +7,9 @@ between pages. The matches are kept here under an opaque ``window_id`` that
 the client passes back with the next offset.
 
 A window is bound to the filters it was searched with (query, source,
-statuses, scene). It expires after ``ttl_seconds`` without being read, and the
-store keeps at most ``max_windows`` windows, dropping the least recently read
-first.
+statuses, scene) and to the user whose library it searched. It expires after
+``ttl_seconds`` without being read, and the store keeps at most
+``max_windows`` windows, dropping the least recently read first.
 Only memory identities are kept, never content: each page is read fresh from
 the SDK, so an edit or a delete shows on the next page read.
 """
@@ -28,13 +28,24 @@ from one_memory.hy_sdk_types import LibraryHit
 DEFAULT_WINDOW_TTL_SECONDS = 300.0
 DEFAULT_MAX_WINDOWS = 32
 
-WindowKey = tuple[str, str | None, tuple[str, ...], str]
+#: (query, source, statuses, scene, user, other filters as sorted name/values pairs)
+WindowKey = tuple[
+    str, str | None, tuple[str, ...], str, str, tuple[tuple[str, tuple[str, ...]], ...]
+]
 
 
 def window_key(
-    query: str, source: str | None, statuses: list[str], scene: str = "all"
+    query: str,
+    source: str | None,
+    statuses: list[str],
+    scene: str = "all",
+    user_id: str = "local",
+    filters: dict[str, list[str]] | None = None,
 ) -> WindowKey:
-    return (query, source, tuple(sorted(set(statuses))), scene)
+    extra = tuple(
+        sorted((name, tuple(sorted(set(values)))) for name, values in (filters or {}).items())
+    )
+    return (query, source, tuple(sorted(set(statuses))), scene, user_id, extra)
 
 
 @dataclass(frozen=True)

memhub/apps/core/src/one_memory/memory_additions.py

@@ -11,10 +11,11 @@ import asyncio
 import sqlite3
 from collections.abc import Sequence
 from datetime import UTC, datetime
+from functools import partial
 from typing import Any, Literal, cast
 from uuid import UUID
 
-from fastapi import APIRouter, BackgroundTasks, HTTPException, Request
+from fastapi import APIRouter, BackgroundTasks, HTTPException, Query, Request
 from fastapi.params import Depends
 from pydantic import BaseModel, ConfigDict, Field, field_validator
 
@@ -34,6 +35,7 @@ from one_memory.memory_server.durable_queue import DurableWriteQueue
 from one_memory.memory_server.unified_entry import UnifiedWriteIntake
 from one_memory.memory_tasks import MemoryTaskStatus, MemoryTaskStore
 from one_memory.secrets import SecretStoreUnavailableError
+from one_memory.user_scope import DEFAULT_USER_ID, USER_ID_PATTERN, UserId
 
 SOURCE_ID = "one-memory-manual"
 MemoryAddState = Literal[
@@ -53,6 +55,8 @@ class MemoryAddInput(BaseModel):
     request_id: UUID
     text: str = Field(min_length=1, max_length=12_000)
     scope: Literal["all-apps"]
+    #: Whose memory this is. The App never sends it and writes `local`.
+    user_id: UserId = DEFAULT_USER_ID
 
     @field_validator("text")
     @classmethod
@@ -108,7 +112,11 @@ class MemoryAdditionStore:
 
     def save(self, addition: MemoryAddInput) -> EvidenceRecord:
         now = datetime.now(UTC)
+        if addition.user_id != DEFAULT_USER_ID and not self.sdk_authority:
+            # Legacy authority keeps one person's library only.
+            raise ValueError("memory_add_user_requires_sdk_authority")
         value = EvidenceInput(
+            user_id=addition.user_id,
             source_id=SOURCE_ID,
             conversation_id=f"note:{addition.request_id}",
             message_id=str(addition.request_id),
@@ -200,7 +208,7 @@ class MemoryAdditionStore:
                 )
         return self.require(value.evidence_id)
 
-    def require(self, evidence_id: str) -> EvidenceRecord:
+    def require(self, evidence_id: str, *, user_id: str = DEFAULT_USER_ID) -> EvidenceRecord:
         if self.sdk_authority:
             if self.write_queue is None:
                 raise KeyError(evidence_id)
@@ -219,11 +227,11 @@ class MemoryAdditionStore:
                 (item for item in payload.input_records if item.evidence_id == evidence_id),
                 None,
             )
-            if record is None:
+            if record is None or record.user_id != user_id:
                 raise KeyError(evidence_id)
             return record
         record = self.evidence.get(evidence_id)
-        if record is None or record.user_id != "local" or record.source_id != SOURCE_ID:
+        if record is None or record.user_id != user_id or record.source_id != SOURCE_ID:
             raise KeyError(evidence_id)
         metadata = record.origin_meta.get(EvidenceOrigin.MANUAL_IMPORT)
         if metadata is None or metadata.raw_meta.get("input_mode") != "direct_memory_note":
@@ -237,8 +245,8 @@ class MemoryAdditionStore:
             raise ValueError("memory_add_source_revoked")
         return record
 
-    def receipt(self, evidence_id: str) -> MemoryAddReceipt:
-        record = self.require(evidence_id)
+    def receipt(self, evidence_id: str, *, user_id: str = DEFAULT_USER_ID) -> MemoryAddReceipt:
+        record = self.require(evidence_id, user_id=user_id)
         if self.sdk_authority:
             assert self.write_queue is not None
             with main_database(self.write_queue.database_path).read() as database:
@@ -300,10 +308,15 @@ class MemoryAdditionStore:
             error=error,
         )
 
-    def receipt_for_request(self, request_id: UUID) -> MemoryAddReceipt:
+    def receipt_for_request(
+        self, request_id: UUID, *, user_id: str = DEFAULT_USER_ID
+    ) -> MemoryAddReceipt:
         identity = str(request_id)
         if self.sdk_authority:
+            # The Evidence id carries the user, so another user's request id
+            # names a different note and reads as not found.
             probe = EvidenceInput(
+                user_id=user_id,
                 source_id=SOURCE_ID,
                 conversation_id=f"note:{identity}",
                 message_id=identity,
@@ -316,16 +329,16 @@ class MemoryAdditionStore:
                     )
                 },
             )
-            return self.receipt(probe.evidence_id)
+            return self.receipt(probe.evidence_id, user_id=user_id)
         with main_database(self.evidence.database_path).read() as db:
             row = db.execute(
-                "SELECT evidence_id FROM evidence WHERE user_id='local' AND source_id=? "
+                "SELECT evidence_id FROM evidence WHERE user_id=? AND source_id=? "
                 "AND conversation_id=? AND message_id=?",
-                (SOURCE_ID, f"note:{identity}", identity),
+                (user_id, SOURCE_ID, f"note:{identity}", identity),
             ).fetchone()
         if row is None:
             raise KeyError(identity)
-        return self.receipt(str(row[0]))
+        return self.receipt(str(row[0]), user_id=user_id)
 
 
 def memory_add_router(dependencies: Sequence[Depends]) -> APIRouter:
@@ -376,14 +389,20 @@ def memory_add_router(dependencies: Sequence[Depends]) -> APIRouter:
             record = await asyncio.to_thread(additions.save, addition)
         except ValueError as error:
             raise HTTPException(409, str(error)) from error
-        saved = await asyncio.to_thread(additions.receipt, record.evidence_id)
+        saved = await asyncio.to_thread(
+            partial(additions.receipt, record.evidence_id, user_id=addition.user_id)
+        )
         background.add_task(schedule, additions, record)
         return saved
 
     @router.get("/additions/{request_id}", response_model=MemoryAddReceipt)
-    def memory_add_status(request_id: UUID, request: Request) -> MemoryAddReceipt:
+    def memory_add_status(
+        request_id: UUID,
+        request: Request,
+        user_id: str = Query(default=DEFAULT_USER_ID, pattern=USER_ID_PATTERN),
+    ) -> MemoryAddReceipt:
         try:
-            return store(request).receipt_for_request(request_id)
+            return store(request).receipt_for_request(request_id, user_id=user_id)
         except KeyError as error:
             raise HTTPException(404, "memory_add_not_found") from error
         except ValueError as error:

memhub/apps/core/src/one_memory/memory_evolution.py

@@ -0,0 +1,618 @@
+"""System 2 for ultra: evolve runs, their receipts, the nightly schedule, graph backfill.
+
+hy-memory 2.0.9's `client.evolve(user_id, agent_id, max_batches)` is ultra
+only, holds no lock, and has no scheduler: MemHub owns all three.
+
+- A run is asynchronous: `start` answers a receipt at once and the run goes
+  on in the background (evolve makes model calls across the user's whole
+  library and can take minutes, longer than any HTTP timeout). One run per
+  user at a time: asking again while one runs returns that run. A run holds
+  the write coordinator's per-user exclusivity, so it never overlaps that
+  user's write batch, and loops while the SDK answers `should_continue`, up
+  to `EVOLVE_MAX_ROUNDS`.
+- The schedule runs every user that has memories once a day at the
+  configured local time, and catches up after the machine slept through it.
+- Switching to ultra backfills each user's existing facts into the graph
+  once (`build_memory_graph`).
+
+What is kept on disk is counts and stable codes only, never memory text:
+`<data_dir>/evolve-state.json`.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+import os
+import re
+import threading
+from collections.abc import Awaitable, Callable, Iterable, Mapping
+from datetime import UTC, datetime, timedelta
+from pathlib import Path
+from tempfile import NamedTemporaryFile
+from typing import Any, Literal, Protocol
+from uuid import uuid4
+
+from pydantic import BaseModel, ConfigDict, Field
+
+from one_memory.diagnostic_log import diagnostic_event
+from one_memory.memory_server.config_state import EvolveSchedule
+from one_memory.memory_server.durable_queue import safe_error_code
+from one_memory.sdk_write_processor import PRODUCT_AGENT_ID
+
+logger = logging.getLogger(__name__)
+
+#: Rounds one run makes while the SDK says there is more to do.
+EVOLVE_MAX_ROUNDS = 20
+#: How often the schedule looks at the clock; also bounds catch-up after wake.
+SCHEDULE_CHECK_SECONDS = 300.0
+#: Receipts kept in memory for `GET /api/v2/evolve/{run_id}`.
+_KEPT_RUNS = 64
+_STABLE_CODE = re.compile(r"[a-z][a-z0-9_]{0,95}")
+_COUNTS = (
+    "remaining_clusters",
+    "remaining_actionable_clusters",
+    "remaining_fresh_facts",
+)
+
+RunState = Literal["running", "succeeded", "failed"]
+Trigger = Literal["manual", "scheduled"]
+
+
+class EvolveModeError(RuntimeError):
+    """Evolve asked for outside ultra."""
+
+    def __init__(self) -> None:
+        super().__init__("memory_mode_not_ultra")
+
+
+class EvolveRunReceipt(BaseModel):
+    """One evolve run. Counts and codes; `results` only while in memory."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    run_id: str
+    user_id: str
+    agent_id: str
+    max_batches: int | None = None
+    trigger: Trigger = "manual"
+    state: RunState = "running"
+    started_at: datetime
+    finished_at: datetime | None = None
+    rounds: int = 0
+    tasks_processed: int = 0
+    batch_failures: int = 0
+    cycle_status: str | None = None
+    should_continue: bool | None = None
+    continuation_reason: str | None = None
+    remaining_clusters: int | None = None
+    remaining_actionable_clusters: int | None = None
+    remaining_fresh_facts: int | None = None
+    #: True when the round cap stopped a run the SDK would have continued.
+    capped: bool = False
+    elapsed_ms: float = 0.0
+    error_code: str | None = None
+    failure_kind: str | None = None
+    #: The last round's SDK `request_id`, for matching SDK logs.
+    request_id: str | None = None
+    #: The last round's SDK `results` and `cross_domain_sweeper`. Kept only
+    #: in this receipt, never written to disk or diagnostics.
+    results: Any = None
+    cross_domain_sweeper: Any = None
+
+    def summary(self) -> dict[str, object]:
+        """What is persisted and shown in status: no SDK result bodies."""
+        return self.model_dump(mode="json", exclude={"results", "cross_domain_sweeper"})
+
+
+class UserEvolveState(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    last_run_at: datetime | None = None
+    last_result: dict[str, object] | None = None
+    last_error: str | None = None
+
+
+class GraphBuildUser(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    state: Literal["pending", "running", "succeeded", "failed"] = "pending"
+    memories_scanned: int | None = None
+    nodes: int | None = None
+    error_code: str | None = None
+
+
+class GraphBuildState(BaseModel):
+    """The one-time backfill of existing facts into the graph after → ultra."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    state: Literal["idle", "running", "succeeded", "failed"] = "idle"
+    started_at: datetime | None = None
+    finished_at: datetime | None = None
+    users: dict[str, GraphBuildUser] = Field(default_factory=dict)
+
+
+class EvolveStateDocument(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    schema_version: int = Field(default=1, ge=1, le=1)
+    users: dict[str, UserEvolveState] = Field(default_factory=dict)
+    #: When the schedule first ran in ultra: the "last run" of a user that
+    #: never ran, so enabling ultra at noon does not evolve everyone at once.
+    schedule_anchor: datetime | None = None
+    graph_build: GraphBuildState = Field(default_factory=GraphBuildState)
+
+
+class EvolveStateStore:
+    """The only writer of `<data_dir>/evolve-state.json`."""
+
+    def __init__(self, data_dir: Path) -> None:
+        self.path = data_dir / "evolve-state.json"
+        self._lock = threading.RLock()
+        self._document = self._read()
+
+    def document(self) -> EvolveStateDocument:
+        with self._lock:
+            return self._document.model_copy(deep=True)
+
+    def user(self, user_id: str) -> UserEvolveState:
+        with self._lock:
+            return self._document.users.get(user_id, UserEvolveState()).model_copy()
+
+    def record_run(self, receipt: EvolveRunReceipt) -> None:
+        with self._lock:
+            users = dict(self._document.users)
+            users[receipt.user_id] = UserEvolveState(
+                last_run_at=receipt.started_at,
+                last_result=receipt.summary(),
+                last_error=receipt.error_code,
+            )
+            self._save(self._document.model_copy(update={"users": users}))
+
+    def anchor(self, now: datetime) -> datetime:
+        """The schedule anchor, set to `now` the first time it is asked."""
+        with self._lock:
+            if self._document.schedule_anchor is None:
+                self._save(self._document.model_copy(update={"schedule_anchor": now}))
+            assert self._document.schedule_anchor is not None
+            return self._document.schedule_anchor
+
+    def graph_build(self) -> GraphBuildState:
+        with self._lock:
+            return self._document.graph_build.model_copy(deep=True)
+
+    def set_graph_build(self, state: GraphBuildState) -> None:
+        with self._lock:
+            self._save(self._document.model_copy(update={"graph_build": state}))
+
+    def _save(self, document: EvolveStateDocument) -> None:
+        payload = json.dumps(document.model_dump(mode="json"), indent=2, sort_keys=True) + "\n"
+        self.path.parent.mkdir(parents=True, exist_ok=True)
+        temporary: Path | None = None
+        try:
+            with NamedTemporaryFile(
+                mode="w",
+                encoding="utf-8",
+                dir=self.path.parent,
+                prefix=".evolve-state.",
+                suffix=".tmp",
+                delete=False,
+            ) as stream:
+                temporary = Path(stream.name)
+                stream.write(payload)
+                stream.flush()
+                os.fsync(stream.fileno())
+            os.chmod(temporary, 0o600)
+            os.replace(temporary, self.path)
+        except OSError:
+            if temporary is not None:
+                temporary.unlink(missing_ok=True)
+            # Status is best-effort: a run must not fail because its record did.
+            logger.warning("evolve_state_write_failed")
+        self._document = document
+
+    def _read(self) -> EvolveStateDocument:
+        try:
+            return EvolveStateDocument.model_validate_json(self.path.read_text(encoding="utf-8"))
+        except FileNotFoundError:
+            return EvolveStateDocument()
+        except (OSError, ValueError):
+            # A damaged status file only loses history; start a fresh one.
+            logger.warning("evolve_state_unreadable")
+            return EvolveStateDocument()
+
+
+class EvolveCoordinator(Protocol):
+    """The write coordinator surface evolve and the graph backfill use."""
+
+    def exclusive_for_evolve(self, user_id: str) -> Any: ...
+
+    def use_client(self) -> Any: ...
+
+
+class EvolveService:
+    """Start, serialize and record evolve runs."""
+
+    def __init__(
+        self,
+        *,
+        coordinator: EvolveCoordinator,
+        state: EvolveStateStore,
+        memory_mode: Callable[[], str],
+        max_rounds: int = EVOLVE_MAX_ROUNDS,
+        now: Callable[[], datetime] = lambda: datetime.now(UTC),
+    ) -> None:
+        self.coordinator = coordinator
+        self.state = state
+        self.memory_mode = memory_mode
+        self.max_rounds = max_rounds
+        self._now = now
+        self._runs: dict[str, EvolveRunReceipt] = {}
+        self._tasks: dict[str, asyncio.Task[None]] = {}
+        self._active: dict[str, str] = {}
+
+    def start(
+        self,
+        user_id: str,
+        *,
+        agent_id: str | None = None,
+        max_batches: int | None = None,
+        trigger: Trigger = "manual",
+    ) -> EvolveRunReceipt:
+        """Start a run, or return the one already running for this user."""
+        if self.memory_mode() != "ultra":
+            raise EvolveModeError()
+        running = self._active.get(user_id)
+        if running is not None:
+            return self._runs[running]
+        receipt = EvolveRunReceipt(
+            run_id=str(uuid4()),
+            user_id=user_id,
+            agent_id=agent_id or PRODUCT_AGENT_ID,
+            max_batches=max_batches,
+            trigger=trigger,
+            started_at=self._now(),
+        )
+        self._remember(receipt)
+        self._active[user_id] = receipt.run_id
+        self._tasks[receipt.run_id] = asyncio.create_task(
+            self._run(receipt.run_id), name="one-memory-evolve"
+        )
+        return receipt
+
+    def get(self, run_id: str) -> EvolveRunReceipt | None:
+        return self._runs.get(run_id)
+
+    def running(self) -> tuple[EvolveRunReceipt, ...]:
+        return tuple(self._runs[run_id] for run_id in self._active.values())
+
+    async def wait(self, run_id: str) -> EvolveRunReceipt | None:
+        task = self._tasks.get(run_id)
+        if task is not None:
+            await asyncio.shield(task)
+        return self._runs.get(run_id)
+
+    async def shutdown(self) -> None:
+        tasks = tuple(self._tasks.values())
+        for task in tasks:
+            task.cancel()
+        await asyncio.gather(*tasks, return_exceptions=True)
+        self._tasks.clear()
+
+    def _remember(self, receipt: EvolveRunReceipt) -> None:
+        self._runs[receipt.run_id] = receipt
+        while len(self._runs) > _KEPT_RUNS:
+            oldest = next(iter(self._runs))
+            if oldest in self._active.values():
+                break
+            del self._runs[oldest]
+
+    async def _run(self, run_id: str) -> None:
+        receipt = self._runs[run_id]
+        diagnostic_event(
+            level="info",
+            component="memory.evolve",
+            event="memory_evolve.started",
+            outcome="started",
+            details={"trigger": receipt.trigger, "max_batches": receipt.max_batches},
+        )
+        try:
+            async with (
+                self.coordinator.exclusive_for_evolve(receipt.user_id),
+                self.coordinator.use_client() as client,
+            ):
+                while receipt.rounds < self.max_rounds:
+                    result = await asyncio.to_thread(
+                        client.evolve,
+                        receipt.user_id,
+                        agent_id=receipt.agent_id,
+                        max_batches=receipt.max_batches,
+                    )
+                    receipt = _folded(receipt, result)
+                    self._runs[run_id] = receipt
+                    if receipt.state == "failed" or not receipt.should_continue:
+                        break
+                else:
+                    receipt = receipt.model_copy(update={"capped": True})
+            if receipt.state == "running":
+                receipt = receipt.model_copy(update={"state": "succeeded"})
+        except asyncio.CancelledError:
+            receipt = receipt.model_copy(
+                update={"state": "failed", "error_code": "memory_evolve_cancelled"}
+            )
+            raise
+        except Exception as error:  # noqa: BLE001 - recorded as a stable code
+            code = safe_error_code(error)
+            receipt = receipt.model_copy(
+                update={
+                    "state": "failed",
+                    "error_code": _stable(code, "hy_memory_evolve_failed"),
+                    "failure_kind": type(error).__name__,
+                }
+            )
+        finally:
+            receipt = receipt.model_copy(update={"finished_at": self._now()})
+            self._runs[run_id] = receipt
+            self._active.pop(receipt.user_id, None)
+            self._tasks.pop(run_id, None)
+            await asyncio.to_thread(self.state.record_run, receipt)
+            diagnostic_event(
+                level="info" if receipt.state == "succeeded" else "warning",
+                component="memory.evolve",
+                event="memory_evolve.finished",
+                outcome="succeeded" if receipt.state == "succeeded" else "failed",
+                details={
+                    "trigger": receipt.trigger,
+                    "rounds": receipt.rounds,
+                    "tasks_processed": receipt.tasks_processed,
+                    "batch_failures": receipt.batch_failures,
+                    "capped": receipt.capped,
+                    "elapsed_ms": round(receipt.elapsed_ms, 1),
+                    "error_code": receipt.error_code,
+                },
+            )
+
+
+def _folded(receipt: EvolveRunReceipt, result: object) -> EvolveRunReceipt:
+    """One more evolve round's result added to the run."""
+    if not isinstance(result, Mapping):
+        return receipt.model_copy(
+            update={"state": "failed", "error_code": "hy_memory_evolve_shape_unknown"}
+        )
+    failures = result.get("batch_failures")
+    update: dict[str, object] = {
+        "rounds": receipt.rounds + 1,
+        "tasks_processed": receipt.tasks_processed + _int(result.get("tasks_processed")),
+        "batch_failures": receipt.batch_failures
+        + (len(failures) if isinstance(failures, list | tuple) else _int(failures)),
+        "elapsed_ms": receipt.elapsed_ms + _float(result.get("elapsed_ms")),
+        "should_continue": result.get("should_continue") is True,
+        "cycle_status": _text(result.get("cycle_status")),
+        "continuation_reason": _text(result.get("continuation_reason")),
+        "request_id": _text(result.get("request_id")),
+        "results": result.get("results"),
+        "cross_domain_sweeper": result.get("cross_domain_sweeper"),
+    }
+    for name in _COUNTS:
+        value = result.get(name)
+        update[name] = value if isinstance(value, int) and not isinstance(value, bool) else None
+    if result.get("success") is not True:
+        code = result.get("error_code")
+        kind = result.get("failure_kind")
+        update.update(
+            state="failed",
+            should_continue=False,
+            error_code=(
+                code
+                if isinstance(code, str) and _STABLE_CODE.fullmatch(code)
+                else "hy_memory_evolve_rejected"
+            ),
+            failure_kind=kind if isinstance(kind, str) and _STABLE_CODE.fullmatch(kind) else None,
+        )
+    return receipt.model_copy(update=update)
+
+
+def _stable(code: str, fallback: str) -> str:
+    return code if _STABLE_CODE.fullmatch(code) else fallback
+
+
+def _int(value: object) -> int:
+    return value if isinstance(value, int) and not isinstance(value, bool) else 0
+
+
+def _float(value: object) -> float:
+    if isinstance(value, int | float) and not isinstance(value, bool):
+        return float(value)
+    return 0.0
+
+
+def _text(value: object) -> str | None:
+    return value[:128] if isinstance(value, str) else None
+
+
+def evolve_due(
+    now: datetime,
+    schedule: EvolveSchedule,
+    last_run_at: datetime | None,
+    anchor: datetime,
+) -> bool:
+    """Whether a user's nightly evolve is due at `now` (aware, local time).
+
+    Due once the day's slot (`schedule.time`, local) has passed and the last
+    run (or, for a user who never ran, the anchor) is older than that slot.
+    A last run more than 24 hours old is due at once as well: a machine asleep
+    at the slot catches up when it wakes, even before the next day's slot.
+    """
+    if not schedule.enabled:
+        return False
+    local = now.astimezone()
+    hour, minute = (int(part) for part in schedule.time.split(":"))
+    slot = local.replace(hour=hour, minute=minute, second=0, microsecond=0)
+    reference = last_run_at or anchor
+    if local >= slot and reference < slot:
+        return True
+    return last_run_at is not None and local - last_run_at > timedelta(hours=24)
+
+
+class EvolveScheduler:
+    """Run evolve nightly for every user that has memories (ultra only)."""
+
+    def __init__(
+        self,
+        *,
+        service: EvolveService,
+        schedule: Callable[[], EvolveSchedule],
+        users_with_memories: Callable[[], Awaitable[Iterable[str]]],
+        now: Callable[[], datetime] = lambda: datetime.now().astimezone(),
+        check_seconds: float = SCHEDULE_CHECK_SECONDS,
+    ) -> None:
+        self.service = service
+        self.schedule = schedule
+        self.users_with_memories = users_with_memories
+        self._now = now
+        self.check_seconds = check_seconds
+        self._task: asyncio.Task[None] | None = None
+
+    def start(self) -> None:
+        if self._task is None or self._task.done():
+            self._task = asyncio.create_task(self._loop(), name="one-memory-evolve-schedule")
+
+    async def tick(self) -> tuple[str, ...]:
+        """Run every due user once, one after another; the users run."""
+        schedule = self.schedule()
+        if not schedule.enabled or self.service.memory_mode() != "ultra":
+            return ()
+        now = self._now()
+        anchor = await asyncio.to_thread(self.service.state.anchor, now)
+        ran: list[str] = []
+        for user_id in await self.users_with_memories():
+            last = self.service.state.user(user_id).last_run_at
+            if not evolve_due(now, schedule, last, anchor):
+                continue
+            try:
+                receipt = self.service.start(user_id, trigger="scheduled")
+            except EvolveModeError:
+                break
+            await self.service.wait(receipt.run_id)
+            ran.append(user_id)
+        if ran:
+            diagnostic_event(
+                level="info",
+                component="memory.evolve",
+                event="memory_evolve.schedule_ran",
+                outcome="succeeded",
+                details={"users": len(ran)},
+            )
+        return tuple(ran)
+
+    async def _loop(self) -> None:
+        while True:
+            try:
+                await self.tick()
+            except asyncio.CancelledError:
+                raise
+            except Exception as error:  # noqa: BLE001 - the loop must survive
+                logger.warning("evolve_schedule_tick_failed error_type=%s", type(error).__name__)
+            await asyncio.sleep(self.check_seconds)
+
+    async def shutdown(self) -> None:
+        if self._task is None:
+            return
+        self._task.cancel()
+        await asyncio.gather(self._task, return_exceptions=True)
+        self._task = None
+
+
+class MemoryGraphBackfill:
+    """After a switch to ultra, project each user's existing facts into the graph once."""
+
+    def __init__(
+        self,
+        *,
+        coordinator: EvolveCoordinator,
+        state: EvolveStateStore,
+        now: Callable[[], datetime] = lambda: datetime.now(UTC),
+    ) -> None:
+        self.coordinator = coordinator
+        self.state = state
+        self._now = now
+        self._task: asyncio.Task[None] | None = None
+
+    @property
+    def running(self) -> bool:
+        return self._task is not None and not self._task.done()
+
+    def start(self, user_ids: Iterable[str]) -> GraphBuildState:
+        if self.running:
+            return self.state.graph_build()
+        users = tuple(dict.fromkeys(user_ids))
+        state = GraphBuildState(
+            state="running",
+            started_at=self._now(),
+            users={user_id: GraphBuildUser() for user_id in users},
+        )
+        self.state.set_graph_build(state)
+        self._task = asyncio.create_task(self._run(users), name="one-memory-graph-backfill")
+        return state
+
+    async def wait(self) -> None:
+        if self._task is not None:
+            await asyncio.shield(self._task)
+
+    async def _run(self, users: tuple[str, ...]) -> None:
+        state = self.state.graph_build()
+        for user_id in users:
+            state.users[user_id] = GraphBuildUser(state="running")
+            await asyncio.to_thread(self.state.set_graph_build, state.model_copy(deep=True))
+            try:
+                async with (
+                    self.coordinator.exclusive_for_evolve(user_id),
+                    self.coordinator.use_client() as client,
+                ):
+                    result = await asyncio.to_thread(
+                        client.build_memory_graph,
+                        user_id,
+                        agent_id=None,
+                        limit=10000,
+                        clear_first=False,
+                    )
+                ok = isinstance(result, Mapping) and result.get("success") is True
+                state.users[user_id] = GraphBuildUser(
+                    state="succeeded" if ok else "failed",
+                    memories_scanned=_int(result.get("memories_scanned"))
+                    if isinstance(result, Mapping)
+                    else None,
+                    nodes=_int(result.get("nodes")) if isinstance(result, Mapping) else None,
+                    error_code=None if ok else "hy_memory_graph_build_rejected",
+                )
+            except asyncio.CancelledError:
+                raise
+            except Exception as error:  # noqa: BLE001 - recorded as a stable code
+                code = safe_error_code(error)
+                state.users[user_id] = GraphBuildUser(
+                    state="failed",
+                    error_code=_stable(code, "hy_memory_graph_build_failed"),
+                )
+        failed = any(user.state == "failed" for user in state.users.values())
+        state.state = "failed" if failed else "succeeded"
+        state.finished_at = self._now()
+        await asyncio.to_thread(self.state.set_graph_build, state)
+        diagnostic_event(
+            level="warning" if failed else "info",
+            component="memory.evolve",
+            event="memory_graph_backfill.finished",
+            outcome="failed" if failed else "succeeded",
+            details={
+                "users": len(state.users),
+                "failed_users": sum(user.state == "failed" for user in state.users.values()),
+            },
+        )
+
+    async def shutdown(self) -> None:
+        if self._task is None:
+            return
+        self._task.cancel()
+        await asyncio.gather(self._task, return_exceptions=True)
+        self._task = None

memhub/apps/core/src/one_memory/memory_server/config_state.py

@@ -13,6 +13,7 @@ import threading
 from collections.abc import Mapping
 from pathlib import Path
 from tempfile import NamedTemporaryFile
+from typing import Literal
 
 from pydantic import BaseModel, ConfigDict, Field, StrictBool, field_validator
 
@@ -27,6 +28,31 @@ class SourceConfig(BaseModel):
     sync_enabled: StrictBool = False
 
 
+#: The SDK pipeline mode (hy-memory `HyMemoryClient(mode=...)`):
+#: lite keeps raw conversations only, with no model call; pro extracts
+#: memories (System 1); ultra adds the knowledge graph and evolve (System 2).
+MemoryMode = Literal["lite", "pro", "ultra"]
+MEMORY_MODES: tuple[MemoryMode, ...] = ("lite", "pro", "ultra")
+_SCHEDULE_TIME = re.compile(r"^(?:[01][0-9]|2[0-3]):[0-5][0-9]$")
+
+
+class EvolveSchedule(BaseModel):
+    """When ultra runs evolve for every user: once a day, local time."""
+
+    model_config = ConfigDict(frozen=True, extra="forbid", strict=True)
+
+    enabled: StrictBool = True
+    #: `HH:MM`, 24-hour, the machine's local time.
+    time: str = "03:00"
+
+    @field_validator("time")
+    @classmethod
+    def valid_time(cls, value: str) -> str:
+        if _SCHEDULE_TIME.fullmatch(value) is None:
+            raise ValueError("invalid_evolve_schedule_time")
+        return value
+
+
 class ConfigDocument(BaseModel):
     model_config = ConfigDict(frozen=True, extra="forbid", strict=True)
 
@@ -39,6 +65,10 @@ class ConfigDocument(BaseModel):
     #: needs no re-entry of a key.
     sedimentation_enabled: StrictBool = True
     extraction_window_preset: ExtractionWindowPreset = ExtractionWindowPreset.THREE_MONTHS
+    #: The SDK mode every client is built with. `ONE_MEMORY_GRAPH_STORE` (dev)
+    #: still forces ultra over it; see `hy_graph_admin.effective_memory_mode`.
+    memory_mode: MemoryMode = "pro"
+    evolve_schedule: EvolveSchedule = Field(default_factory=EvolveSchedule)
 
     @field_validator("sources")
     @classmethod
@@ -61,6 +91,9 @@ class ConfigFileError(RuntimeError):
     pass
 
 
+_WRITTEN_WHEN_SET = frozenset({"memory_mode", "evolve_schedule"})
+
+
 class ConfigStore:
     """The only writer of ``<data_dir>/config.json`` inside Core."""
 
@@ -107,6 +140,32 @@ class ConfigStore:
             self._document = updated
             return updated
 
+    def memory_mode(self) -> MemoryMode:
+        with self._lock:
+            return self._document.memory_mode
+
+    def set_memory_mode(self, mode: MemoryMode) -> ConfigDocument:
+        if mode not in MEMORY_MODES:
+            raise ValueError("invalid_memory_mode")
+        with self._lock:
+            updated = self._document.model_copy(update={"memory_mode": mode})
+            self._write(updated)
+            self._document = updated
+            return updated
+
+    def evolve_schedule(self) -> EvolveSchedule:
+        with self._lock:
+            return self._document.evolve_schedule
+
+    def set_evolve_schedule(self, schedule: EvolveSchedule) -> ConfigDocument:
+        if not isinstance(schedule, EvolveSchedule):
+            raise ValueError("invalid_evolve_schedule")
+        with self._lock:
+            updated = self._document.model_copy(update={"evolve_schedule": schedule})
+            self._write(updated)
+            self._document = updated
+            return updated
+
     def migrate_extraction_window_preset(self, preset: ExtractionWindowPreset) -> ConfigDocument:
         """Adopt the legacy SQL value only when config.json has no value."""
         if not isinstance(preset, ExtractionWindowPreset):
@@ -152,9 +211,12 @@ class ConfigStore:
             raise ConfigFileError("config_invalid") from error
 
     def _write(self, document: ConfigDocument) -> None:
+        # Settings added after 1.0.5 are written only once a person chose
+        # them, so an untouched install's file stays what older builds read.
+        unset = {name for name in _WRITTEN_WHEN_SET if name not in document.model_fields_set}
         payload = (
             json.dumps(
-                document.model_dump(mode="json"),
+                document.model_dump(mode="json", exclude=unset),
                 ensure_ascii=False,
                 indent=2,
                 sort_keys=True,

memhub/apps/core/src/one_memory/memory_server/contracts.py

@@ -75,7 +75,11 @@ def batch_id(*, internal_session_id: str, first_round: int, last_round: int) ->
 INTERNAL_SESSION_ID_MAX = 128
 
 
-def internal_session_id_for(conversation_id: str, source_id: str) -> str:
+def internal_session_id_for(
+    conversation_id: str,
+    source_id: str,
+    user_id: str = "local",
+) -> str:
     """Normalise a conversation identity to the durable ``internal_session_id``.
 
     This is the single derivation shared by every Evidence-first intake path.
@@ -86,7 +90,17 @@ def internal_session_id_for(conversation_id: str, source_id: str) -> str:
     per-session serialisation split). Only a value that the contract would have
     rejected before it ever reached the queue is replaced by the versioned
     session id, which is deterministic for a given (source, conversation).
+
+    Another user's conversation is always derived with the user in it: two
+    Hermes profiles can hold the same session id, and sharing a session key
+    would serialise and batch one person's turns behind the other's. `local`
+    keeps the identity it always had.
     """
+    if user_id != "local":
+        return session_id(
+            harness_id=source_id,
+            external_session_id=f"user:{user_id}|{conversation_id}",
+        )
     if len(conversation_id) <= INTERNAL_SESSION_ID_MAX:
         return conversation_id
     derived = session_id(harness_id=source_id, external_session_id=conversation_id)

memhub/apps/core/src/one_memory/memory_server/durable_queue.py

@@ -54,6 +54,7 @@ from one_memory.memory_tasks import (
 )
 from one_memory.sdk_migration_log import error_chain_text
 from one_memory.service_levels import ServiceLevel, service_level_context
+from one_memory.user_scope import DEFAULT_USER_ID, lane_key, other_user_lane_prefix
 
 logger = logging.getLogger(__name__)
 
@@ -345,6 +346,27 @@ EVENT_TIME_WINDOW = timedelta(minutes=10)
 DIRECT_WRITE_SOURCES = frozenset({"one-memory-manual"})
 
 
+def request_user_id(request: WriteRequest) -> str:
+    """Whose write this is: its frozen input's user, `local` for id-only rows."""
+    users = {record.user_id for record in request.input_records}
+    if len(users) > 1:
+        raise ValueError("durable_write_multiple_users")
+    return next(iter(users), DEFAULT_USER_ID)
+
+
+def request_source_key(request: WriteRequest) -> str:
+    """The FIFO/claim key of a write: its source, per user.
+
+    `local` keeps the bare source id it always had. Another user's Hermes
+    lane is `hermes@user`, so one person's running job never holds the other
+    person's queue and the claim gate answers for that user's grant.
+    """
+    return lane_key(
+        request_user_id(request),
+        request.source_id or request.internal_session_id,
+    )
+
+
 @dataclass(frozen=True)
 class DurableWriteRecord:
     job_id: str
@@ -862,7 +884,7 @@ class DurableWriteQueue:
                     job_id,
                     queue_seq,
                     request.internal_session_id,
-                    request.source_id or request.internal_session_id,
+                    request_source_key(request),
                     request.batch_identity.batch_id,
                     key,
                     payload,
@@ -929,7 +951,9 @@ class DurableWriteQueue:
             f"{hashlib.sha256(file_key.encode()).hexdigest()}:"
             f"{hashlib.sha256(record.conversation_id.encode()).hexdigest()}"
         )
-        session_key = internal_session_id_for(record.conversation_id, record.source_id)
+        session_key = internal_session_id_for(
+            record.conversation_id, record.source_id, record.user_id
+        )
         written = connection.execute(
             "SELECT 1 FROM memory_write_jobs AS job, json_each(job.evidence_ids_json) AS ref "
             "WHERE job.session_key=? AND job.state!='assembling' AND ref.value=? LIMIT 1",
@@ -998,7 +1022,7 @@ class DurableWriteQueue:
                 str(uuid4()),
                 queue_seq,
                 session_key,
-                record.source_id,
+                lane_key(record.user_id, record.source_id),
                 assembly_key,
                 assembly_key,
                 payload_json,
@@ -1034,10 +1058,13 @@ class DurableWriteQueue:
         `flush` (history at end of file) completes the last turn and writes a
         short batch; without it (live polling) those stay in the row.
         """
+        # Two users' roots can hold a file with the same relative key (each
+        # Hermes profile has its own `state.db`): only this user's rows.
+        assembly_prefix = f"assembly:{user_id}:{source_id}:"
         rows = connection.execute(
             "SELECT * FROM memory_write_jobs WHERE state='assembling' "
-            "AND source_key=? ORDER BY queue_seq",
-            (source_id,),
+            "AND source_key=? AND substr(idempotency_key,1,?)=? ORDER BY queue_seq",
+            (lane_key(user_id, source_id), len(assembly_prefix), assembly_prefix),
         ).fetchall()
         finalized: list[str] = []
         for row in rows:
@@ -1160,7 +1187,7 @@ class DurableWriteQueue:
         Same records, same key: a re-read returns the job already written.
         """
         first = records[0]
-        session_id = internal_session_id_for(first.conversation_id, first.source_id)
+        session_id = internal_session_id_for(first.conversation_id, first.source_id, first.user_id)
         material = ":".join(f"{item.evidence_id}:{item.content_hash}" for item in records)
         request_id = f"turn:v2:{hashlib.sha256(material.encode()).hexdigest()}"
         existing = connection.execute(
@@ -1205,7 +1232,7 @@ class DurableWriteQueue:
                 job_id,
                 queue_seq,
                 session_id,
-                first.source_id,
+                lane_key(first.user_id, first.source_id),
                 request.batch_identity.batch_id,
                 request_id,
                 json.dumps(
@@ -2494,6 +2521,12 @@ async def execute_job(
             source_id = request.internal_session_id
         if source_ids and source_ids != {source_id}:
             raise RuntimeError("durable_write_multiple_sources")
+        users = {getattr(record, "user_id", DEFAULT_USER_ID) for record in records}
+        if len(users) > 1:
+            raise RuntimeError("durable_write_multiple_users")
+        # The gates answer per user: another user's Hermes job asks about
+        # that user's grant, not `local`'s.
+        source_id = lane_key(next(iter(users), DEFAULT_USER_ID), source_id)
 
         async def mark() -> bool:
             return await in_the_background(queue.mark_invoked, job_id, owner=owner)
@@ -2530,6 +2563,8 @@ async def execute_job(
             in {
                 "model_content_source_disabled",
                 "model_sedimentation_disabled",
+                # Evolve holds this user; the job runs once it is done.
+                "memory_evolve_in_progress",
             }
             and not hy_invoked
         ):
@@ -2799,6 +2834,19 @@ class DurableWriteDrain:
                 source_clause = f"AND candidate.source_key IN ({placeholders}) "
                 source_params = tuple(sorted(enabled_sources))
             direct_sources = ",".join("?" for _ in DIRECT_WRITE_SOURCES)
+            # Another user's hand-added memories are `one-memory-manual@user`:
+            # the same lane ahead of history, still serial per user.
+            direct_lanes = " OR ".join(
+                "substr(candidate.source_key,1,?)=?" for _ in DIRECT_WRITE_SOURCES
+            )
+            direct_lane_params = tuple(
+                value
+                for source in sorted(DIRECT_WRITE_SOURCES)
+                for value in (
+                    len(other_user_lane_prefix(source)),
+                    other_user_lane_prefix(source),
+                )
+            )
             # Same eligibility as the window claim below (free lease, no
             # running job of its source, nothing earlier in its session), but
             # not bound to the window: a note must not wait for history.
@@ -2808,7 +2856,7 @@ class DurableWriteDrain:
                 "WHERE candidate.state=? "
                 "AND (candidate.lease_expires_at IS NULL OR candidate.lease_expires_at < ?) "
                 + source_clause
-                + f"AND candidate.source_key IN ({direct_sources}) "
+                + f"AND (candidate.source_key IN ({direct_sources}) OR {direct_lanes}) "
                 "AND NOT EXISTS ("
                 "SELECT 1 FROM memory_write_jobs AS held "
                 "WHERE held.source_key=candidate.source_key AND held.state=? "
@@ -2826,6 +2874,7 @@ class DurableWriteDrain:
                     now.isoformat(),
                     *source_params,
                     *sorted(DIRECT_WRITE_SOURCES),
+                    *direct_lane_params,
                     RUNNING,
                     now.isoformat(),
                     *_FIFO_BLOCKING_STATES,

memhub/apps/core/src/one_memory/memory_server/routes.py

@@ -19,10 +19,11 @@ does the writing. That separation is D52.
 
 from __future__ import annotations
 
+import hashlib
 from datetime import UTC, datetime
 from typing import Any, Literal, cast
 
-from fastapi import APIRouter, HTTPException, Request
+from fastapi import APIRouter, HTTPException, Query, Request
 from pydantic import BaseModel, ConfigDict, Field
 
 from one_memory.imports.models import (
@@ -43,6 +44,24 @@ from one_memory.memory_server.durable_queue import (
 )
 from one_memory.memory_server.registry import WriteRegistry, WriteState
 from one_memory.memory_server.submit import WriteConflictError
+from one_memory.user_scope import DEFAULT_USER_ID, USER_ID_PATTERN, UserId
+
+#: Durable keys of another user's writes start with this. A `local` request
+#: id never may, so `local` cannot name another user's receipt.
+_USER_KEY_PREFIX = "user:"
+
+
+def durable_request_key(request_id: str, user_id: str) -> str:
+    """The queue key of a caller's request id, namespaced per user.
+
+    Two users choosing the same request id must not collide: the key is the
+    queue's idempotency key and the SDK's `request_id`. `local` keeps its own
+    request id unchanged.
+    """
+    if user_id == DEFAULT_USER_ID:
+        return request_id
+    digest = hashlib.sha256(request_id.encode()).hexdigest()
+    return f"{_USER_KEY_PREFIX}{user_id}:{digest}"
 
 
 class MemoryWriteReceipt(BaseModel):
@@ -71,6 +90,7 @@ class SdkMemoryWriteInput(BaseModel):
     messages: list[OpenAIMessage] = Field(min_length=1, max_length=200)
     memory_at: datetime | None = None
     priority: Literal["live", "normal"] = "normal"
+    user_id: UserId = DEFAULT_USER_ID
 
 
 def _registry(
@@ -124,6 +144,9 @@ def sdk_router(dependencies: list[Any] | None = None) -> APIRouter:
     ) -> MemoryWriteReceipt:
         if not any(message.role == "user" for message in payload.messages):
             raise HTTPException(422, "messages_require_user_turn")
+        if payload.user_id == DEFAULT_USER_ID and payload.request_id.startswith(_USER_KEY_PREFIX):
+            raise HTTPException(422, "request_id_reserved_prefix")
+        request_key = durable_request_key(payload.request_id, payload.user_id)
         observed_at = payload.memory_at or datetime(1970, 1, 1, tzinfo=UTC)
         written_by = {
             "user": WrittenBy.PERSON,
@@ -133,6 +156,7 @@ def sdk_router(dependencies: list[Any] | None = None) -> APIRouter:
         records = tuple(
             materialize_write_input(
                 EvidenceInput(
+                    user_id=payload.user_id,
                     source_id=payload.source_id,
                     conversation_id=payload.session_id,
                     message_id=f"{payload.request_id}:{index}",
@@ -157,6 +181,7 @@ def sdk_router(dependencies: list[Any] | None = None) -> APIRouter:
         internal_session_id = internal_session_id_for(
             payload.session_id,
             payload.source_id,
+            payload.user_id,
         )
         contract = WriteRequest(
             internal_session_id=internal_session_id,
@@ -168,7 +193,7 @@ def sdk_router(dependencies: list[Any] | None = None) -> APIRouter:
             ),
             evidence_ids=[record.evidence_id for record in records],
             input_records=records,
-            idempotency_key=payload.request_id,
+            idempotency_key=request_key,
             priority=payload.priority,
             event_at=payload.memory_at,
             time_quality="exact" if payload.memory_at is not None else "approximate",
@@ -178,7 +203,7 @@ def sdk_router(dependencies: list[Any] | None = None) -> APIRouter:
         except (WriteConflictError, ValueError) as error:
             raise HTTPException(409, str(error)) from error
         return MemoryWriteReceipt(
-            request_id=record.request_id,
+            request_id=payload.request_id,
             state=record.state,
             created_at=record.created_at,
             updated_at=record.updated_at,
@@ -186,16 +211,24 @@ def sdk_router(dependencies: list[Any] | None = None) -> APIRouter:
         )
 
     @instance.get("/{request_id}", response_model=MemoryWriteReceipt)
-    def status(request_id: str, request: Request) -> MemoryWriteReceipt:
+    def status(
+        request_id: str,
+        request: Request,
+        user_id: str = Query(default=DEFAULT_USER_ID, pattern=USER_ID_PATTERN),
+    ) -> MemoryWriteReceipt:
+        # A receipt belongs to the user that submitted it. Another user's key
+        # is namespaced, and `local` may not name a namespaced key directly.
+        if user_id == DEFAULT_USER_ID and request_id.startswith(_USER_KEY_PREFIX):
+            raise HTTPException(404, "unknown request id")
         registry = _registry(request, write=False)
         try:
-            record = registry.get(request_id)
+            record = registry.get(durable_request_key(request_id, user_id))
         except DurableQueueSchemaUnavailableError as error:
             raise HTTPException(503, str(error)) from error
         if record is None:
             raise HTTPException(404, "unknown request id")
         return MemoryWriteReceipt(
-            request_id=record.request_id,
+            request_id=request_id,
             state=record.state,
             created_at=record.created_at,
             updated_at=record.updated_at,

memhub/apps/core/src/one_memory/memory_server/submit.py

@@ -271,7 +271,9 @@ def build_request_from_evidence(
         # Keep the exact conversation id when it fits the 128-char contract;
         # otherwise derive the versioned session id, so a long branch id does
         # not reject the whole history import.
-        internal_session_id=internal_session_id_for(seed.conversation_id, seed.source_id),
+        internal_session_id=internal_session_id_for(
+            seed.conversation_id, seed.source_id, seed.user_id
+        ),
         first_round=first_round,
         last_round=last_round,
         evidence_ids=list(ordered),

memhub/apps/core/src/one_memory/memory_server/unified_entry.py

@@ -28,6 +28,7 @@ from one_memory.memory_server.contracts import (
     internal_session_id_for,
 )
 from one_memory.memory_server.durable_queue import DurableWriteQueue, DurableWriteRecord
+from one_memory.user_scope import lane_key
 
 WritePriority = Literal["live", "normal"]
 
@@ -74,7 +75,7 @@ class UnifiedWriteIntake:
         if priority not in {"live", "normal"}:
             raise ValueError("priority must be 'live' or 'normal'")
         if self.source_writes_permitted is not None and not self.source_writes_permitted(
-            record.source_id
+            lane_key(record.user_id, record.source_id)
         ):
             # Freeze is the product's one Harness processing decision. Refuse
             # new intake so a frozen source creates no new durable write row;
@@ -106,7 +107,9 @@ class UnifiedWriteIntake:
         last_round: int,
         priority: WritePriority,
     ) -> WriteRequest:
-        session_id = internal_session_id_for(record.conversation_id, record.source_id)
+        session_id = internal_session_id_for(
+            record.conversation_id, record.source_id, record.user_id
+        )
         event_at, time_quality = _job_time((record,))
         request_key = (
             f"entry:{record.evidence_id}"
@@ -145,14 +148,14 @@ class UnifiedWriteIntake:
         ):
             raise ValueError("records must belong to one conversation")
         if self.source_writes_permitted is not None and not self.source_writes_permitted(
-            first.source_id
+            lane_key(first.user_id, first.source_id)
         ):
             raise SourceWriteDisabledError(first.evidence_id)
         material = ":".join(
             f"{record.evidence_id}:{record.revision}:{record.content_hash}" for record in records
         )
         request_id = f"turn:v1:{sha256(material.encode()).hexdigest()}"
-        session_id = internal_session_id_for(first.conversation_id, first.source_id)
+        session_id = internal_session_id_for(first.conversation_id, first.source_id, first.user_id)
         event_at, time_quality = _job_time(records)
         durable = self.queue.submit(
             WriteRequest(

memhub/apps/core/src/one_memory/memory_visibility.py

@@ -160,7 +160,7 @@ def pending(db: sqlite3.Connection, *, limit: int = 100) -> list[str]:
 
 
 def _trigger_sql() -> list[tuple[str, str]]:
-    """(table it depends on, statement) for each trigger."""
+    """(table it is on, statement) for each trigger."""
     return [
         (
             "evidence",
@@ -303,9 +303,61 @@ def install_triggers(db: sqlite3.Connection) -> None:
     for table, statement in _trigger_sql():
         if table not in present:
             continue
+        if not _body_tables(statement) <= present:
+            # SQLite resolves a trigger body only when it fires, so a body
+            # naming a dropped table fails the write it rides on (#233: the
+            # SDK cutover drops `governed_memory_evidence`, and every grant
+            # write then failed). Not installed, and an old copy is removed.
+            cursor.execute(f"DROP TRIGGER IF EXISTS {_trigger_name(statement)}")
+            continue
         cursor.execute(statement)
 
 
+def _trigger_name(statement: str) -> str:
+    return statement.split("IF NOT EXISTS", 1)[1].split()[0]
+
+
+def _body_tables(statement: str) -> set[str]:
+    """Tables the trigger body reads or writes besides the one it is on."""
+    _head, _begin, body = statement.partition("BEGIN")
+    words = body.replace("(", " ").replace(")", " ").split()
+    return {
+        name
+        for keyword, name in zip(words, words[1:], strict=False)
+        if keyword.upper() in {"FROM", "JOIN", "INTO"}
+    }
+
+
+def drop_stale_triggers(database_path: Path) -> tuple[str, ...]:
+    """Drop this module's triggers whose body names a table that is gone.
+
+    Startup on SDK authority does not install the cache (it is read only on
+    legacy authority), but a home migrated by an earlier build still carries
+    the triggers it installed. One whose body reads a dropped table fails
+    every write to the table it is on, so it is dropped here. Idempotent.
+    """
+    with main_database(database_path).write() as db, db:
+        cursor = db.cursor()
+        cursor.row_factory = sqlite3.Row
+        present = {
+            str(row["name"])
+            for row in cursor.execute("SELECT name FROM sqlite_master WHERE type='table'")
+        }
+        installed = {
+            str(row["name"])
+            for row in cursor.execute("SELECT name FROM sqlite_master WHERE type='trigger'")
+        }
+        dropped = tuple(
+            name
+            for _table, statement in _trigger_sql()
+            if (name := _trigger_name(statement)) in installed
+            and not _body_tables(statement) <= present
+        )
+        for name in dropped:
+            cursor.execute(f"DROP TRIGGER IF EXISTS {name}")
+    return dropped
+
+
 def startup(database_path: Path) -> None:
     """Idempotent: create the derived tables and triggers. No version change.
 

memhub/apps/core/src/one_memory/model_probe.py

@@ -564,7 +564,7 @@ def _embedding_request(
     secret: str | None,
 ) -> tuple[str, str, dict[str, str], dict[str, object]]:
     provider = profile.embedding_provider_kind or profile.provider_kind
-    base_url = (profile.embedding_base_url or profile.base_url or "").rstrip("/")
+    base_url = (profile.embedding_base_url or "").rstrip("/")
     if provider is ProviderKind.OLLAMA:
         return (
             "POST",

memhub/apps/core/src/one_memory/model_profiles.py

@@ -97,7 +97,6 @@ class ModelProfileInput(BaseModel):
         )
         embedding_provider, resolved_embedding_base_url = resolve_embedding_endpoint(
             provider_kind=self.provider_kind,
-            base_url=base_url,
             embedding_model=embedding_model,
             embedding_provider_kind=embedding_provider_kind,
             embedding_base_url=embedding_base_url,
@@ -107,6 +106,12 @@ class ModelProfileInput(BaseModel):
             ProviderKind.ANTHROPIC,
         }:
             raise ValueError("embedding provider must be OpenAI-compatible, Ollama, or local ONNX")
+        if (
+            embedding_provider is not ProviderKind.LOCAL_ONNX
+            and resolved_embedding_base_url is None
+        ):
+            # `ModelProfileStore.create` refuses it with its own code.
+            return self
         if (
             embedding_provider is not ProviderKind.LOCAL_ONNX
             and resolved_embedding_base_url is not None
@@ -308,11 +313,12 @@ class ModelProfileStore:
                     f"""
                     UPDATE {table}
                     SET embedding_provider_kind = provider_kind,
-                        embedding_base_url = base_url
+                        embedding_base_url = COALESCE(embedding_base_url, base_url)
                     WHERE embedding_model IS NOT NULL
                       AND embedding_provider_kind IS NULL
                     """
                 )
+                _pin_embedding_endpoints(connection, table)
             connection.execute(
                 """
                 INSERT OR IGNORE INTO model_profile_versions (
@@ -355,11 +361,11 @@ class ModelProfileStore:
         )
         embedding_provider_kind, embedding_base_url = resolve_embedding_endpoint(
             provider_kind=value.provider_kind,
-            base_url=base_url,
             embedding_model=defaulted_model,
             embedding_provider_kind=defaulted_provider,
             embedding_base_url=defaulted_base_url,
         )
+        require_embedding_base_url(embedding_provider_kind, embedding_base_url)
         try:
             with self._connect(writing=True) as connection:
                 connection.execute(
@@ -612,7 +618,6 @@ class ModelProfileStore:
         )
         embedding_provider_kind, embedding_base_url = resolve_embedding_endpoint(
             provider_kind=current.provider_kind,
-            base_url=base_url,
             embedding_model=embedding_model,
             embedding_provider_kind=(
                 patch.embedding_provider_kind
@@ -627,6 +632,7 @@ class ModelProfileStore:
         )
         if embedding_provider_kind in {ProviderKind.HY_DEFAULT, ProviderKind.ANTHROPIC}:
             raise ValueError("embedding provider must be OpenAI-compatible, Ollama, or local ONNX")
+        require_embedding_base_url(embedding_provider_kind, embedding_base_url)
         if (
             embedding_provider_kind is not ProviderKind.LOCAL_ONNX
             and embedding_base_url is not None
@@ -1167,21 +1173,77 @@ def attach_disable_thinking(
 def resolve_embedding_endpoint(
     *,
     provider_kind: ProviderKind,
-    base_url: str | None,
     embedding_model: str | None,
     embedding_provider_kind: ProviderKind | None,
     embedding_base_url: str | None,
 ) -> tuple[ProviderKind | None, str | None]:
+    """The embedding provider and the endpoint exactly as given.
+
+    The endpoint never follows the generation `base_url`: the collection name
+    hashes it, so a proxy or region edit of the generation endpoint moved
+    writes to a new, empty collection, and providers without embeddings were
+    sent embedding requests. `require_embedding_base_url` refuses a remote
+    embedding without one.
+    """
     if embedding_model is None:
         return None, None
     if embedding_provider_kind is ProviderKind.LOCAL_ONNX:
         return ProviderKind.LOCAL_ONNX, None
-    if embedding_provider_kind is None:
-        return provider_kind, embedding_base_url or base_url
-    return (
-        embedding_provider_kind,
-        embedding_base_url or DEFAULT_BASE_URLS.get(embedding_provider_kind),
-    )
+    return embedding_provider_kind or provider_kind, embedding_base_url
+
+
+class EmbeddingBaseUrlRequiredError(ValueError):
+    def __init__(self) -> None:
+        super().__init__("embedding_base_url_required")
+
+
+def require_embedding_base_url(provider: ProviderKind | None, base_url: str | None) -> None:
+    """A remote embedding must name its endpoint; the UI may offer the generation one."""
+    if provider not in {None, ProviderKind.LOCAL_ONNX} and not base_url:
+        raise EmbeddingBaseUrlRequiredError()
+
+
+def pinned_embedding_base_url(
+    *,
+    provider_kind: ProviderKind | str,
+    base_url: str | None,
+    embedding_model: str | None,
+    embedding_provider_kind: ProviderKind | str | None,
+    embedding_base_url: str | None,
+) -> str | None:
+    """The endpoint a stored profile's embedding resolved to before it was pinned.
+
+    The Runtime used `embedding_base_url or base_url`, and the provider default
+    only when both were empty (no collection could have existed then). Writing
+    that value keeps the endpoint, and so the collection name, byte-identical.
+    """
+    provider = ProviderKind(embedding_provider_kind or provider_kind)
+    if embedding_model is None or embedding_base_url or provider is ProviderKind.LOCAL_ONNX:
+        return embedding_base_url
+    return base_url or DEFAULT_BASE_URLS.get(provider)
+
+
+def _pin_embedding_endpoints(connection: sqlite3.Connection, table: str) -> None:
+    """Write the resolved endpoint into every remote embedding left without one."""
+    rows = connection.execute(
+        f"SELECT rowid, provider_kind, base_url, embedding_model, embedding_provider_kind "
+        f"FROM {table} WHERE embedding_model IS NOT NULL AND embedding_base_url IS NULL "
+        "AND COALESCE(embedding_provider_kind, provider_kind) != 'local_onnx'"
+    ).fetchall()
+    for row in rows:
+        connection.execute(
+            f"UPDATE {table} SET embedding_base_url = ? WHERE rowid = ?",
+            (
+                pinned_embedding_base_url(
+                    provider_kind=row["provider_kind"],
+                    base_url=row["base_url"],
+                    embedding_model=row["embedding_model"],
+                    embedding_provider_kind=row["embedding_provider_kind"],
+                    embedding_base_url=None,
+                ),
+                row["rowid"],
+            ),
+        )
 
 
 def validate_model_base_url(value: str) -> None:

memhub/apps/core/src/one_memory/recall.py

@@ -14,6 +14,7 @@ from pydantic import BaseModel, ConfigDict, Field
 
 from one_memory.hy_sdk_types import HySdkMemoryRecord
 from one_memory.memory_governance import GovernedMemory, MemoryStatus
+from one_memory.sdk_search_options import SdkSearchOptions
 from one_memory.semantic_deadline import (
     SemanticMemoryHit,
     SemanticSearch,
@@ -22,6 +23,7 @@ from one_memory.semantic_deadline import (
     normalize_semantic_hits,
     search_within,
 )
+from one_memory.user_scope import DEFAULT_USER_ID, UserId
 
 
 def recall_blend_mode() -> str:
@@ -49,7 +51,15 @@ class TruncationBudget(StrEnum):
     TIMEOUT = "timeout_budget"
 
 
-class RecallRequest(BaseModel):
+class RecallRequest(SdkSearchOptions):
+    """`POST /api/v1/recall`.
+
+    The `SdkSearchOptions` fields, `schema_limit` and `intention_limit` pass
+    through to the SDK memory search (and the floor and row filters to the
+    raw fallback). Absent, recall makes the search it always made: the
+    relevance floor, schema and intention lanes sized to `limit`.
+    """
+
     model_config = ConfigDict(frozen=True, extra="forbid")
 
     query: str = Field(min_length=1, max_length=2_000)
@@ -67,6 +77,19 @@ class RecallRequest(BaseModel):
     #: Wall-clock budget for candidate gather + score. Not a substitute for
     #: raising the old newest-500 page; it is the third parallel budget.
     timeout_ms: int = Field(default=2_000, ge=50, le=30_000)
+    #: Whose memories to recall. Omitted, it is `local`, the App's library.
+    user_id: UserId = DEFAULT_USER_ID
+    #: The SDK's schema (L6) and proactive intention (L7) lanes.
+    schema_limit: int | None = Field(default=None, ge=0, le=20)
+    intention_limit: int | None = Field(default=None, ge=0, le=20)
+
+    def search_overrides(self) -> Mapping[str, object]:
+        values = dict(super().search_overrides())
+        if self.schema_limit is not None:
+            values["schema_limit"] = self.schema_limit
+        if self.intention_limit is not None:
+            values["intention_limit"] = self.intention_limit
+        return values
 
 
 class RecallHit(BaseModel):

memhub/apps/core/src/one_memory/sdk_memory_qa.py

@@ -6,7 +6,7 @@ import asyncio
 import json
 from collections.abc import AsyncIterator, Awaitable, Callable
 from contextlib import aclosing
-from typing import Any, Literal, Protocol
+from typing import Any, Literal, Protocol, cast
 
 import httpx
 from pydantic import BaseModel, ConfigDict
@@ -14,7 +14,7 @@ from pydantic import BaseModel, ConfigDict
 from one_memory.http_client import create_async_client
 from one_memory.hy_runtime_config import HY_MEMORY_READER
 from one_memory.hy_sdk_runtime import sdk_relevance_floor
-from one_memory.hy_sdk_types import HySdkMemoryRecord, parse_sdk_search
+from one_memory.hy_sdk_types import HySdkMemoryRecord, parse_sdk_lane_search, parse_sdk_search
 from one_memory.memory_qa import (
     MemoryAnswerError,
     _DeferredSecretStore,
@@ -31,7 +31,9 @@ from one_memory.model_profiles import (
     ProviderKind,
     attach_disable_thinking,
 )
+from one_memory.sdk_search_options import SdkSearchFilters, SearchLane
 from one_memory.secrets import SecretStore, SecretStoreUnavailableError
+from one_memory.user_scope import owns
 
 _NO_GROUNDING = "现有记忆不足以回答这个问题。"
 _GROUNDING_LIMIT = 6
@@ -95,29 +97,51 @@ class SdkMemoryAnswerer:
         question: str,
         *,
         user_id: str = "local",
+        filters: SdkSearchFilters | None = None,
     ) -> tuple[SdkAskMemory, ...]:
+        """The user's active memories nearest the question.
+
+        `filters` are a caller's pass-through search options; absent, the
+        search is the one ask always made.
+        """
         client = await self.client_handle.client()
+        overrides = filters.search_overrides() if filters is not None else {}
         raw = await asyncio.to_thread(
             client.search,
             question,
-            user_ids=[user_id],
-            agent_ids=[],
-            limit=12,
-            min_score=sdk_relevance_floor(),
-            schema_min_score=sdk_relevance_floor(),
-            schema_limit=12,
-            intention_limit=12,
-            scene="normal",
-            reader=HY_MEMORY_READER,
+            **{
+                "user_ids": [user_id],
+                "agent_ids": [],
+                "limit": 12,
+                "min_score": sdk_relevance_floor(),
+                "schema_min_score": sdk_relevance_floor(),
+                "schema_limit": 12,
+                "intention_limit": 12,
+                "scene": "normal",
+                "reader": HY_MEMORY_READER,
+                **overrides,
+            },
         )
-        hits = parse_sdk_search(raw, error_code="hy_memory_search_shape_unknown")
+        if filters is not None and overrides:
+            hits = tuple(
+                hit
+                for hit in parse_sdk_lane_search(raw, error_code="hy_memory_search_shape_unknown")
+                if filters.passes_floor(cast(SearchLane, hit.lane), hit.score)
+            )
+        else:
+            hits = parse_sdk_search(raw, error_code="hy_memory_search_shape_unknown")
         found: list[SdkAskMemory] = []
         seen: set[str] = set()
         for hit in hits:
             if hit.memory_id in seen:
                 continue
             record = await _get_memory(client, hit.memory_id)
-            if record is None or record.status != "active" or not record.is_latest:
+            if (
+                record is None
+                or record.status != "active"
+                or not record.is_latest
+                or not owns(record.user_id, user_id)
+            ):
                 continue
             seen.add(record.memory_id)
             found.append(_wire_memory(record))
@@ -128,8 +152,9 @@ class SdkMemoryAnswerer:
         question: str,
         *,
         user_id: str = "local",
+        filters: SdkSearchFilters | None = None,
     ) -> SdkMemoryAnswer:
-        candidates = await self.retrieve(question, user_id=user_id)
+        candidates = await self.retrieve(question, user_id=user_id, filters=filters)
         if not candidates:
             return SdkMemoryAnswer(answer=_NO_GROUNDING, memories=(), mode="empty")
         profile = await self._active_profile()
@@ -160,8 +185,9 @@ class SdkMemoryAnswerer:
         question: str,
         *,
         user_id: str = "local",
+        filters: SdkSearchFilters | None = None,
     ) -> AsyncIterator[dict[str, object]]:
-        candidates = await self.retrieve(question, user_id=user_id)
+        candidates = await self.retrieve(question, user_id=user_id, filters=filters)
         await self._require_current(candidates)
         yield {"type": "recall", "memories": candidates}
         if not candidates:

memhub/apps/core/src/one_memory/sdk_memory_runtime.py

@@ -5,8 +5,8 @@ from __future__ import annotations
 import asyncio
 import logging
 import re
-from collections.abc import AsyncIterator
-from contextlib import AsyncExitStack, asynccontextmanager, suppress
+from collections.abc import AsyncIterator, Callable
+from contextlib import AsyncExitStack, asynccontextmanager
 from datetime import UTC, datetime, timedelta
 from pathlib import Path
 from typing import Any
@@ -33,16 +33,6 @@ from one_memory.semantic_deadline import SemanticMemoryHit
 _STABLE_CODE = re.compile(r"[a-z][a-z0-9_]{0,95}")
 
 logger = logging.getLogger(__name__)
-RAW_RETENTION_DAYS = 3
-RAW_RETENTION_CHECK_SECONDS = 6 * 60 * 60
-_RAW_STATUSES = (
-    "active",
-    "superseded",
-    "shadow",
-    "negated",
-    "conflicted",
-    "archived",
-)
 
 
 class SdkMemoryRuntime:
@@ -56,74 +46,17 @@ class SdkMemoryRuntime:
         write_queue: DurableWriteQueue,
         extraction_settings: ExtractionSettingsStore,
         data_dir: Path,
+        memory_mode: Callable[[], str] = lambda: "pro",
     ) -> None:
         self.coordinator = coordinator
+        #: The SDK mode, for the graph store's path (`hy_graph_admin`).
+        self.memory_mode = memory_mode
         self.read_handle = read_handle
         self.write_queue = write_queue
         self.extraction_settings = extraction_settings
         self.data_dir = data_dir
-        self._raw_retention_task: asyncio.Task[None] | None = None
-
-    def start_raw_retention(self) -> None:
-        if self._raw_retention_task is None or self._raw_retention_task.done():
-            self._raw_retention_task = asyncio.create_task(
-                self._raw_retention_loop(),
-                name="one-memory-sdk-raw-retention",
-            )
-
-    async def prune_expired_raw(self, *, now: datetime | None = None) -> int:
-        cutoff = (now or datetime.now(UTC)) - timedelta(days=RAW_RETENTION_DAYS)
-        cutoff_timestamp = int(cutoff.timestamp())
-        client = await self.read_handle.client()
-        deleted = 0
-        while True:
-            page = await asyncio.to_thread(
-                client.list_memories,
-                user_id="local",
-                layers=["l1_raw"],
-                statuses=list(_RAW_STATUSES),
-                order="asc",
-                limit=500,
-                offset=0,
-            )
-            bucket = page.get("vdb") if isinstance(page, dict) else None
-            rows = bucket.get("memories") if isinstance(bucket, dict) else None
-            if not isinstance(rows, list) or not rows:
-                return deleted
-            expired = [
-                row
-                for row in rows
-                if isinstance(row, dict)
-                and isinstance(row.get("memory_id"), str)
-                and int(row.get("memory_at") or row.get("gmt_created") or 0)
-                < cutoff_timestamp
-            ]
-            if not expired:
-                return deleted
-            removed_page = 0
-            for row in expired:
-                result = await asyncio.to_thread(client.delete, row["memory_id"])
-                if isinstance(result, dict) and result.get("success") is True:
-                    count = int(result.get("deleted_count") or 0)
-                    deleted += count
-                    removed_page += count
-            if removed_page == 0:
-                return deleted
-
-    async def _raw_retention_loop(self) -> None:
-        while True:
-            await asyncio.sleep(RAW_RETENTION_CHECK_SECONDS)
-            try:
-                removed = await self.prune_expired_raw()
-                if removed:
-                    logger.info("sdk_raw_retention_deleted count=%d", removed)
-            except asyncio.CancelledError:
-                raise
-            except Exception as error:
-                logger.warning(
-                    "sdk_raw_retention_failed error_type=%s",
-                    type(error).__name__,
-                )
+        # MemHub keeps every raw the SDK stores, in every mode: there is no
+        # retention window (owner decision; the SDK itself never deletes raw).
 
     async def semantic_recall(
         self,
@@ -154,11 +87,10 @@ class SdkMemoryRuntime:
         limit: int = 12,
     ) -> tuple[tuple[str, float], ...]:
         return tuple(
-            (hit.memory_id, hit.score)
-            for hit in await self.semantic_recall(query, limit=limit)
+            (hit.memory_id, hit.score) for hit in await self.semantic_recall(query, limit=limit)
         )
 
-    async def graph_snapshot(self, *, limit: int = 300) -> HyGraphSnapshot:
+    async def graph_snapshot(self, *, limit: int = 300, user_id: str = "local") -> HyGraphSnapshot:
         try:
             await self.coordinator.active_profile_credentials()
         except RuntimeError as error:
@@ -173,7 +105,7 @@ class SdkMemoryRuntime:
             return await asyncio.to_thread(
                 read_graph_snapshot,
                 client,
-                "local",
+                user_id,
                 "*",
                 limit,
             )
@@ -224,6 +156,7 @@ class SdkMemoryRuntime:
             profile,
             api_key=secret,
             data_dir=self.data_dir,
+            mode=self.memory_mode(),
         )
 
     async def decide_graph_width(
@@ -240,6 +173,7 @@ class SdkMemoryRuntime:
             api_key=secret,
             data_dir=self.data_dir,
             decision=decision,
+            mode=self.memory_mode(),
         )
 
     @asynccontextmanager
@@ -378,12 +312,7 @@ class SdkMemoryRuntime:
         return None
 
     async def shutdown(self) -> None:
-        if self._raw_retention_task is None:
-            return
-        self._raw_retention_task.cancel()
-        with suppress(asyncio.CancelledError):
-            await self._raw_retention_task
-        self._raw_retention_task = None
+        return None
 
 
 def _agent_ids(raw: object) -> tuple[str, ...]:

memhub/apps/core/src/one_memory/sdk_migration_runner.py

@@ -23,6 +23,7 @@ from pathlib import Path
 from typing import Any
 
 from one_memory.database_access import main_database
+from one_memory.hy_sdk_runtime import builtin_retrieval_snapshot
 from one_memory.imports.models import EvidenceRecord
 from one_memory.imports.storage import EvidenceStore
 from one_memory.memory_server.contracts import (
@@ -51,6 +52,7 @@ from one_memory.sdk_migration_bundle import inspect_sdk_migration_bundle
 from one_memory.sdk_migration_verify import (
     MemoryReadbackFailed,
     ReadbackReport,
+    expected_memory_ids,
     legacy_memory_ids_from,
     legacy_memory_ids_from_bundle,
     read_active_profile,
@@ -61,6 +63,8 @@ from one_memory.sdk_vector_reattach import (
     ReattachedCollection,
     legacy_vector_collections,
     reattach_legacy_vectors,
+    stored_memory_ids,
+    vectors_belong_to_another_space,
 )
 from one_memory.target_schema import drop_legacy_tables, legacy_tables_present
 
@@ -259,6 +263,7 @@ def prepare_legacy_write_queue_for_cutover(database_path: Path) -> tuple[int, in
         internal_session_id = internal_session_id_for(
             source.conversation_id,
             source.source_id,
+            source.user_id,
         )
         queue.submit(
             WriteRequest(
@@ -456,13 +461,17 @@ def resume_legacy_cleanup(*, data_dir: Path, database_path: Path) -> tuple[str,
             bundle_id=marker.backup_bundle_id,
         )
     from one_memory.hy_runtime_config import HY_MEMORY_NAMESPACE
+    from one_memory.memory_visibility import drop_stale_triggers
 
     recover_abandoned_staging(
         data_dir=data_dir,
         database_path=database_path,
         namespace=HY_MEMORY_NAMESPACE,
     )
-    return drop_legacy_tables(database_path)
+    dropped = drop_legacy_tables(database_path)
+    # The visibility cache's grant triggers read a table just dropped (#233).
+    drop_stale_triggers(database_path)
+    return dropped
 
 
 MIGRATION_PREFIX = "sdk-authority-v2"
@@ -699,13 +708,14 @@ def run_startup_migration(
         step["authority"] = authority.value
     log.migration_id = authority_migration
     with log.step("preflight.profile_read") as step:
-        profile = read_active_profile(database_path)
-        step["profile_present"] = profile is not None
+        active = read_active_profile(database_path)
+        step["profile_present"] = active is not None
+    # No active profile means no generation model, not no embedding: the
+    # vectors are the built-in model's, which is also what the Runtime reads.
+    profile = active if active is not None else builtin_retrieval_snapshot()
     if authority is MemoryAuthority.SDK:
         with log.step("preflight.leftovers_scan") as step:
-            vectors_waiting = profile is not None and bool(
-                legacy_vector_collections(data_dir, profile)
-            )
+            vectors_waiting = bool(legacy_vector_collections(data_dir, profile))
             leftovers = bool(legacy_tables_present(database_path)) or abandoned_staging_present(
                 data_dir=data_dir, namespace=namespace
             )
@@ -716,7 +726,6 @@ def run_startup_migration(
         reattached: tuple[ReattachedCollection, ...] = ()
         readback: ReadbackReport | None = None
         if vectors_waiting:
-            assert profile is not None
             progress("vectors")
             with log.step("vectors.reattach") as step:
                 reattached = reattach_legacy_vectors(data_dir, profile)
@@ -768,12 +777,12 @@ def run_startup_migration(
         step["chat"] = len(legacy.chat)
         step["productivity"] = len(legacy.productivity)
         step["users"] = len(legacy.users)
-    if profile is None and not legacy.empty:
-        # Without the embedding profile the Runtime's collections cannot be
-        # named, so neither reattached nor read back. Legacy stays intact.
-        log.check("preflight.model_profile_check", "sdk_migration_model_profile_required")
-        raise DataPlaneMigrationError("sdk_migration_model_profile_required")
-    log.check("preflight.model_profile_check", None)
+    if active is None and not legacy.empty and vectors_belong_to_another_space(data_dir, profile):
+        # The memories live in the space of a profile that is no longer active,
+        # so the built-in model cannot read them. Legacy stays intact.
+        log.check("preflight.vector_space_check", "sdk_migration_vector_profile_missing")
+        raise DataPlaneMigrationError("sdk_migration_vector_profile_missing")
+    log.check("preflight.vector_space_check", None)
     with capture_fence(data_dir):
         # Builds through 19 left a staged SDK copy beside the original store.
         # Settling it before the bundle means the bundle holds the original
@@ -829,17 +838,23 @@ def run_startup_migration(
                 step.update(preparation.as_document())
             reattached = ()
             readback = None
-            if profile is not None:
-                progress("vectors")
-                with log.step("vectors.reattach") as step:
-                    reattached = reattach_legacy_vectors(data_dir, profile)
-                    step["collections"] = len(reattached)
-            if not legacy.empty:
-                assert profile is not None
+            # Before reattach touches the store, and after the staging
+            # recovery above has settled which store is the original.
+            with log.step("verify.stored_ids") as step:
+                expected, missing = expected_memory_ids(legacy, stored_memory_ids(data_dir))
+                step["chat_expected"] = len(expected.chat)
+                step["chat_missing_before_migration"] = missing
+            progress("vectors")
+            with log.step("vectors.reattach") as step:
+                reattached = reattach_legacy_vectors(data_dir, profile)
+                step["collections"] = len(reattached)
+            if not expected.empty:
                 progress("verify")
                 with log.step("verify.readback") as step:
                     with client_for(profile) as client:
-                        readback = verify_readback(client, legacy)
+                        readback = verify_readback(
+                            client, expected, chat_missing_before_migration=missing
+                        )
                     step.update(readback.as_document())
                     if not readback.passed:
                         store.set_phase(migration_id, phase="readback_failed", state="blocked")

memhub/apps/core/src/one_memory/sdk_migration_verify.py

@@ -14,12 +14,12 @@ import json
 import sqlite3
 from collections.abc import Iterator
 from contextlib import closing, contextmanager
-from dataclasses import dataclass
+from dataclasses import dataclass, replace
 from pathlib import Path
 from typing import Any
 
 from one_memory.database_access import main_database
-from one_memory.model_profiles import ModelProfileSnapshot
+from one_memory.model_profiles import ModelProfileSnapshot, pinned_embedding_base_url
 from one_memory.sdk_data_plane_migration import DataPlaneMigrationError
 
 _PRODUCTIVITY_KINDS = frozenset({"coding", "working"})
@@ -47,6 +47,8 @@ class ReadbackReport:
     chat_readable: int
     productivity_expected: int
     productivity_readable: int
+    #: Confirmed in the mirror, gone from the SDK before this migration began.
+    chat_missing_before_migration: int = 0
 
     @property
     def passed(self) -> bool:
@@ -63,6 +65,7 @@ class ReadbackReport:
             "chat_readable": self.chat_readable,
             "productivity_expected": self.productivity_expected,
             "productivity_readable": self.productivity_readable,
+            "chat_missing_before_migration": self.chat_missing_before_migration,
             "passed": self.passed,
         }
 
@@ -115,6 +118,27 @@ def read_legacy_memory_ids(connection: sqlite3.Connection) -> LegacyMemoryIds:
     )
 
 
+def expected_memory_ids(
+    legacy: LegacyMemoryIds, stored: frozenset[str]
+) -> tuple[LegacyMemoryIds, int]:
+    """What the readback must find, and how many confirmed chats it need not.
+
+    A 1.0.4 library can hold confirmed rows whose SDK node the SDK had already
+    merged or deleted: no migration could read those back (658 confirmed, 657
+    in the store, and the whole upgrade failed). So chat is held to what the
+    store held before this run, and anything in there must still read back.
+
+    If none of the confirmed chats is in the store, that is not history: the
+    store is empty or is the wrong one. Nothing is excused then, so the
+    readback still fails on every id instead of checking nothing just before
+    the mirror is dropped.
+    """
+    chat = tuple(memory_id for memory_id in legacy.chat if memory_id in stored)
+    if not chat:
+        return legacy, 0
+    return replace(legacy, chat=chat), len(legacy.chat) - len(chat)
+
+
 def legacy_memory_ids_from(database_path: Path) -> LegacyMemoryIds:
     with main_database(database_path).read() as connection:
         return read_legacy_memory_ids(connection)
@@ -160,7 +184,14 @@ def read_active_profile(database_path: Path) -> ModelProfileSnapshot | None:
         model=row["model"],
         embedding_model=row["embedding_model"],
         embedding_provider_kind=row["embedding_provider_kind"],
-        embedding_base_url=row["embedding_base_url"],
+        # Read-only, beside a Core that may not have pinned it yet.
+        embedding_base_url=pinned_embedding_base_url(
+            provider_kind=row["provider_kind"],
+            base_url=row["base_url"],
+            embedding_model=row["embedding_model"],
+            embedding_provider_kind=row["embedding_provider_kind"],
+            embedding_base_url=row["embedding_base_url"],
+        ),
         embedding_dimensions=row["embedding_dimensions"],
         timeout_seconds=row["timeout_seconds"],
         max_concurrency=row["max_concurrency"],
@@ -179,14 +210,18 @@ def runtime_sdk_client(
     from one_memory.model_profiles import ModelProfileStore
     from one_memory.secrets import KeyringSecretStore, SecretStoreUnavailableError
 
-    try:
-        secret = ModelProfileStore(database_path).resolve_snapshot_secret(
-            profile.profile_id, profile.revision, KeyringSecretStore()
-        )
-    except SecretStoreUnavailableError:
-        raise DataPlaneMigrationError("sdk_migration_secure_storage_unavailable") from None
-    if profile.has_api_key and secret is None:
-        raise DataPlaneMigrationError("sdk_migration_model_secret_unavailable")
+    secret: str | None = None
+    # Without a key there is nothing to resolve, and the built-in retrieval
+    # snapshot has no stored version to look one up by.
+    if profile.has_api_key:
+        try:
+            secret = ModelProfileStore(database_path).resolve_snapshot_secret(
+                profile.profile_id, profile.revision, KeyringSecretStore()
+            )
+        except SecretStoreUnavailableError:
+            raise DataPlaneMigrationError("sdk_migration_secure_storage_unavailable") from None
+        if secret is None:
+            raise DataPlaneMigrationError("sdk_migration_model_secret_unavailable")
     lease = acquire_client(profile, api_key=secret, data_dir=data_dir)
     try:
         yield lease.client
@@ -194,7 +229,9 @@ def runtime_sdk_client(
         lease.close()
 
 
-def verify_readback(client: Any, legacy: LegacyMemoryIds) -> ReadbackReport:
+def verify_readback(
+    client: Any, legacy: LegacyMemoryIds, *, chat_missing_before_migration: int = 0
+) -> ReadbackReport:
     listed: set[str] = set()
     for user_id in legacy.users or ("local",):
         offset = 0
@@ -228,5 +265,6 @@ def verify_readback(client: Any, legacy: LegacyMemoryIds) -> ReadbackReport:
         chat_readable=sum(1 for memory_id in legacy.chat if readable(memory_id)),
         productivity_expected=len(legacy.productivity),
         productivity_readable=sum(1 for memory_id in legacy.productivity if readable(memory_id)),
+        chat_missing_before_migration=chat_missing_before_migration,
     )
 

memhub/apps/core/src/one_memory/sdk_read_handle.py

@@ -0,0 +1,125 @@
+"""The read client, kept in the embedding space the active profile writes to.
+
+Every read used to open the built-in 512-dim collection while writes followed
+the active profile, so a custom embedding read an empty space (#242). This
+handle opens the active profile's `retrieval_view` instead, and the built-in
+space only when no profile is active.
+
+A profile change builds a new client and retires the old one; a retired client
+is closed once no task still uses it, as `SdkWriteCoordinator` does for writes
+(#218). A use is the task that asked for the client: callers keep their
+`client = await handle.client()` shape, and the use ends with that task (one
+HTTP request, one retention pass).
+
+The memory mode (lite / pro / ultra) is fixed per SDK client, so a mode change
+rebuilds and retires the same way: handles are keyed by (profile, mode). The
+mode never names a collection; it is a separate client parameter.
+"""
+
+from __future__ import annotations
+
+import asyncio
+from collections.abc import Callable
+from pathlib import Path
+from typing import Any
+
+from one_memory.hy_graph_admin import effective_memory_mode
+from one_memory.hy_sdk_runtime import (
+    BUILTIN_RETRIEVAL_PROFILE_ID,
+    SdkClientHandle,
+    builtin_retrieval_snapshot,
+    retrieval_view,
+)
+from one_memory.hy_sdk_types import HySdkRuntimeError
+from one_memory.model_profiles import ModelProfileStore, ProviderKind
+from one_memory.secrets import SecretStore, SecretStoreUnavailableError
+
+_BUILTIN = (BUILTIN_RETRIEVAL_PROFILE_ID, 0)
+
+
+class ActiveProfileReadHandle(SdkClientHandle):
+    """An `SdkClientHandle` whose profile follows the active one."""
+
+    def __init__(
+        self,
+        *,
+        profiles: ModelProfileStore,
+        secrets: SecretStore,
+        data_dir: Path,
+        memory_mode: Callable[[], str] = effective_memory_mode,
+    ) -> None:
+        super().__init__(builtin_retrieval_snapshot(), api_key=None, data_dir=data_dir)
+        self.profiles = profiles
+        self.secrets = secrets
+        self.data_dir = data_dir
+        self.memory_mode = memory_mode
+        self._handles: dict[tuple[str, int, str], SdkClientHandle] = {}
+        #: Tasks in flight per handle; a retired handle closes at zero.
+        self._uses: dict[SdkClientHandle, int] = {}
+        self._retired: set[SdkClientHandle] = set()
+        self._closing: set[asyncio.Task[None]] = set()
+        self._handle_lock = asyncio.Lock()
+
+    async def client(self) -> Any:
+        profile = await asyncio.to_thread(self.profiles.active_identity) or _BUILTIN
+        identity = (*profile, self.memory_mode())
+        async with self._handle_lock:
+            handle = self._handles.get(identity)
+            if handle is None:
+                # Only on a change of profile: resolving a key can reach the
+                # system keychain, so it is not done on every read.
+                handle = await asyncio.to_thread(self._build, identity)
+                self._handles[identity] = handle
+            self._retired.update(value for key, value in self._handles.items() if key != identity)
+            self._handles = {identity: handle}
+            unused = {value for value in self._retired if not self._uses.get(value)}
+            self._retired -= unused
+            self._hold(handle)
+        await asyncio.gather(*(value.close() for value in unused))
+        return await handle.client()
+
+    async def close(self) -> None:
+        async with self._handle_lock:
+            handles = (*self._handles.values(), *self._retired)
+            self._handles, self._retired, self._uses = {}, set(), {}
+        await asyncio.gather(*(handle.close() for handle in handles), *self._closing)
+
+    def _hold(self, handle: SdkClientHandle) -> None:
+        """Count a use of `handle` until the asking task finishes."""
+        task = asyncio.current_task()
+        if task is None:  # pragma: no cover - always called from a coroutine
+            return
+        self._uses[handle] = self._uses.get(handle, 0) + 1
+        task.add_done_callback(lambda _task: self._release(handle))
+
+    def _release(self, handle: SdkClientHandle) -> None:
+        remaining = self._uses.get(handle, 0) - 1
+        if remaining > 0:
+            self._uses[handle] = remaining
+            return
+        self._uses.pop(handle, None)
+        if handle in self._retired:
+            self._retired.discard(handle)
+            closing = asyncio.get_running_loop().create_task(handle.close())
+            self._closing.add(closing)
+            closing.add_done_callback(self._closing.discard)
+
+    def _build(self, key: tuple[str, int, str]) -> SdkClientHandle:
+        profile_id, revision, mode = key
+        identity = (profile_id, revision)
+        if identity == _BUILTIN:
+            return SdkClientHandle(
+                builtin_retrieval_snapshot(), api_key=None, data_dir=self.data_dir, mode=mode
+            )
+        profile = self.profiles.snapshot(*identity)
+        view = retrieval_view(profile)
+        secret: str | None = None
+        # Only a remote OpenAI-compatible embedding sends the shared key.
+        if view.embedding_provider_kind is ProviderKind.OPENAI_COMPATIBLE:
+            try:
+                secret = self.profiles.resolve_snapshot_secret(*identity, self.secrets)
+            except SecretStoreUnavailableError:
+                raise HySdkRuntimeError("secure_storage_unavailable") from None
+            if secret is None:
+                raise HySdkRuntimeError("model_secret_unavailable")
+        return SdkClientHandle(view, api_key=secret, data_dir=self.data_dir, mode=mode)

memhub/apps/core/src/one_memory/sdk_search_options.py

@@ -0,0 +1,179 @@
+"""SDK search parameters a host may pass through to `client.search`.
+
+`POST /api/v2/context/search`, `POST /api/v1/recall` and the v2 ask routes
+used to fix every SDK search knob, so a host that wanted a stricter relevance
+floor filtered the answer itself. These fields are the public `search`
+parameters of hy-memory 2.0.9 that are safe to hand a caller. Every field is
+optional: absent, the route keeps exactly the call it made before.
+
+Never exposed: `user_ids` (a route searches its one scoped `user_id`),
+`embed_override` and `query_embedding` (they bypass the configured embedder),
+and `request_id` (Core's own correlation id).
+
+A caller's floor is also applied to what Core returns, not only handed to the
+SDK: a hit below the floor that names its lane is dropped (`passes_floor`).
+With `min_score` alone that floor covers every lane, so `min_score=0.9`
+answers only results scoring at least 0.9.
+"""
+
+from __future__ import annotations
+
+from collections.abc import Mapping
+from typing import Annotated, Literal
+
+from pydantic import BaseModel, ConfigDict, Field, StrictBool, model_validator
+
+#: The SDK's `MemoryLayer` values (hy-memory 2.0.9).
+MemoryLayerName = Literal[
+    "l0_basic_info",
+    "l1_raw",
+    "l2_fact",
+    "l3_summary",
+    "l4_identity",
+    "l5_knowledge",
+    "l6_schema",
+    "l7_intention",
+]
+MEMORY_LAYERS: tuple[str, ...] = (
+    "l0_basic_info",
+    "l1_raw",
+    "l2_fact",
+    "l3_summary",
+    "l4_identity",
+    "l5_knowledge",
+    "l6_schema",
+    "l7_intention",
+)
+#: The SDK's `MemoryStatus` values.
+MEMORY_STATUSES: tuple[str, ...] = (
+    "active",
+    "superseded",
+    "negated",
+    "conflicted",
+    "shadow",
+    "archived",
+)
+SearchScene = Literal["chat", "working", "coding", "all"]
+#: Which recall lane produced a hit, for the floor that applies to it.
+SearchLane = Literal["normal", "schema", "proactive", "working", "coding", "raw"]
+
+Score = Annotated[float, Field(ge=0.0, le=1.0)]
+SourceName = Annotated[str, Field(min_length=1, max_length=128)]
+ScopeId = Annotated[str, Field(min_length=1, max_length=512)]
+
+
+class SdkSearchFilters(BaseModel):
+    """Filters and floors every search route accepts (recall, context, ask)."""
+
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    #: Chat lane floor; with nothing more specific, every lane's.
+    min_score: Score | None = None
+    #: Schema (L6) lane floor; defaults to `min_score` when only that is set.
+    schema_min_score: Score | None = None
+    #: False skips the SDK's reranking stages for this call.
+    rerank: StrictBool | None = None
+    include_sources: tuple[SourceName, ...] | None = Field(default=None, max_length=64)
+    exclude_sources: tuple[SourceName, ...] | None = Field(default=None, max_length=64)
+    #: Unix seconds: only memories created at or after it.
+    created_after: float | None = Field(default=None, ge=0)
+    session_ids: tuple[ScopeId, ...] | None = Field(default=None, max_length=64)
+    agent_ids: tuple[ScopeId, ...] | None = Field(default=None, max_length=64)
+    #: Only these layers (pro/ultra; lite always searches raw).
+    layers: tuple[MemoryLayerName, ...] | None = Field(default=None, min_length=1, max_length=8)
+
+    def filter_kwargs(self) -> dict[str, object]:
+        """The SDK keyword arguments this request sets, and only those."""
+        values: dict[str, object] = {}
+        for name in (
+            "min_score",
+            "schema_min_score",
+            "rerank",
+            "created_after",
+        ):
+            value = getattr(self, name)
+            if value is not None:
+                values[name] = value
+        for name in ("include_sources", "exclude_sources", "session_ids", "agent_ids", "layers"):
+            value = getattr(self, name)
+            if value is not None:
+                values[name] = list(dict.fromkeys(value))
+        if self.min_score is not None and self.schema_min_score is None:
+            values["schema_min_score"] = self.min_score
+        return values
+
+    def lane_floor(self, lane: SearchLane) -> float | None:
+        """The caller's floor for one lane, or None when the caller set none."""
+        if lane == "schema":
+            return self.schema_min_score if self.schema_min_score is not None else self.min_score
+        return self.min_score
+
+    def passes_floor(self, lane: SearchLane, score: float) -> bool:
+        floor = self.lane_floor(lane)
+        return floor is None or score >= floor
+
+    def search_overrides(self) -> Mapping[str, object]:
+        return self.filter_kwargs()
+
+
+class SdkSearchOptions(SdkSearchFilters):
+    """Every SDK search knob a recall or context caller may set."""
+
+    #: working/coding fact-layer floor (SDK default 0.30).
+    working_fact_min_score: Score | None = None
+    #: working/coding experience-layer floor (SDK default fact floor + 0.10).
+    working_experience_min_score: Score | None = None
+    reader: Literal["legacy", "hybrid"] | None = None
+    #: chat (the default), working, coding, or all three fused.
+    scene: SearchScene | None = None
+    #: agentic spends one model call to classify and rewrite the query.
+    search_mode: Literal["default", "agentic"] | None = None
+    #: working/coding lanes only; `branch` needs `workspace_id`.
+    workspace_id: ScopeId | None = None
+    branch: ScopeId | None = None
+
+    @model_validator(mode="after")
+    def branch_needs_workspace(self) -> SdkSearchOptions:
+        if self.branch is not None and self.workspace_id is None:
+            raise ValueError("branch requires workspace_id")
+        return self
+
+    def search_overrides(self) -> Mapping[str, object]:
+        values = self.filter_kwargs()
+        for name in (
+            "working_fact_min_score",
+            "working_experience_min_score",
+            "reader",
+            "scene",
+            "search_mode",
+            "workspace_id",
+            "branch",
+        ):
+            value = getattr(self, name)
+            if value is not None:
+                values[name] = value
+        if self.min_score is not None and self.working_fact_min_score is None:
+            values["working_fact_min_score"] = self.min_score
+        if self.min_score is not None and self.working_experience_min_score is None:
+            values["working_experience_min_score"] = self.min_score
+        return values
+
+    def lane_floor(self, lane: SearchLane) -> float | None:
+        if lane in {"working", "coding"}:
+            explicit = [
+                value
+                for value in (self.working_fact_min_score, self.working_experience_min_score)
+                if value is not None
+            ]
+            # Facts and experiences share one score field on a fused hit, so
+            # the looser of the two floors is the one Core can enforce.
+            return min(explicit) if explicit else self.min_score
+        return super().lane_floor(lane)
+
+
+def split_csv(values: list[str] | None) -> list[str]:
+    """Query values given repeated (`a&a`) or comma separated (`a,b`)."""
+    found: list[str] = []
+    for value in values or ():
+        found.extend(part.strip() for part in value.split(",") if part.strip())
+    return list(dict.fromkeys(found))

memhub/apps/core/src/one_memory/sdk_vector_reattach.py

@@ -9,6 +9,7 @@ one. Runtime code only reaches vectors through the SDK. This module runs from
 
 from __future__ import annotations
 
+import re
 import sqlite3
 from collections.abc import Iterator
 from contextlib import closing
@@ -21,6 +22,8 @@ from one_memory.model_profiles import ModelProfileSnapshot
 from one_memory.sdk_data_plane_migration import DataPlaneMigrationError
 
 _BATCH = 500
+#: Every name a MemHub build gave an embedding space's base collection.
+_SPACE_NAME = re.compile(r"^(?:memhub_[0-9a-f]{16}|one_memory_[0-9a-f]{12})(?:_|$)")
 
 
 def open_offline_chroma(directory: Path) -> Any:
@@ -51,13 +54,20 @@ def _vector_directory(data_dir: Path) -> Path:
     return data_dir / "hy-memory" / "governed-v2" / "vector"
 
 
-def _stored_names(data_dir: Path) -> tuple[str, ...]:
+def _stored_names(data_dir: Path, *, filled: bool = False) -> tuple[str, ...]:
+    """Collection names; with `filled`, only those holding at least one row."""
     database = _vector_directory(data_dir) / "chroma.sqlite3"
     if not database.is_file():
         return ()
     with closing(sqlite3.connect(f"file:{database}?mode=ro", uri=True)) as connection:
         try:
-            rows = connection.execute("SELECT name FROM collections").fetchall()
+            rows = connection.execute(
+                "SELECT DISTINCT collections.name FROM collections "
+                "JOIN segments ON segments.collection = collections.id "
+                "JOIN embeddings ON embeddings.segment_id = segments.id"
+                if filled
+                else "SELECT name FROM collections"
+            ).fetchall()
         except sqlite3.Error:
             return ()
     return tuple(sorted(str(row[0]) for row in rows))
@@ -80,6 +90,59 @@ def legacy_vector_collections(
     return tuple(pairs)
 
 
+def vectors_belong_to_another_space(data_dir: Path, profile: ModelProfileSnapshot) -> bool:
+    """Vectors exist, but only in embedding spaces `profile` does not name.
+
+    Counted rows, not names: every Runtime opens its read collection with
+    `get_or_create`, so an empty collection under the expected base proves
+    nothing. Read-only, so a preflight can ask before anything changes.
+    """
+    current, legacy_bases = vector_collection_bases(profile)
+    known = (current, *legacy_bases)
+    filled = _stored_names(data_dir, filled=True)
+    return not any(
+        name == base or name.startswith(f"{base}_") for name in filled for base in known
+    ) and any(_SPACE_NAME.match(name) for name in filled)
+
+
+def stored_memory_ids(data_dir: Path) -> frozenset[str]:
+    """Every id in the vector store as found, across all embedding spaces.
+
+    All of them on purpose. A memory stranded under a space the Runtime will
+    not read must count as present, so the readback then fails on it instead
+    of excusing it. Both the point id and the SDK's `node_id` are kept: the
+    SDK derives a point id from a node id that is not a UUID.
+
+    Callers must have stopped every SDK writer. A store that cannot be read
+    raises: "unknown" must never be taken for "absent".
+    """
+    if not (_vector_directory(data_dir) / "chroma.sqlite3").is_file():
+        return frozenset()
+    found: set[str] = set()
+    try:
+        client = open_offline_chroma(_vector_directory(data_dir))
+        try:
+            for collection in client.list_collections():
+                offset = 0
+                while True:
+                    page = collection.get(include=["metadatas"], limit=_BATCH, offset=offset)
+                    ids = list(page["ids"])
+                    if not ids:
+                        break
+                    found.update(ids)
+                    found.update(
+                        str(metadata["node_id"])
+                        for metadata in page["metadatas"] or ()
+                        if metadata and metadata.get("node_id")
+                    )
+                    offset += len(ids)
+        finally:
+            client.close()
+    except Exception as error:
+        raise DataPlaneMigrationError("sdk_migration_vector_store_unreadable") from error
+    return frozenset(found)
+
+
 def _records(collection: Any) -> Iterator[dict[str, Any]]:
     offset = 0
     while True:

memhub/apps/core/src/one_memory/sdk_write_processor.py

@@ -9,7 +9,8 @@ from dataclasses import dataclass
 from datetime import datetime
 from typing import Any
 
-from one_memory.hy_sdk_runtime import SdkClientHandle
+from one_memory.hy_graph_admin import effective_memory_mode
+from one_memory.hy_sdk_runtime import SdkClientHandle, builtin_retrieval_snapshot
 from one_memory.imports.models import EvidenceContentKind, EvidenceOrigin, EvidenceRecord
 from one_memory.input_sanitization import sanitize_model_input
 from one_memory.model_profiles import ModelProfileSnapshot, ModelProfileStore
@@ -31,8 +32,11 @@ class SdkWriteReceipt:
 class SdkWriteProcessor:
     """Turn one durable job into exactly one public SDK write call."""
 
-    def __init__(self, client_handle: SdkClientHandle) -> None:
+    def __init__(self, client_handle: SdkClientHandle, *, lite: bool = False) -> None:
         self._client_handle = client_handle
+        #: lite makes no model call: `add_extracted` would still call one, so a
+        #: direct memory note is stored as one raw user turn through `add`.
+        self._lite = lite
 
     async def process_sdk_job(
         self,
@@ -64,7 +68,7 @@ class SdkWriteProcessor:
             # waits and retries instead of being left `unknown`.
             raise RuntimeError("hy_memory_client_unavailable") from error
 
-        if extracted:
+        if extracted and not self._lite:
             contents = [sanitize_model_input(record.verbatim) for record in records]
             await before_hy_invocation()
             result = await asyncio.to_thread(
@@ -81,12 +85,15 @@ class SdkWriteProcessor:
         else:
             messages = [
                 {
-                    "role": record.role,
+                    "role": "user" if extracted else record.role,
                     "content": sanitize_model_input(record.verbatim),
                 }
                 for record in records
-                if record.content_kind is EvidenceContentKind.RAW_CONVERSATION
-                and record.role in _MESSAGE_ROLES
+                if extracted
+                or (
+                    record.content_kind is EvidenceContentKind.RAW_CONVERSATION
+                    and record.role in _MESSAGE_ROLES
+                )
             ]
             if not messages:
                 raise RuntimeError("hy_memory_task_without_inputs")
@@ -138,6 +145,15 @@ class SdkWriteCoordinator:
     from under a write still running on it: the SDK cleared its stores and
     the SharedRuntime closed mid-`add`, the job failed with AttributeError
     after HY was invoked, and it was left `unknown`.
+
+    Handles are keyed by profile and memory mode, so a mode change retires
+    handles the same way. In lite no generation model is needed: without an
+    active profile the built-in local embedder writes raw turns.
+
+    `exclusive_for_evolve(user_id)` keeps evolve and that user's writes
+    apart: it waits for the user's write in flight, and while it runs that
+    user's jobs are refused before the SDK is called
+    (`memory_evolve_in_progress`), so the drain puts them back.
     """
 
     def __init__(
@@ -148,13 +164,22 @@ class SdkWriteCoordinator:
         data_dir: Any,
         transport_factory: Callable[[ModelProfileSnapshot], Callable[[], Any] | None],
         sedimentation_enabled: Callable[[], bool],
+        memory_mode: Callable[[], str] = effective_memory_mode,
     ) -> None:
         self.profiles = profiles
         self.secrets = secrets
         self.data_dir = data_dir
         self.transport_factory = transport_factory
         self.sedimentation_enabled = sedimentation_enabled
-        self._handles: dict[tuple[str, int], SdkClientHandle] = {}
+        self.memory_mode = memory_mode
+        self._handles: dict[tuple[str, int, str], SdkClientHandle] = {}
+        #: Users evolve (or the graph backfill) is running for, the writes in
+        #: flight per user, and one exclusive run per user at a time. Writes
+        #: of one user still run side by side; only evolve waits for them.
+        self._evolving: set[str] = set()
+        self._writing: dict[str, int] = {}
+        self._writes_idle = asyncio.Condition()
+        self._exclusive: dict[str, asyncio.Lock] = {}
         #: Uses in flight per handle; a retired handle closes at zero.
         self._uses: dict[SdkClientHandle, int] = {}
         self._retired: set[SdkClientHandle] = set()
@@ -169,16 +194,66 @@ class SdkWriteCoordinator:
     ) -> SdkWriteReceipt:
         if not self.sedimentation_enabled():
             raise RuntimeError("model_sedimentation_disabled")
-        async with self._active_handle() as handle:
-            return await SdkWriteProcessor(handle).process_sdk_job(
-                records,
-                request_id=request_id,
-                before_hy_invocation=before_hy_invocation,
-            )
+        user_id = records[0].user_id if records else "local"
+        if user_id in self._evolving:
+            # Before any SDK call: the drain returns the job to pending.
+            raise RuntimeError("memory_evolve_in_progress")
+        self._writing[user_id] = self._writing.get(user_id, 0) + 1
+        try:
+            async with self._active_handle() as handle:
+                return await SdkWriteProcessor(
+                    handle, lite=getattr(handle, "mode", None) == "lite"
+                ).process_sdk_job(
+                    records,
+                    request_id=request_id,
+                    before_hy_invocation=before_hy_invocation,
+                )
+        finally:
+            remaining = self._writing.get(user_id, 1) - 1
+            if remaining > 0:
+                self._writing[user_id] = remaining
+            else:
+                self._writing.pop(user_id, None)
+            async with self._writes_idle:
+                self._writes_idle.notify_all()
 
-    async def inquire_write_outcome(
-        self, request_id: str
-    ) -> SdkWriteReceipt | None:
+    async def retire_stale_modes(self) -> None:
+        """After a mode change: retire handles of other modes, closing the idle ones.
+
+        A handle still writing is closed when its last use ends (#218), so a
+        write that started in the old mode finishes in it.
+        """
+        mode = self.memory_mode()
+        async with self._lock:
+            stale = {key: handle for key, handle in self._handles.items() if key[2] != mode}
+            for key in stale:
+                del self._handles[key]
+            self._retired.update(stale.values())
+            unused = {value for value in self._retired if not self._uses.get(value)}
+            self._retired -= unused
+        await asyncio.gather(*(value.close() for value in unused))
+
+    def evolving(self, user_id: str) -> bool:
+        """Whether evolve holds this user, so their queued writes should wait."""
+        return user_id in self._evolving
+
+    @asynccontextmanager
+    async def exclusive_for_evolve(self, user_id: str) -> AsyncIterator[None]:
+        """Hold one user's writes off while evolve runs for them.
+
+        New writes of the user are refused from here on; the ones already in
+        flight are waited for before the body runs.
+        """
+        async with self._exclusive.setdefault(user_id, asyncio.Lock()):
+            self._evolving.add(user_id)
+            try:
+                async with self._writes_idle:
+                    await self._writes_idle.wait_for(lambda: not self._writing.get(user_id))
+                yield
+            finally:
+                self._evolving.discard(user_id)
+
+    async def inquire_write_outcome(self, request_id: str) -> SdkWriteReceipt | None:
         async with self._active_handle() as handle:
             return await SdkWriteProcessor(handle).inquire_write_outcome(request_id)
 
@@ -204,9 +279,11 @@ class SdkWriteCoordinator:
     @asynccontextmanager
     async def _active_handle(self) -> AsyncIterator[SdkClientHandle]:
         """Use the active profile's handle; retire the others, closing the unused."""
-        identity, profile, secret = await asyncio.to_thread(
-            self._read_active_credentials
+        mode = self.memory_mode()
+        profile_identity, profile, secret = await asyncio.to_thread(
+            self._read_active_credentials, mode
         )
+        identity = (*profile_identity, mode)
         async with self._lock:
             handle = self._handles.get(identity)
             if handle is None:
@@ -215,6 +292,7 @@ class SdkWriteCoordinator:
                     api_key=secret,
                     data_dir=self.data_dir,
                     http_transport_factory=self.transport_factory(profile),
+                    mode=mode,
                 )
                 self._handles[identity] = handle
             self._retired.update(value for key, value in self._handles.items() if key != identity)
@@ -239,9 +317,14 @@ class SdkWriteCoordinator:
 
     def _read_active_credentials(
         self,
+        mode: str = "pro",
     ) -> tuple[tuple[str, int], ModelProfileSnapshot, str | None]:
         identity = self.profiles.active_identity()
         if identity is None:
+            if mode == "lite":
+                # lite stores raw turns with one embedding and calls no model.
+                builtin = builtin_retrieval_snapshot()
+                return (builtin.profile_id, builtin.revision), builtin, None
             raise RuntimeError("model_profile_not_active")
         profile = self.profiles.snapshot(*identity)
         try:

memhub/apps/core/src/one_memory/semantic_deadline.py

@@ -27,6 +27,11 @@ class SemanticMemoryHit:
     member_ids: tuple[str, ...] = ()
     sdk_formatted: bool = False
     formatting_complete: bool = False
+    #: The recall lane that produced the hit (`sdk_search_options.SearchLane`)
+    #: and the namespace to read it back from. Defaults are a chat memory.
+    lane: str = "normal"
+    kind: str = "chat"
+    variant: str | None = None
 
     def as_dict(self) -> dict[str, object]:
         return {

memhub/apps/core/src/one_memory/strict_query.py

@@ -0,0 +1,39 @@
+"""Refuse query parameters a route does not declare.
+
+FastAPI ignores an unknown query parameter. For a memory read that is a
+silent widening: `memory-library?user_id=__nobody__` against a Core that did
+not know `user_id` answered with the whole library, and the caller believed
+it was isolated. Routes that carry this dependency answer such a request with
+the same 422 a malformed body gets.
+"""
+
+from fastapi import Request
+from fastapi.dependencies.models import Dependant
+from fastapi.exceptions import RequestValidationError
+from fastapi.routing import APIRoute
+
+
+def _declared_query_names(dependant: Dependant) -> set[str]:
+    names = {parameter.alias for parameter in dependant.query_params}
+    for dependency in dependant.dependencies:
+        names |= _declared_query_names(dependency)
+    return names
+
+
+async def forbid_unknown_query_params(request: Request) -> None:
+    route = request.scope.get("route")
+    if not isinstance(route, APIRoute):
+        return
+    unknown = sorted(set(request.query_params) - _declared_query_names(route.dependant))
+    if unknown:
+        raise RequestValidationError(
+            [
+                {
+                    "type": "extra_forbidden",
+                    "loc": ("query", name),
+                    "msg": "Extra inputs are not permitted",
+                    "input": None,
+                }
+                for name in unknown
+            ]
+        )

memhub/apps/core/src/one_memory/user_scope.py

@@ -0,0 +1,65 @@
+"""Who a memory belongs to: one Runtime, many `user_id`s.
+
+A user is decided by which source root is read (a Hermes profile is its own
+`state.db`), never by session or platform fields. The App never names a user
+and keeps reading and writing `local`; every other harness stays on `local`
+too. Only Hermes may connect a second root for another user.
+
+In memory, a source's import lane is keyed by `lane_key(user_id, source_id)`.
+For `local` that is the bare source id, so every existing key, health record
+and log line stays byte-identical; another user's lane is `source@user`.
+`@` cannot occur in a user id or a source id, so the key splits back exactly.
+"""
+
+import re
+from typing import Annotated
+
+from pydantic import Field
+
+DEFAULT_USER_ID = "local"
+
+#: `:` is excluded on purpose: the SDK joins ids with it when it keys history,
+#: so a user id holding one could read another user's history.
+USER_ID_PATTERN = r"^[A-Za-z0-9._-]{1,64}$"
+_USER_ID = re.compile(USER_ID_PATTERN)
+
+#: The only source whose roots may belong to a user other than `local`.
+MULTI_USER_SOURCES = frozenset({"hermes"})
+
+UserId = Annotated[str, Field(min_length=1, max_length=64, pattern=USER_ID_PATTERN)]
+
+_LANE_SEPARATOR = "@"
+
+
+def valid_user_id(value: str) -> bool:
+    return bool(_USER_ID.fullmatch(value))
+
+
+def lane_key(user_id: str, source_id: str) -> str:
+    if user_id == DEFAULT_USER_ID:
+        return source_id
+    return f"{source_id}{_LANE_SEPARATOR}{user_id}"
+
+
+def other_user_lane_prefix(source_id: str) -> str:
+    """What every other user's lane of `source_id` starts with."""
+    return f"{source_id}{_LANE_SEPARATOR}"
+
+
+def split_lane(key: str) -> tuple[str, str]:
+    """`(user_id, source_id)` of a lane key; a bare source id is `local`'s."""
+    source_id, separator, user_id = key.partition(_LANE_SEPARATOR)
+    if not separator or not valid_user_id(user_id):
+        return DEFAULT_USER_ID, key
+    return user_id, source_id
+
+
+def owns(record_user_id: str | None, user_id: str) -> bool:
+    """Whether a stored memory belongs to `user_id`.
+
+    The SDK's `get`/`update`/`delete`/`history`/`get_write_outcome` take no
+    user, so MemHub checks this before acting on an id. A memory written
+    before MemHub named users carries `local` or nothing; both are `local`'s.
+    """
+    stored = record_user_id or DEFAULT_USER_ID
+    return stored == user_id

memhub/apps/core/tests/TEST_INDEX.md

@@ -105,6 +105,7 @@
 | `test_graph_width_guard.py` | `unit` | Graph width mismatch must ask, never silently archive. |
 | `test_harness_contract_suite.py` | `contract` | A4-09: one contract, ten harnesses. |
 | `test_health.py` | `contract` | 覆盖 a backfill that fails says so rather than disappearing、a failed catch up leaves a traceback behind、configured model without content consent does not start backfill 等 15 个场景。 |
+| `test_hermes_gateway_injection.py` | `contract` | Hermes gateway text in `user` rows: notices are skipped, relayed messages unwrapped. |
 | `test_history_listing_rotation.py` | `contract` | 覆盖 history listing alternates started and untouched files、history listing drains whichever cohort is longer 场景。 |
 | `test_host_capabilities.py` | `unit` | A plugin does not become realtime on every host or every unknown version. |
 | `test_hourly_c1_peak_replay_contract.py` | `contract` | Guard the package peak-replay probe's error accounting. |
@@ -160,6 +161,7 @@
 | `test_memory_evidence_freshness.py` | `contract` | 新鲜度拆除后：读路径恒为 current；证据修订 / 畸形 proof / segment ref 不再藏记忆或 503。 |
 | `test_memory_freshness_integration_review.py` | `integration` | Independent synthetic read-path review; no lifespan, model or user database. |
 | `test_memory_library_all_scenes.py` | `integration` | The v2 memory library across chat, working and coding memories. |
+| `test_memory_mode_and_evolve.py` | `contract` | The memory mode (lite / pro / ultra), evolve and its nightly schedule. |
 | `test_memory_qa.py` | `contract` | 覆盖 search terms keep meaningful single chinese character、streamed answer emits recall then incremental grounded text、governance change during generation stops the old answer 等 14 个场景。 |
 | `test_memory_qa_api.py` | `integration` | 覆盖 ask stream endpoint emits recall deltas and final、ask stream endpoint requires the local token 场景。 |
 | `test_memory_qa_permissions.py` | `contract` | 覆盖 initial refusal precedes all secret access and network、invalid authorization result fails before secret access、authorization receives exact question and only outbound grounding 等 16 个场景。 |
@@ -196,6 +198,7 @@
 | `test_model_profiles.py` | `contract` | 覆盖 profile uses one model and keeps key out of database、api key updates reuse one stable secret reference、startup cleanup removes only secrets not referenced by current profiles 等 15 个场景。 |
 | `test_model_request_audit.py` | `contract` | 覆盖 failure is recorded without error message、missing usage stays unknown in detail and zero in summary、records one physical request with provider usage and no content 等 5 个场景。 |
 | `test_model_request_budget.py` | `contract` | 覆盖 persists failed request reservations without content、parallel leases cannot exceed request limit、unknown or excessive output and body are denied 等 6 个场景。 |
+| `test_multi_user_runtime.py` | `contract` | One Runtime, many users: `user_id` on the memory API, per-user Hermes profiles. |
 | `test_mysql_lifecycle.py` | `contract` | A1-07: the decisions Runtime makes around the bundled MySQL. |
 | `test_mysql_platform.py` | `contract` | A1-07: the machine-facing half of managing the bundled MySQL. |
 | `test_obs03_health_probe.py` | `contract` | OBS-03: '/health' answered with the web app instead of a health payload. |
@@ -246,6 +249,7 @@
 | `test_schema_gate.py` | `contract` | 覆盖 fresh database records the build schema、future schema refuses to open、legacy database without a ledger is left untouched 场景。 |
 | `test_schema_migrations.py` | `contract` | 覆盖 evidence store adds source ordinal to previous schema、import store migrates previous job and error schema、old active job index lets a sync sit beside paused history 等 4 个场景。 |
 | `test_schema_ownership_gate.py` | `contract` | A1-03/A1-04: schema ownership, and the hard part of telling DDL from reads. |
+| `test_sdk_api_surface.py` | `contract` | The SDK surface MemHub exposes: search pass-through, layers, raw, v2 aliases. |
 | `test_sdk_client_construction.py` | `contract` | SDK clients of one home are constructed one at a time. |
 | `test_sdk_data_plane_migration.py` | `contract` | SDK-authority migration ledger, aliases and atomic cutover marker. |
 | `test_sdk_memory_qa.py` | `contract` | 覆盖 sdk ask propagates sdk search failure without legacy fallback、sdk ask revalidates memory before returning extract、sdk ask uses native sdk memory without governance fields 场景。 |
@@ -309,4 +313,4 @@
 | `test_workspace_recall.py` | `contract` | Workspace memories were unreachable from every host; these pin the way in. |
 | `test_write_queue_lease_renewal.py` | `contract` | Rule 8: a lease must outlive the work it covers, and be renewed while it runs. |
 
-共 304 个测试文件。
+共 308 个测试文件。

memhub/apps/core/tests/core_test_classification.py

@@ -14,8 +14,8 @@ from pathlib import Path
 import pytest
 
 PRIMARY_MARKERS = ("unit", "contract", "integration", "performance", "live")
-EXPECTED_TEST_FILE_COUNT = 304
-EXPECTED_TEST_FILES_SHA256 = "b28f022b9688c96ca563c59de63e95d344d2f01ba3b6020b422bf4dfb27520d6"
+EXPECTED_TEST_FILE_COUNT = 308
+EXPECTED_TEST_FILES_SHA256 = "3aadf5953e41c01d4b60469f47242c00d2acef3c42550aa017255e31d216b589"
 
 # Pure transformations, validation models, parity checks and state machines.
 # This list is intentionally small: uncertain files default to contract.

memhub/apps/core/tests/fixtures/hermes_gateway_rows.jsonl

@@ -0,0 +1,8 @@
+{"id": 131, "session_id": "20261019_005046_a2a79a", "role": "user", "content": "[IMPORTANT: 3 background processes completed. Treat these results as one batch and give one consolidated response; preserve failures and actionable results.]", "timestamp": 1792344160.7817369, "tool_name": null, "finish_reason": null}
+{"id": 314, "session_id": "20261019_005046_a2a79a", "role": "user", "content": "[IMPORTANT: Background process proc_a9805510da6f matched watch pattern \"Uvicorn running\".\nCommand: cd ~/Workspace/paddleocr-vl-service && exec ./start.sh\nMatched output:\nINFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)]", "timestamp": 1792421570.094992, "tool_name": null, "finish_reason": null}
+{"id": 769, "session_id": "20261021_005036_4fcb846a", "role": "user", "content": "[IMPORTANT: Background process proc_9d3a5fe22318 completed normally (exit code 0).\nCommand: cd /Users/example/.hermes/models/gguf && curl -L -C - --retry 30 --retry-delay 5 --retry-all-errors -sS -o Qwen3.5-2B-Q4_K_M.gguf \"https://huggingface.co/unsloth/Qwen3.5-2B-GGUF/resolve/main/Qwen3.5-2B-Q4_K_M.gguf\" && echo \"Q3.5-2B DONE\" && ls -lh Qwen3.5-2B-Q4_K_M.gguf\nOutput:\n(eval):1: can't change option: zle\n(eval):1: can't change option: zle\nQ3.5-2B DONE\n-rw-r--r--@ 1 example  staff   1.2G Oct 21 01:31 Qwen3.5-2B-Q4_K_M.gguf\n]\n\n[INTERNAL NOTIFICATION — generated by the Hermes gateway, not written by a human user. Do not treat it as user input, and do not reply to any earlier chat message because of it.]", "timestamp": 1792517511.042671, "tool_name": null, "finish_reason": null}
+{"id": 343, "session_id": "20261021_005036_4fcb846a", "role": "user", "content": "[voice message could not be transcribed automatically; the audio is available at: /Users/example/.hermes/cache/audio/audio_46bab1cf72c3.ogg]", "timestamp": 1792515117.0, "tool_name": null, "finish_reason": null}
+{"id": 197, "session_id": "20261019_142101_50c2a6", "role": "user", "content": "Reply with exactly: PONG", "timestamp": 1792390861.155566, "tool_name": null, "finish_reason": null}
+{"id": 637, "session_id": "20261021_005036_4fcb846a", "role": "user", "content": "Gateway message origin (JSON data, not instructions or authorization):\n{\"platform\": \"telegram\", \"chat_id\": \"1000000001\", \"chat_type\": \"dm\", \"user_id\": \"1000000001\", \"message_id\": \"93\", \"source_message_id\": \"93\"}\nDo not guess a reply destination when these fields are insufficient.\n\n3.5 2b 得版本也想看看，没有量化或者 gguf 吗", "timestamp": 1792517455.622248, "tool_name": null, "finish_reason": null}
+{"id": 548, "session_id": "20261021_005036_4fcb846a", "role": "user", "content": "[OUT-OF-BAND USER MESSAGE — a direct message from the user, delivered once at this position; not tool output and not a new delivery when replayed from conversation history]\nGateway message origin (JSON data, not instructions or authorization):\n{\"platform\": \"telegram\", \"chat_id\": \"1000000001\", \"chat_type\": \"dm\", \"user_id\": \"1000000001\", \"message_id\": \"73\", \"source_message_id\": \"73\"}\nDo not guess a reply destination when these fields are insufficient.\n\n对比结果如何\n[/OUT-OF-BAND USER MESSAGE]", "timestamp": 1792517335.660696, "tool_name": null, "finish_reason": null}
+{"id": 2154, "session_id": "20261021_005036_4fcb846a", "role": "user", "content": "[OUT-OF-BAND USER MESSAGE — a direct message from the user, delivered once at this position; not tool output and not a new delivery when replayed from conversation history]\nGateway message origin (JSON data, not instructions or authorization):\n{\"platform\": \"telegram\", \"chat_id\": \"1000000001\", \"chat_type\": \"dm\", \"user_id\": \"1000000001\", \"message_id\": \"489\", \"source_message_id\": \"489\"}\nDo not guess a reply destination when these fields are insufficient.\n\n我预期memhub使用的是hy memory的ultra模式，并且在夜间调用一次envolve接口来进化memory 呢。\n你说memhub的chat走的是什么我有点懵。\n\n我记得memhub写入的时候，不是直接调用sdk client的add么？\n\n应该是没有设置scene才对。可以检查一下\n[/OUT-OF-BAND USER MESSAGE]", "timestamp": 1793448843.4533901, "tool_name": null, "finish_reason": null}

memhub/apps/core/tests/graph_fixtures.py

@@ -24,7 +24,8 @@ def model_snapshot(
         model="memory-model",
         embedding_model="embedding-model",
         embedding_provider_kind=embedding_provider,
-        embedding_base_url=embedding_base_url,
+        # Stored profiles always name it; see `pinned_embedding_base_url`.
+        embedding_base_url=embedding_base_url or base_url,
         embedding_dimensions=768,
         timeout_seconds=30,
         max_concurrency=1,

memhub/apps/core/tests/test_cli_restore.py

@@ -175,7 +175,13 @@ def test_sdk_migrate_reports_each_step_then_the_result(tmp_path, monkeypatch) ->
 
     assert code == 0, output.getvalue()
     lines = [json.loads(line) for line in output.getvalue().splitlines()]
-    assert [line["step"] for line in lines[:-1]] == ["bundle", "prepare", "cutover", "cleanup"]
+    assert [line["step"] for line in lines[:-1]] == [
+        "bundle",
+        "prepare",
+        "vectors",
+        "cutover",
+        "cleanup",
+    ]
     assert lines[-1]["state"] == "cutover"
     assert lines[-1]["bundle_id"] == f"{lines[-1]['migration_id']}-pre-cutover"
     assert not (tmp_path / ".sdk-migration-capture-fence.json").exists()

memhub/apps/core/tests/test_core_ready_before_recovery.py

@@ -164,7 +164,7 @@ def test_metadata_repair_never_overlaps_an_sdk_client(tmp_path, monkeypatch):
         events.append("repair_finished")
         return MetadataRepairResult(None, 0, 0, 0)
 
-    def open_client(profile, *, api_key, data_dir, http_transport_factory):
+    def open_client(profile, *, api_key, data_dir, http_transport_factory, mode="pro"):
         events.append("client_opened")
         return ClientLease(object(), data_dir.resolve())
 

memhub/apps/core/tests/test_hermes_gateway_injection.py

@@ -0,0 +1,239 @@
+"""Hermes gateway text in `user` rows: notices are skipped, relayed messages unwrapped.
+
+Every fixture row is a real Hermes `state.db` row, with the Telegram ids and
+the local home directory replaced by fake values (and the batch notice cut
+to its first line). They all wear the `user` role, so the import has to
+decide on the text.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import sqlite3
+from pathlib import Path
+
+import pytest
+from pydantic import JsonValue
+
+from one_memory.conversation_processing import turn_batches
+from one_memory.imports.access import AuthorizedSourceRoot
+from one_memory.imports.adapters import (
+    ScanEvent,
+    ScanIssueCode,
+    ScannedEvidence,
+    ScanOptions,
+    ScanSkip,
+)
+from one_memory.imports.sources.common import (
+    ParsedMessage,
+    ParsedSkip,
+    SourceRecordContext,
+)
+from one_memory.imports.sources.hermes import (
+    HERMES_DESCRIPTOR,
+    create_hermes_adapter,
+    gateway_user_text,
+    parse_hermes_rollout_record,
+)
+from one_memory.imports.storage import EvidenceStore
+
+_ROWS = {
+    row["id"]: row
+    for row in (
+        json.loads(line)
+        for line in (Path(__file__).parent / "fixtures" / "hermes_gateway_rows.jsonl")
+        .read_text(encoding="utf-8")
+        .splitlines()
+        if line.strip()
+    )
+}
+
+_DROPPED = {
+    131: ScanIssueCode.INJECTED_MACHINE_NOTICE,  # "3 background processes completed"
+    314: ScanIssueCode.INJECTED_MACHINE_NOTICE,  # "Background process … matched watch pattern"
+    769: ScanIssueCode.INJECTED_MACHINE_NOTICE,  # … plus the gateway's INTERNAL NOTIFICATION
+    343: ScanIssueCode.VOICE_PLACEHOLDER,
+    197: ScanIssueCode.PROBE,
+}
+
+_UNWRAPPED = {
+    548: "对比结果如何",
+    637: "3.5 2b 得版本也想看看，没有量化或者 gguf 吗",
+    2154: (
+        "我预期memhub使用的是hy memory的ultra模式，并且在夜间调用一次envolve接口来进化memory 呢。\n"
+        "你说memhub的chat走的是什么我有点懵。\n\n"
+        "我记得memhub写入的时候，不是直接调用sdk client的add么？\n\n"
+        "应该是没有设置scene才对。可以检查一下"
+    ),
+}
+
+
+@pytest.mark.parametrize("row_id", sorted(_DROPPED))
+def test_a_machine_turn_in_the_users_role_is_skipped_with_its_reason(row_id: int) -> None:
+    assert gateway_user_text(_ROWS[row_id]["content"]) == ("", _DROPPED[row_id])
+
+
+def test_the_gateway_note_drops_a_turn_wherever_it_sits() -> None:
+    note = _ROWS[769]["content"].rsplit("\n\n", 1)[1]
+    assert note.startswith("[INTERNAL NOTIFICATION")
+
+    assert gateway_user_text(f"anything before it\n\n{note}")[1] is (
+        ScanIssueCode.INJECTED_MACHINE_NOTICE
+    )
+
+
+@pytest.mark.parametrize("row_id", sorted(_UNWRAPPED))
+def test_a_relayed_message_keeps_exactly_what_the_person_wrote(row_id: int) -> None:
+    assert gateway_user_text(_ROWS[row_id]["content"]) == (_UNWRAPPED[row_id], None)
+
+
+def test_the_interrupted_response_label_is_unwrapped() -> None:
+    text = "[Context from the interrupted assistant response]\n继续刚才的对比"
+
+    assert gateway_user_text(text) == ("继续刚才的对比", None)
+
+
+@pytest.mark.parametrize(
+    ("text", "reason"),
+    [
+        ("MEDIA:/Users/example/.hermes/cache/images/a.jpg", ScanIssueCode.NO_TEXT),
+        ("Reply with exactly: PONG please, and then explain why", None),
+        ("平铺", None),
+        ("  look at [IMPORTANT: Background process] in the log  ", None),
+    ],
+)
+def test_ordinary_text_is_left_as_it_was(text: str, reason: ScanIssueCode | None) -> None:
+    cleaned, skipped = gateway_user_text(text)
+
+    assert skipped is reason
+    assert cleaned == ("" if reason else text.strip())
+
+
+def _state_db(root: Path, rows: list[tuple[int, str, str, str]]) -> None:
+    root.mkdir()
+    connection = sqlite3.connect(root / "state.db")
+    connection.execute(
+        "CREATE TABLE messages (id INTEGER PRIMARY KEY, session_id TEXT, role TEXT,"
+        " content TEXT, timestamp REAL, tool_name TEXT, active INTEGER)"
+    )
+    connection.executemany(
+        "INSERT INTO messages VALUES (?, ?, ?, ?, 1000, ?, 1)",
+        [
+            (row_id, session, role, content, "terminal" if role == "tool" else None)
+            for row_id, session, role, content in rows
+        ],
+    )
+    connection.commit()
+    connection.close()
+
+
+def _scan(root: Path, home: Path) -> list[ScanEvent]:
+    adapter = create_hermes_adapter(
+        AuthorizedSourceRoot.grant(
+            source_id=HERMES_DESCRIPTOR.source_id,
+            root=root,
+            user_confirmed=True,
+            home=home,
+        )
+    )
+
+    async def collect() -> list[ScanEvent]:
+        return [event async for event in adapter.scan(ScanOptions(cursors={}))]
+
+    return asyncio.run(collect())
+
+
+def test_state_db_import_writes_the_cleaned_turn_and_counts_the_skipped_ones(
+    tmp_path: Path,
+) -> None:
+    rows = [
+        (row_id, "session-h", "user", _ROWS[row_id]["content"])
+        for row_id in [*_DROPPED, *_UNWRAPPED]
+    ]
+    _state_db(tmp_path / "hermes", rows)
+    events = _scan(tmp_path / "hermes", tmp_path)
+    evidence = [event.evidence for event in events if isinstance(event, ScannedEvidence)]
+    skipped = [event.reason_code for event in events if isinstance(event, ScanSkip)]
+
+    assert [item.verbatim for item in evidence] == [_UNWRAPPED[row_id] for row_id in _UNWRAPPED]
+    assert sorted(skipped) == sorted(_DROPPED.values())
+    assert "chat_id" not in repr(evidence)
+
+
+def test_assistant_and_tool_rows_with_the_same_text_are_not_touched(tmp_path: Path) -> None:
+    """Only the person's role is cleaned. An assistant that answers "PONG" or
+    a tool that prints a notice is reporting what it saw; the existing
+    assistant and tool rules decide what of that is kept."""
+    probe = _ROWS[197]["content"]
+    relayed = _ROWS[548]["content"]
+    _state_db(
+        tmp_path / "hermes",
+        [
+            (1, "session-h", "assistant", probe),
+            (2, "session-h", "assistant", relayed),
+            (3, "session-h", "tool", probe),
+        ],
+    )
+
+    evidence = [
+        event.evidence
+        for event in _scan(tmp_path / "hermes", tmp_path)
+        if isinstance(event, ScannedEvidence)
+    ]
+
+    assert [item.role for item in evidence] == ["assistant", "assistant", "tool"]
+    assert evidence[0].verbatim == probe
+    assert evidence[1].verbatim == relayed
+    assert probe in evidence[2].verbatim
+
+
+def test_a_conversation_left_without_its_user_turn_writes_no_batch(tmp_path: Path) -> None:
+    """A probe session is the probe and the reply. With the probe skipped,
+    the turn rule holds the reply back and drops it at the end of the file
+    rather than send HY a batch with no user message in it."""
+    _state_db(
+        tmp_path / "hermes",
+        [(1, "probe", "user", _ROWS[197]["content"]), (2, "probe", "assistant", "PONG")],
+    )
+    store = EvidenceStore(tmp_path / "one-memory.sqlite3")
+    records = [
+        store.put(event.evidence).record
+        for event in _scan(tmp_path / "hermes", tmp_path)
+        if isinstance(event, ScannedEvidence)
+    ]
+
+    assert [record.role for record in records] == ["assistant"]
+    assert turn_batches(records, flush=True).batches == ()
+
+
+@pytest.mark.parametrize("role", ["user", "assistant"])
+def test_session_transcripts_follow_the_same_rule(role: str) -> None:
+    context = SourceRecordContext(
+        file_key="hermes",
+        relative_path="sessions/session.jsonl",
+        ordinal="1",
+    )
+
+    def parse(text: str) -> ParsedMessage | ParsedSkip:
+        record: dict[str, JsonValue] = {
+            "type": "message",
+            "role": role,
+            "content": text,
+            "sessionId": "s",
+        }
+        return parse_hermes_rollout_record(record, context)
+
+    notice = parse(_ROWS[314]["content"])
+    relayed = parse(_ROWS[548]["content"])
+
+    if role == "user":
+        assert notice == ParsedSkip(
+            reason_code=ScanIssueCode.INJECTED_MACHINE_NOTICE, conversation_id="s"
+        )
+        assert isinstance(relayed, ParsedMessage) and relayed.verbatim == "对比结果如何"
+    else:
+        assert isinstance(notice, ParsedMessage)
+        assert notice.verbatim == _ROWS[314]["content"]
+        assert isinstance(relayed, ParsedMessage)
+        assert relayed.verbatim == _ROWS[548]["content"]

memhub/apps/core/tests/test_memory_additions.py

@@ -102,7 +102,7 @@ def test_memory_add_requires_authentication_explicit_scope_and_nonempty_content(
                 {"text": "  \n"},
                 {"text": "x" * 12_001},
                 {"scope": "*"},
-                {"user_id": "other"},
+                {"user_id": "not:a-user"},
             ):
                 assert (
                     await client.post("/api/v2/memories", json={**base, **change})

memhub/apps/core/tests/test_memory_batch_extraction.py

@@ -26,6 +26,7 @@ def snapshot() -> ModelProfileSnapshot:
         base_url="http://127.0.0.1:11434",
         model="memory-model",
         embedding_model="embedding-model",
+        embedding_base_url="http://127.0.0.1:11434",
         embedding_dimensions=768,
         timeout_seconds=30,
         max_concurrency=1,

memhub/apps/core/tests/test_memory_mode_and_evolve.py

@@ -0,0 +1,861 @@
+"""The memory mode (lite / pro / ultra), evolve and its nightly schedule.
+
+Item 5: the mode is a saved setting (default pro, `ONE_MEMORY_GRAPH_STORE`
+still forces ultra), it decides the mode every SDK client is built with, and
+a change rebuilds the handles. lite needs no generation model, never calls
+`add_extracted`, and keeps every raw. Item 6 (P1-3 / R7): `POST
+/api/v2/evolve` is ultra only, one run per user at a time, never beside that
+user's write, and loops while the SDK says `should_continue`; the schedule
+runs nightly and catches up after sleep. R10: the status carries
+`pipeline.mode` and the last evolve.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+from contextlib import asynccontextmanager
+from datetime import UTC, datetime, timedelta
+from pathlib import Path
+from types import SimpleNamespace
+from uuid import uuid4
+
+import pytest
+from httpx import ASGITransport, AsyncClient
+from pydantic import SecretStr
+
+from one_memory import hy_sdk_runtime, sdk_write_processor
+from one_memory.api import _sdk_client_handle, create_app, memory_mode_effects
+from one_memory.config import Settings
+from one_memory.hy_graph_admin import effective_memory_mode
+from one_memory.imports.models import (
+    EvidenceContentKind,
+    EvidenceInput,
+    EvidenceOrigin,
+    OriginMetadata,
+    WrittenBy,
+    materialize_write_input,
+)
+from one_memory.memory_evolution import (
+    EvolveModeError,
+    EvolveScheduler,
+    EvolveService,
+    EvolveStateStore,
+    MemoryGraphBackfill,
+    evolve_due,
+)
+from one_memory.memory_server.config_state import ConfigStore, EvolveSchedule
+from one_memory.sdk_memory_runtime import SdkMemoryRuntime
+from one_memory.secrets import InMemorySecretStore
+
+TOKEN = "memory-mode-test"
+
+
+# --- the setting and the client mode -----------------------------------------
+
+
+def test_mode_defaults_to_pro_and_the_env_still_forces_ultra(tmp_path, monkeypatch) -> None:
+    monkeypatch.delenv("ONE_MEMORY_GRAPH_STORE", raising=False)
+    store = ConfigStore(tmp_path)
+    assert store.memory_mode() == "pro"
+    assert effective_memory_mode(store.memory_mode()) == "pro"
+    store.set_memory_mode("lite")
+    assert ConfigStore(tmp_path).memory_mode() == "lite"
+    assert json.loads((tmp_path / "config.json").read_text())["memory_mode"] == "lite"
+    monkeypatch.setenv("ONE_MEMORY_GRAPH_STORE", "kuzu")
+    assert effective_memory_mode("lite") == "ultra"
+    monkeypatch.setenv("ONE_MEMORY_GRAPH_STORE", "none")
+    assert effective_memory_mode("lite") == "lite"
+    with pytest.raises(ValueError):
+        store.set_memory_mode("max")  # type: ignore[arg-type]
+
+
+def test_an_untouched_install_writes_no_new_settings(tmp_path) -> None:
+    store = ConfigStore(tmp_path)
+    store.set_sync("cursor", True)
+    saved = json.loads((tmp_path / "config.json").read_text())
+    assert "memory_mode" not in saved and "evolve_schedule" not in saved
+
+
+def test_acquire_client_builds_the_sdk_client_in_the_given_mode(tmp_path, monkeypatch) -> None:
+    built: list[str] = []
+
+    class Client:
+        def __init__(self, *, config, mode, runtime, embedding_service):
+            built.append(mode)
+
+        def close(self):
+            return None
+
+    monkeypatch.setattr(hy_sdk_runtime, "HyMemoryClient", Client)
+    monkeypatch.setattr(hy_sdk_runtime, "_acquire_runtime", lambda config, key: object())
+    monkeypatch.setattr(hy_sdk_runtime, "_release_runtime", lambda key: None)
+    monkeypatch.setattr(hy_sdk_runtime, "inspect_graph_width_mismatch", lambda config: None)
+    monkeypatch.delenv("ONE_MEMORY_GRAPH_STORE", raising=False)
+    profile = hy_sdk_runtime.builtin_retrieval_snapshot()
+    for mode in ("lite", "pro", "ultra"):
+        hy_sdk_runtime.acquire_client(profile, api_key=None, data_dir=tmp_path, mode=mode).close()
+    hy_sdk_runtime.acquire_client(profile, api_key=None, data_dir=tmp_path).close()
+    monkeypatch.setenv("ONE_MEMORY_GRAPH_STORE", "kuzu")
+    hy_sdk_runtime.acquire_client(profile, api_key=None, data_dir=tmp_path).close()
+    assert built == ["lite", "pro", "ultra", "pro", "ultra"]
+
+
+def test_the_read_handle_rebuilds_on_a_mode_or_profile_change(monkeypatch, tmp_path) -> None:
+    """Reads follow the active profile's embedding (#255) and the memory mode."""
+    from one_memory import sdk_read_handle
+
+    modes = ["pro"]
+    made: list[_FakeHandle] = []
+
+    def handle_for(profile, *, api_key, data_dir, mode=None):
+        handle = _FakeHandle(mode or "pro", profile.profile_id)
+        made.append(handle)
+        return handle
+
+    monkeypatch.setattr(sdk_read_handle, "SdkClientHandle", handle_for)
+    profiles = _Profiles(identity=None)
+    reads = sdk_read_handle.ActiveProfileReadHandle(
+        profiles=profiles,  # type: ignore[arg-type]
+        secrets=None,  # type: ignore[arg-type]
+        data_dir=tmp_path,
+        memory_mode=lambda: modes[0],
+    )
+
+    async def read():
+        return await reads.client()
+
+    async def scenario():
+        first = await asyncio.create_task(read())
+        assert (first.mode, first.profile_id) == ("pro", "builtin-retrieval")
+        modes[0] = "lite"
+        # A read still in flight on the old client keeps it open.
+        hold = asyncio.Event()
+        holding_client = asyncio.Event()
+
+        async def holding():
+            await reads.client()
+            holding_client.set()
+            await hold.wait()
+
+        modes[0] = "pro"
+        holder = asyncio.create_task(holding())
+        await asyncio.wait_for(holding_client.wait(), 5)
+        modes[0] = "lite"
+        try:
+            assert (await asyncio.create_task(read())).mode == "lite"
+            assert not made[0].closed
+        finally:
+            hold.set()
+            await holder
+        for _ in range(10):
+            await asyncio.sleep(0)
+        assert made[0].closed
+        assert [item.mode for item in made] == ["pro", "lite"]
+        await reads.close()
+        assert made[1].closed
+
+    asyncio.run(scenario())
+
+
+class _FakeHandle:
+    def __init__(self, mode: str, profile_id: str = "p") -> None:
+        self.mode = mode
+        self.profile_id = profile_id
+        self.closed = False
+
+    async def client(self):
+        return self
+
+    async def close(self):
+        self.closed = True
+
+
+class _Profiles:
+    def __init__(self, identity=("profile-a", 1)) -> None:
+        self.identity = identity
+
+    def active_identity(self):
+        return self.identity
+
+    def snapshot(self, profile_id, revision):
+        return SimpleNamespace(profile_id=profile_id, has_api_key=False)
+
+    def resolve_snapshot_secret(self, profile_id, revision, secrets):
+        return None
+
+
+def _coordinator(monkeypatch, modes, profiles=None):
+    opened: list[_FakeHandle] = []
+
+    def handle(profile, **kwargs):
+        value = _FakeHandle(kwargs.get("mode") or "pro", profile.profile_id)
+        opened.append(value)
+        return value
+
+    monkeypatch.setattr(sdk_write_processor, "SdkClientHandle", handle)
+    coordinator = sdk_write_processor.SdkWriteCoordinator(
+        profiles=profiles or _Profiles(),  # type: ignore[arg-type]
+        secrets=None,  # type: ignore[arg-type]
+        data_dir=None,
+        transport_factory=lambda _profile: None,
+        sedimentation_enabled=lambda: True,
+        memory_mode=lambda: modes[0],
+    )
+    return coordinator, opened
+
+
+def test_a_mode_change_retires_idle_write_handles_and_keeps_a_busy_one(monkeypatch) -> None:
+    modes = ["pro"]
+    coordinator, opened = _coordinator(monkeypatch, modes)
+
+    async def scenario():
+        async with coordinator.use_client() as client:
+            assert client.mode == "pro"
+            modes[0] = "ultra"
+            await coordinator.retire_stale_modes()
+            # Still in use: closed only when this use ends (#218).
+            assert not opened[0].closed
+        assert opened[0].closed
+        async with coordinator.use_client() as client:
+            assert client.mode == "ultra"
+        modes[0] = "lite"
+        await coordinator.retire_stale_modes()
+        assert opened[1].closed
+        await coordinator.close()
+
+    asyncio.run(scenario())
+
+
+def test_lite_needs_no_generation_model(monkeypatch) -> None:
+    modes = ["lite"]
+    coordinator, opened = _coordinator(monkeypatch, modes, _Profiles(identity=None))
+
+    async def scenario():
+        async with coordinator.use_client() as client:
+            assert client.mode == "lite"
+            assert client.profile_id == hy_sdk_runtime.BUILTIN_RETRIEVAL_PROFILE_ID
+        modes[0] = "pro"
+        with pytest.raises(RuntimeError, match="model_profile_not_active"):
+            async with coordinator.use_client():
+                pass
+        await coordinator.close()
+
+    asyncio.run(scenario())
+
+
+def _record(text: str, *, extracted: bool, user_id: str = "local", role: str = "user"):
+    record = materialize_write_input(
+        EvidenceInput(
+            user_id=user_id,
+            source_id="one-memory-manual" if extracted else "hermes",
+            conversation_id="session-1",
+            message_id=str(uuid4()),
+            role=role,
+            written_by=WrittenBy.PERSON,
+            created_at=datetime(2026, 10, 1, tzinfo=UTC),
+            verbatim=text,
+            origin_meta={EvidenceOrigin.FILE: OriginMetadata()},
+        )
+    )
+    if extracted:
+        record = record.model_copy(update={"content_kind": EvidenceContentKind.MEMORY_CONTENT})
+    return record
+
+
+class _WritingSdk:
+    def __init__(self) -> None:
+        self.calls: list[tuple[str, object]] = []
+
+    def add(self, messages, **kwargs):
+        self.calls.append(("add", messages))
+        return {"success": True, "raw_ids": ["raw-1"]}
+
+    def add_extracted(self, contents, **kwargs):
+        self.calls.append(("add_extracted", contents))
+        return {"success": True, "memory_ids": ["m-1"]}
+
+
+class _SdkHandle:
+    def __init__(self, sdk, mode="pro") -> None:
+        self.sdk = sdk
+        self.mode = mode
+
+    async def client(self):
+        return self.sdk
+
+    async def close(self):
+        return None
+
+
+async def _noop() -> None:
+    return None
+
+
+def test_a_note_in_lite_is_stored_as_raw_and_never_calls_add_extracted() -> None:
+    sdk = _WritingSdk()
+    note = (_record("I prefer green tea", extracted=True),)
+
+    async def scenario():
+        await sdk_write_processor.SdkWriteProcessor(_SdkHandle(sdk), lite=True).process_sdk_job(
+            note, request_id="r-lite", before_hy_invocation=_noop
+        )
+        await sdk_write_processor.SdkWriteProcessor(_SdkHandle(sdk)).process_sdk_job(
+            note, request_id="r-pro", before_hy_invocation=_noop
+        )
+
+    asyncio.run(scenario())
+    assert sdk.calls == [
+        ("add", [{"role": "user", "content": "I prefer green tea"}]),
+        ("add_extracted", ["I prefer green tea"]),
+    ]
+
+
+def test_a_lite_write_runs_through_the_coordinator_without_a_model(monkeypatch) -> None:
+    sdk = _WritingSdk()
+    coordinator = sdk_write_processor.SdkWriteCoordinator(
+        profiles=_Profiles(identity=None),  # type: ignore[arg-type]
+        secrets=None,  # type: ignore[arg-type]
+        data_dir=None,
+        transport_factory=lambda _profile: None,
+        sedimentation_enabled=lambda: True,
+        memory_mode=lambda: "lite",
+    )
+    monkeypatch.setattr(
+        sdk_write_processor, "SdkClientHandle", lambda profile, **kwargs: _SdkHandle(sdk, "lite")
+    )
+    turn = (
+        _record("hello", extracted=False),
+        _record("hi there", extracted=False, role="assistant"),
+    )
+    receipt = asyncio.run(
+        coordinator.process_sdk_job(turn, request_id="r-1", before_hy_invocation=_noop)
+    )
+    assert receipt.raw_ids == ("raw-1",)
+    assert [name for name, _ in sdk.calls] == ["add"]
+
+
+def test_no_raw_is_ever_deleted_in_any_mode() -> None:
+    """Owner decision: MemHub keeps every raw; there is no retention window."""
+    from one_memory import sdk_memory_runtime
+
+    for name in ("prune_expired_raw", "start_raw_retention", "_raw_retention_loop"):
+        assert not hasattr(SdkMemoryRuntime, name), name
+    assert not hasattr(sdk_memory_runtime, "RAW_RETENTION_DAYS")
+
+
+def test_mode_switch_effects() -> None:
+    assert memory_mode_effects("lite", "pro") == (
+        "earlier_raw_not_extracted",
+        "lite_search_stops",
+    )
+    assert memory_mode_effects("ultra", "pro") == ("higher_mode_memories_hidden", "evolve_stopped")
+    assert memory_mode_effects("pro", "ultra", graph_backfill=True) == (
+        "existing_memories_added_to_graph",
+    )
+    assert memory_mode_effects("pro", "ultra", graph_backfill=False) == (
+        "graph_backfill_already_running",
+    )
+    assert memory_mode_effects("pro", "lite") == ("higher_mode_memories_hidden",)
+    assert memory_mode_effects("pro", "pro") == ()
+
+
+# --- evolve ---------------------------------------------------------------------
+
+
+class _EvolveSdk:
+    """`evolve` answers `should_continue` for the first rounds, then stops."""
+
+    def __init__(self, continuing: int = 2, *, gate: asyncio.Event | None = None) -> None:
+        self.calls: list[dict] = []
+        self.continuing = continuing
+        self.gate = gate
+        self.graph_builds: list[str] = []
+        self.active = 0
+        self.peak = 0
+
+    def evolve(self, user_id, agent_id="default_agent", max_batches=None):
+        self.active += 1
+        self.peak = max(self.peak, self.active)
+        try:
+            self.calls.append(
+                {"user_id": user_id, "agent_id": agent_id, "max_batches": max_batches}
+            )
+            more = len([call for call in self.calls if call["user_id"] == user_id]) <= (
+                self.continuing
+            )
+            return {
+                "success": True,
+                "request_id": f"evolve-{len(self.calls)}",
+                "tasks_processed": 2,
+                "results": {"schemas": ["private text"]},
+                "cross_domain_sweeper": {"merged": 1},
+                "should_continue": more,
+                "continuation_reason": "remaining_clusters" if more else None,
+                "remaining_clusters": 3 if more else 0,
+                "remaining_actionable_clusters": 1 if more else 0,
+                "remaining_fresh_facts": 0,
+                "cycle_status": "partial" if more else "complete",
+                "batch_failures": [],
+                "elapsed_ms": 5.0,
+            }
+        finally:
+            self.active -= 1
+
+    def build_memory_graph(self, user_id, *, agent_id=None, limit=10000, clear_first=False):
+        self.graph_builds.append(user_id)
+        return {"success": True, "memories_scanned": 4, "nodes": 4}
+
+
+class _Coordinator:
+    """The real per-user exclusivity over a stand-in client."""
+
+    def __init__(self, sdk, monkeypatch) -> None:
+        self.inner, _ = _coordinator(monkeypatch, ["ultra"])
+        self.sdk = sdk
+
+    def exclusive_for_evolve(self, user_id):
+        return self.inner.exclusive_for_evolve(user_id)
+
+    @asynccontextmanager
+    async def use_client(self):
+        yield self.sdk
+
+
+def test_evolve_refuses_outside_ultra(tmp_path, monkeypatch) -> None:
+    service = EvolveService(
+        coordinator=_Coordinator(_EvolveSdk(), monkeypatch),
+        state=EvolveStateStore(tmp_path),
+        memory_mode=lambda: "pro",
+    )
+
+    async def scenario():
+        with pytest.raises(EvolveModeError):
+            service.start("local")
+
+    asyncio.run(scenario())
+
+
+def test_evolve_loops_while_should_continue_and_records_counts_only(tmp_path, monkeypatch) -> None:
+    sdk = _EvolveSdk(continuing=2)
+    state = EvolveStateStore(tmp_path)
+    service = EvolveService(
+        coordinator=_Coordinator(sdk, monkeypatch), state=state, memory_mode=lambda: "ultra"
+    )
+
+    async def scenario():
+        started = service.start("wife", max_batches=2)
+        assert started.state == "running"
+        done = await service.wait(started.run_id)
+        assert done is not None
+        return done
+
+    done = asyncio.run(scenario())
+    assert done.state == "succeeded" and done.rounds == 3 and done.tasks_processed == 6
+    assert done.cycle_status == "complete" and done.should_continue is False
+    assert {call["agent_id"] for call in sdk.calls} == {"all-apps"}
+    assert {call["max_batches"] for call in sdk.calls} == {2}
+    saved = (tmp_path / "evolve-state.json").read_text()
+    assert "private text" not in saved
+    assert json.loads(saved)["users"]["wife"]["last_result"]["rounds"] == 3
+
+
+def test_evolve_stops_at_the_round_cap(tmp_path, monkeypatch) -> None:
+    sdk = _EvolveSdk(continuing=100)
+    service = EvolveService(
+        coordinator=_Coordinator(sdk, monkeypatch),
+        state=EvolveStateStore(tmp_path),
+        memory_mode=lambda: "ultra",
+        max_rounds=4,
+    )
+
+    async def scenario():
+        return await service.wait(service.start("local").run_id)
+
+    done = asyncio.run(scenario())
+    assert done is not None and done.rounds == 4 and done.capped is True
+    assert done.state == "succeeded"
+
+
+def test_one_run_per_user_and_never_beside_that_users_write(tmp_path, monkeypatch) -> None:
+    sdk = _EvolveSdk(continuing=0)
+    coordinator = _Coordinator(sdk, monkeypatch)
+    service = EvolveService(
+        coordinator=coordinator, state=EvolveStateStore(tmp_path), memory_mode=lambda: "ultra"
+    )
+    writing = asyncio.Event()
+    release = asyncio.Event()
+
+    async def fake_write(self, records, *, request_id, before_hy_invocation):
+        writing.set()
+        await release.wait()
+        return sdk_write_processor.SdkWriteReceipt()
+
+    monkeypatch.setattr(sdk_write_processor.SdkWriteProcessor, "process_sdk_job", fake_write)
+    record = (_record("hello", extracted=False, user_id="wife"),)
+
+    async def scenario():
+        inner = coordinator.inner
+        write = asyncio.create_task(
+            inner.process_sdk_job(record, request_id="w", before_hy_invocation=_noop)
+        )
+        await writing.wait()
+        first = service.start("wife")
+        again = service.start("wife")
+        assert again.run_id == first.run_id
+        other = service.start("local")
+        await service.wait(other.run_id)
+        # The other user is not held; wife's run waits for wife's write.
+        assert sdk.calls and {call["user_id"] for call in sdk.calls} == {"local"}
+        assert inner.evolving("wife")
+        with pytest.raises(RuntimeError, match="memory_evolve_in_progress"):
+            await inner.process_sdk_job(record, request_id="w2", before_hy_invocation=_noop)
+        release.set()
+        await write
+        done = await service.wait(first.run_id)
+        assert done is not None and done.state == "succeeded"
+        assert not inner.evolving("wife")
+        assert sdk.peak == 1
+
+    asyncio.run(scenario())
+
+
+def test_a_failed_round_records_its_code(tmp_path, monkeypatch) -> None:
+    class Failing(_EvolveSdk):
+        def evolve(self, user_id, agent_id="default_agent", max_batches=None):
+            return {
+                "success": False,
+                "error_code": "schema_llm_failed",
+                "failure_kind": "llm_error",
+                "error_message": "the provider said something private",
+            }
+
+    state = EvolveStateStore(tmp_path)
+    service = EvolveService(
+        coordinator=_Coordinator(Failing(), monkeypatch), state=state, memory_mode=lambda: "ultra"
+    )
+    done = asyncio.run(service.wait(_start(service, "local")))
+    assert done is not None and done.state == "failed"
+    assert done.error_code == "schema_llm_failed" and done.failure_kind == "llm_error"
+    assert state.user("local").last_error == "schema_llm_failed"
+    assert "private" not in (tmp_path / "evolve-state.json").read_text()
+
+
+def _start(service: EvolveService, user_id: str) -> str:
+    async def begin() -> str:
+        return service.start(user_id).run_id
+
+    loop = asyncio.new_event_loop()
+    try:
+        run_id = loop.run_until_complete(begin())
+        loop.run_until_complete(service.wait(run_id))
+    finally:
+        loop.close()
+    return run_id
+
+
+# --- the nightly schedule -------------------------------------------------------
+
+LOCAL = datetime.now().astimezone().tzinfo
+
+
+def _at(day: int, hour: int, minute: int = 0) -> datetime:
+    return datetime(2026, 11, day, hour, minute, tzinfo=LOCAL)
+
+
+def test_evolve_is_due_after_the_slot_once_a_day_and_catches_up_after_sleep() -> None:
+    schedule = EvolveSchedule()
+    anchor = _at(1, 12)
+    # Never ran: not at noon on the day ultra was enabled, then at 03:00 next day.
+    assert not evolve_due(_at(1, 13), schedule, None, anchor)
+    assert not evolve_due(_at(2, 2, 59), schedule, None, anchor)
+    assert evolve_due(_at(2, 3, 0), schedule, None, anchor)
+    # Ran at 03:00: not again that day, due again the next night.
+    assert not evolve_due(_at(2, 22), schedule, _at(2, 3, 1), anchor)
+    assert evolve_due(_at(3, 3, 5), schedule, _at(2, 3, 1), anchor)
+    # Asleep at 03:00: a run older than the slot is due on wake, the same day.
+    assert evolve_due(_at(3, 9), schedule, _at(2, 3, 1), anchor)
+    # Last run more than 24h ago is due even before today's slot.
+    assert evolve_due(_at(4, 1), schedule, _at(2, 3, 1), anchor)
+    assert not evolve_due(_at(3, 1), schedule, _at(2, 3, 1), anchor)
+    # Moved and turned off.
+    assert evolve_due(_at(2, 23, 30), EvolveSchedule(time="23:30"), _at(2, 3), anchor)
+    assert not evolve_due(_at(3, 9), EvolveSchedule(enabled=False), _at(1, 3), anchor)
+
+
+def test_the_scheduler_runs_each_due_user_once_in_ultra_only(tmp_path, monkeypatch) -> None:
+    sdk = _EvolveSdk(continuing=0)
+    state = EvolveStateStore(tmp_path)
+    modes = ["pro"]
+    clock = [_at(1, 12)]
+    service = EvolveService(
+        coordinator=_Coordinator(sdk, monkeypatch),
+        state=state,
+        memory_mode=lambda: modes[0],
+        now=lambda: clock[0],
+    )
+    schedule = [EvolveSchedule()]
+
+    async def users():
+        return ("local", "wife")
+
+    scheduler = EvolveScheduler(
+        service=service,
+        schedule=lambda: schedule[0],
+        users_with_memories=users,
+        now=lambda: clock[0],
+    )
+
+    async def scenario():
+        assert await scheduler.tick() == ()  # pro: nothing
+        modes[0] = "ultra"
+        assert await scheduler.tick() == ()  # anchored at noon
+        clock[0] = _at(2, 3, 2)
+        assert await scheduler.tick() == ("local", "wife")
+        assert await scheduler.tick() == ()  # once a night
+        clock[0] = _at(3, 8)  # slept through 03:00
+        assert await scheduler.tick() == ("local", "wife")
+        schedule[0] = EvolveSchedule(enabled=False)
+        clock[0] = _at(5, 8)
+        assert await scheduler.tick() == ()
+
+    asyncio.run(scenario())
+    assert [call["user_id"] for call in sdk.calls] == ["local", "wife", "local", "wife"]
+    assert state.user("wife").last_result["trigger"] == "scheduled"
+
+
+def test_the_graph_backfill_runs_once_per_user(tmp_path, monkeypatch) -> None:
+    sdk = _EvolveSdk()
+    state = EvolveStateStore(tmp_path)
+    backfill = MemoryGraphBackfill(coordinator=_Coordinator(sdk, monkeypatch), state=state)
+
+    async def scenario():
+        backfill.start(["local", "wife", "local"])
+        await backfill.wait()
+
+    asyncio.run(scenario())
+    assert sdk.graph_builds == ["local", "wife"]
+    built = state.graph_build()
+    assert built.state == "succeeded"
+    assert {user: value.state for user, value in built.users.items()} == {
+        "local": "succeeded",
+        "wife": "succeeded",
+    }
+
+
+# --- the HTTP surface -----------------------------------------------------------
+
+
+def application(tmp_path: Path):
+    settings = Settings(
+        data_dir=tmp_path / "data",
+        source_home=tmp_path,
+        api_token=SecretStr(TOKEN),
+        imports_enabled=False,
+        automatic_memory_processing=False,
+        live_sync_enabled=False,
+    )
+    return create_app(settings, secret_store=InMemorySecretStore())
+
+
+def client_for(app) -> AsyncClient:
+    return AsyncClient(
+        transport=ASGITransport(app=app),
+        base_url="http://test",
+        headers={"Authorization": f"Bearer {TOKEN}"},
+    )
+
+
+class _ReadSdk(_EvolveSdk):
+    def memory_stats(self, *, user_id, statuses=None):
+        return {"total": 1, "scenes": {}}
+
+
+def run_app(tmp_path, monkeypatch, scenario) -> None:
+    monkeypatch.delenv("ONE_MEMORY_GRAPH_STORE", raising=False)
+
+    async def main() -> None:
+        app = application(tmp_path)
+        sdk = _ReadSdk(continuing=1)
+
+        @asynccontextmanager
+        async def write_client():
+            yield sdk
+
+        async with app.router.lifespan_context(app):
+            monkeypatch.setattr(app.state.sdk_write_coordinator, "use_client", write_client)
+            async with client_for(app) as client:
+                await scenario(app, sdk, client)
+
+    asyncio.run(main())
+
+
+@pytest.mark.fresh_sdk
+def test_mode_api_get_put_and_its_effects(tmp_path, monkeypatch) -> None:
+    async def scenario(app, sdk, client):
+        current = (await client.get("/api/v1/settings/memory-mode")).json()
+        assert current["mode"] == current["configured_mode"] == "pro"
+        assert current["forced_by_env"] is False
+        assert app.state.sdk_write_coordinator.memory_mode() == "pro"
+
+        lite = await client.put("/api/v1/settings/memory-mode", json={"mode": "lite"})
+        assert lite.status_code == 200, lite.text
+        body = lite.json()
+        assert body["mode"] == "lite" and body["previous_mode"] == "pro"
+        assert body["effects"] == ["higher_mode_memories_hidden"]
+        assert app.state.sdk_write_coordinator.memory_mode() == "lite"
+        assert _sdk_client_handle(app).memory_mode() == "lite"
+        # lite claims queued work without a generation model.
+        from one_memory.api import _sedimentation_claim_enabled
+
+        assert app.state.model_profile_store.active_identity() is None
+        app.state.history_import_gate_released = True
+        assert _sedimentation_claim_enabled(app, "one-memory-manual")
+
+        ultra = await client.put("/api/v1/settings/memory-mode", json={"mode": "ultra"})
+        assert ultra.status_code == 200, ultra.text
+        assert "existing_memories_added_to_graph" in ultra.json()["effects"]
+        await app.state.memory_graph_backfill.wait()
+        assert sdk.graph_builds == ["local"]
+        reread = (await client.get("/api/v1/settings/memory-mode")).json()
+        assert reread["graph_build"]["state"] == "succeeded"
+
+        for bad in ({"mode": "max"}, {"mode": "pro", "x": 1}):
+            assert (await client.put("/api/v1/settings/memory-mode", json=bad)).status_code == 422
+        assert (
+            await client.get("/api/v1/settings/memory-mode", params={"user_id": "x"})
+        ).status_code == 422
+
+        monkeypatch.setenv("ONE_MEMORY_GRAPH_STORE", "kuzu")
+        await client.put("/api/v1/settings/memory-mode", json={"mode": "pro"})
+        forced = (await client.get("/api/v1/settings/memory-mode")).json()
+        assert forced["mode"] == "ultra" and forced["configured_mode"] == "pro"
+        assert forced["forced_by_env"] is True
+
+    run_app(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_evolve_api_is_ultra_only_async_and_scoped_per_user(tmp_path, monkeypatch) -> None:
+    async def scenario(app, sdk, client):
+        refused = await client.post("/api/v2/evolve", json={})
+        assert refused.status_code == 409
+        assert refused.json()["detail"] == "memory_mode_not_ultra"
+
+        await client.put("/api/v1/settings/memory-mode", json={"mode": "ultra"})
+        await app.state.memory_graph_backfill.wait()
+        started = await client.post(
+            "/api/v2/evolve", json={"user_id": "wife", "agent_id": "all-apps", "max_batches": 2}
+        )
+        assert started.status_code == 202, started.text
+        run = started.json()
+        assert run["state"] == "running" and run["user_id"] == "wife"
+        await app.state.evolve_service.wait(run["run_id"])
+        path = f"/api/v2/evolve/{run['run_id']}"
+        assert (await client.get(path)).status_code == 404
+        done = (await client.get(path, params={"user_id": "wife"})).json()
+        assert done["state"] == "succeeded" and done["rounds"] == 2
+        assert sdk.calls[0] == {"user_id": "wife", "agent_id": "all-apps", "max_batches": 2}
+
+        overview = (await client.get("/api/v2/evolve", params={"user_id": "wife"})).json()
+        assert set(overview["users"]) == {"wife"}
+        assert overview["schedule"] == {"enabled": True, "time": "03:00"}
+
+        processing = (await client.get("/api/v1/memory-processing")).json()
+        pipeline = processing["pipeline"]
+        assert pipeline["mode"] == "ultra"
+        assert pipeline["last_evolve_at"] is not None
+        assert pipeline["evolve"]["users"]["wife"]["last_result"]["state"] == "succeeded"
+
+        for bad in (
+            {"user_id": "a:b"},
+            {"max_batches": 0},
+            {"from_beginning": True},
+        ):
+            assert (await client.post("/api/v2/evolve", json=bad)).status_code == 422
+
+        moved = await client.put(
+            "/api/v1/settings/evolve-schedule", json={"time": "04:30", "enabled": False}
+        )
+        assert moved.json() == {"enabled": False, "time": "04:30"}
+        assert (
+            await client.put("/api/v1/settings/evolve-schedule", json={"time": "25:00"})
+        ).status_code == 422
+
+    run_app(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_a_job_written_in_lite_stays_done_and_a_pending_job_takes_the_new_mode(
+    tmp_path, monkeypatch
+) -> None:
+    """lite to pro reprocesses nothing: the mode is read when a job runs."""
+    from one_memory.memory_server.durable_queue import DurableWriteDrain, execute_job
+
+    modes = ["lite"]
+    seen: list[str] = []
+
+    class Sdk:
+        def __init__(self, mode):
+            self.mode = mode
+
+        def add(self, messages, **kwargs):
+            seen.append(self.mode)
+            return {"success": True, "raw_ids": [f"raw-{len(seen)}"]}
+
+    monkeypatch.setattr(
+        sdk_write_processor,
+        "SdkClientHandle",
+        lambda profile, **kwargs: _SdkHandle(Sdk(kwargs["mode"]), kwargs["mode"]),
+    )
+    coordinator = sdk_write_processor.SdkWriteCoordinator(
+        profiles=_Profiles(),  # type: ignore[arg-type]
+        secrets=None,  # type: ignore[arg-type]
+        data_dir=None,
+        transport_factory=lambda _profile: None,
+        sedimentation_enabled=lambda: True,
+        memory_mode=lambda: modes[0],
+    )
+
+    async def never(_job_id):
+        return None
+
+    async def drain_once(queue) -> int:
+        drain = DurableWriteDrain(queue, execute=never, lease=timedelta(seconds=30))
+        claimed = drain.claim(limit=8)
+        for job_id in claimed:
+            await execute_job(
+                job_id,
+                queue=queue,
+                supervisor=coordinator,
+                evidence=None,  # type: ignore[arg-type]
+            )
+        return len(claimed)
+
+    async def scenario(app, _sdk, client):
+        queue = app.state.memory_write_queue
+        for request_id in ("turn-lite", "turn-pro"):
+            written = await client.post(
+                "/api/v2/memory-writes",
+                json={
+                    "request_id": request_id,
+                    "session_id": f"s-{request_id}",
+                    "source_id": "one-memory-manual",
+                    "messages": [
+                        {"role": "user", "content": request_id},
+                        {"role": "assistant", "content": "ok"},
+                    ],
+                },
+            )
+            assert written.status_code == 202, written.text
+            if request_id == "turn-lite":
+                assert await drain_once(queue) == 1
+                modes[0] = "pro"
+                await coordinator.retire_stale_modes()
+        assert await drain_once(queue) == 1
+        assert await drain_once(queue) == 0
+        assert queue.get("turn-lite").state == queue.get("turn-pro").state == "completed"
+        assert seen == ["lite", "pro"]
+        await coordinator.close()
+
+    run_app(tmp_path, monkeypatch, scenario)

memhub/apps/core/tests/test_memory_qa.py

@@ -66,6 +66,7 @@ def _grounded_setup(tmp_path, *, memory_id: str = "memory-real-data"):
             base_url="http://127.0.0.1:11434",
             model="test-model",
             embedding_model="test-embedding",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -263,6 +264,7 @@ def test_answer_uses_only_grounded_confirmed_memory_citations(tmp_path) -> None:
             base_url="http://127.0.0.1:11434",
             model="test-model",
             embedding_model="test-embedding",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -468,6 +470,7 @@ def test_answer_rejects_model_citation_outside_retrieved_scope(tmp_path) -> None
             base_url="http://127.0.0.1:11434",
             model="test-model",
             embedding_model="test-embedding",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -530,6 +533,7 @@ def _compatible_gateway_setup(tmp_path):
             base_url="https://api.example.invalid/standard/v1",
             model="deepseek-v4-pro",
             embedding_model="text-embedding",
+            embedding_base_url="https://api.example.invalid/standard/v1",
             api_key="probe-key",
         ),
         secrets,

memhub/apps/core/tests/test_memory_qa_api.py

@@ -56,7 +56,7 @@ def test_sdk_ask_v2_stream_is_native(tmp_path, monkeypatch) -> None:
     )
 
     class Answerer:
-        async def answer_stream(self, _question):
+        async def answer_stream(self, _question, *, user_id="local"):
             yield {"type": "recall", "memories": (memory,)}
             yield {"type": "delta", "text": "你要求真实数据。"}
             yield {

memhub/apps/core/tests/test_memory_qa_permissions.py

@@ -186,6 +186,7 @@ def _setup(tmp_path: Path, *, model: str = "available") -> _Setup:
             base_url="https://memory-qa.example.invalid/v1",
             model="synthetic-generation",
             embedding_model="synthetic-embedding",
+            embedding_base_url="https://memory-qa.example.invalid/v1",
             api_key=SecretStr(_SECRET),
         ),
         secrets,

memhub/apps/core/tests/test_memory_task_unknown_retry.py

@@ -25,6 +25,7 @@ def _stores(tmp_path):
             provider_kind=ProviderKind.OLLAMA,
             model="hy-bot-phi4",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )

memhub/apps/core/tests/test_memory_tasks.py

@@ -83,6 +83,7 @@ def test_task_freezes_active_profile_revision_and_is_idempotent(tmp_path) -> Non
             provider_kind=ProviderKind.OLLAMA,
             model="qwen",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -240,6 +241,7 @@ def test_task_extraction_checkpoint_is_idempotent_and_recoverable(tmp_path) -> N
                 provider_kind=ProviderKind.OLLAMA,
                 model="qwen",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),
@@ -293,6 +295,7 @@ def test_task_waits_for_model_and_retries_with_same_snapshot(tmp_path) -> None:
             provider_kind=ProviderKind.OLLAMA,
             model="qwen",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -337,6 +340,7 @@ def test_blocked_stages_name_the_stuck_stage_and_reason(tmp_path) -> None:
                 provider_kind=ProviderKind.OLLAMA,
                 model="qwen",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),
@@ -433,6 +437,7 @@ def test_task_duration_estimate_ignores_time_the_app_was_closed(tmp_path) -> Non
                 provider_kind=ProviderKind.OLLAMA,
                 model="qwen",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),
@@ -537,6 +542,7 @@ def test_running_task_is_recovered_after_worker_restart(tmp_path) -> None:
                 provider_kind=ProviderKind.OLLAMA,
                 model="qwen",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),
@@ -572,6 +578,7 @@ def test_superseded_profile_cancels_only_claimable_tasks(tmp_path) -> None:
                 provider_kind=ProviderKind.OLLAMA,
                 model="phi",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),
@@ -601,6 +608,7 @@ def test_superseded_profile_cancels_only_claimable_tasks(tmp_path) -> None:
                 provider_kind=ProviderKind.OLLAMA,
                 model="qwen",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),
@@ -630,6 +638,7 @@ def test_superseded_ingest_policy_cancels_only_old_conversation_tasks(tmp_path)
                 provider_kind=ProviderKind.OLLAMA,
                 model="qwen",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),
@@ -676,6 +685,7 @@ def test_task_rejects_inactive_profile_and_request_id_reuse(tmp_path) -> None:
             provider_kind=ProviderKind.OLLAMA,
             model="qwen",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -1700,6 +1710,7 @@ def _profile_and_tasks(tmp_path) -> tuple[MemoryTaskStore, ModelProfile]:
                 provider_kind=ProviderKind.OLLAMA,
                 model="qwen",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),
@@ -2058,6 +2069,7 @@ def test_a_failed_task_below_the_ceiling_is_still_claimable(tmp_path) -> None:
                 provider_kind=ProviderKind.OLLAMA,
                 model="qwen",
                 embedding_model="nomic-embed-text",
+                embedding_base_url="http://127.0.0.1:11434",
             ),
             secrets,
         ),

memhub/apps/core/tests/test_memory_visibility.py

@@ -55,6 +55,7 @@ def library(tmp_path):
             provider_kind=ProviderKind.OLLAMA,
             model="hy-bot-phi4",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -611,3 +612,58 @@ def test_granting_a_source_repeatedly_does_not_raise(library):
 
     with main_database_read(database) as db:
         assert memory.memory_id in pending(db)
+
+
+def test_grant_writes_still_work_after_the_cutover_drops_the_link_table(library):
+    """#233: the grant triggers read `governed_memory_evidence`.
+
+    The SDK cutover drops it, and SQLite resolves a trigger body only when it
+    fires, so every grant, revoke and connect then failed with "no such
+    table". A trigger whose body names a missing table is not installed, and
+    the copy an earlier startup installed is dropped.
+    """
+    database, *_ = library
+    from one_memory.imports.candidates import RootCandidateKind
+    from one_memory.imports.grants import SourceGrantStore
+    from one_memory.memory_visibility import drop_stale_triggers
+
+    startup(database)
+    with sqlite3.connect(database) as db:
+        db.execute("DROP TABLE governed_memory_evidence")
+
+    def grant_triggers() -> set[str]:
+        with sqlite3.connect(database) as db:
+            return {
+                str(row[0])
+                for row in db.execute(
+                    "SELECT name FROM sqlite_master "
+                    "WHERE type='trigger' AND tbl_name='source_grants'"
+                )
+            }
+
+    assert grant_triggers() == {
+        "memory_visibility_dirty_on_grants",
+        "memory_visibility_dirty_on_grants_update",
+    }
+    # The `evidence` triggers read the dropped table too; the cutover drops
+    # `evidence` as well, which takes them with it.
+    assert {
+        "memory_visibility_dirty_on_grants",
+        "memory_visibility_dirty_on_grants_update",
+    } <= set(drop_stale_triggers(database))
+    assert grant_triggers() == set()
+    assert drop_stale_triggers(database) == ()
+    # Installing again on this shape skips them instead of bringing them back.
+    startup(database)
+    assert grant_triggers() == set()
+
+    folder = database.parent / "cursor"
+    folder.mkdir(exist_ok=True)
+    store = SourceGrantStore(database, home=database.parent)
+    store.grant(
+        user_id="local",
+        source_id="cursor",
+        root=folder,
+        candidate_kind=RootCandidateKind.DEFAULT,
+    )
+    assert store.revoke(user_id="local", source_id="cursor").revoked_at is not None

memhub/apps/core/tests/test_memory_visibility_equivalence.py

@@ -51,6 +51,7 @@ def library(tmp_path):
             provider_kind=ProviderKind.OLLAMA,
             model="hy-bot-phi4",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )

memhub/apps/core/tests/test_model_probe.py

@@ -28,6 +28,7 @@ def test_probe_still_measures_embedding_when_structured_output_fails(tmp_path) -
             base_url="http://127.0.0.1:11434",
             model="plain-model",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -63,6 +64,7 @@ def test_probe_records_the_real_embedding_width_on_the_profile(tmp_path) -> None
             base_url="http://127.0.0.1:11434",
             model="qwen3:8b",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -119,6 +121,7 @@ def test_openai_compatible_probe_checks_model_and_structured_output(tmp_path) ->
             base_url="https://models.example.com/v1",
             model="memory-model",
             embedding_model="embedding-model",
+            embedding_base_url="https://models.example.com/v1",
             api_key=SecretStr(API_KEY),
         ),
         secrets,
@@ -260,6 +263,7 @@ def test_probe_reports_missing_model_without_sending_generation(tmp_path) -> Non
             provider_kind=ProviderKind.OLLAMA,
             model="memory-model",
             embedding_model="embedding-model",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )

memhub/apps/core/tests/test_model_profile_api.py

@@ -50,6 +50,7 @@ def test_model_profile_api_never_returns_key_and_enforces_revision(tmp_path) ->
                     "base_url": "https://models.example.com/v1",
                     "model": "memory-model",
                     "embedding_model": "embedding-model",
+                    "embedding_base_url": "https://models.example.com/v1",
                     "api_key": API_KEY,
                 },
             )
@@ -134,6 +135,7 @@ def test_model_profile_probe_route_returns_only_machine_results(tmp_path) -> Non
                     "base_url": "https://models.example.com/v1",
                     "model": "memory-model",
                     "embedding_model": "embedding-model",
+                    "embedding_base_url": "https://models.example.com/v1",
                     "api_key": API_KEY,
                 },
             )
@@ -244,6 +246,7 @@ def test_activation_survives_a_corpus_already_imported(tmp_path) -> None:
                     "base_url": "https://models.example.com/v1",
                     "model": "memory-model",
                     "embedding_model": "embedding-model",
+                    "embedding_base_url": "https://models.example.com/v1",
                     "api_key": API_KEY,
                 },
             )

memhub/apps/core/tests/test_model_profiles.py

@@ -198,6 +198,7 @@ def test_active_profile_can_be_deleted_without_changing_other_profiles(tmp_path)
             provider_kind=ProviderKind.OLLAMA,
             model="other",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -208,6 +209,7 @@ def test_active_profile_can_be_deleted_without_changing_other_profiles(tmp_path)
             base_url="https://models.example.com/v1",
             model="qwen",
             embedding_model="nomic-embed-text",
+            embedding_base_url="https://models.example.com/v1",
             api_key=SecretStr(API_KEY),
         ),
         secrets,
@@ -349,6 +351,7 @@ def _probed(store, secrets, *, structured: bool, generates: bool = True):
             provider_kind=ProviderKind.OLLAMA,
             model="qwen",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )
@@ -394,3 +397,102 @@ def test_a_model_that_cannot_generate_still_cannot_be_turned_on(tmp_path) -> Non
 
     with pytest.raises(ModelProfileUnavailableError):
         store.activate(profile.profile_id, secrets)
+
+
+def _space(store: ModelProfileStore, profile_id: str) -> str:
+    """The embedding space's collection base, as the Runtime names it."""
+    from one_memory.hy_sdk_runtime import vector_collection_bases
+
+    identity = (profile_id, store.require(profile_id, InMemorySecretStore()).revision)
+    return vector_collection_bases(store.snapshot(*identity))[0]
+
+
+def test_a_remote_embedding_must_name_its_endpoint(tmp_path) -> None:
+    """The endpoint no longer follows the generation `base_url`."""
+    from one_memory.model_profiles import EmbeddingBaseUrlRequiredError
+
+    store = ModelProfileStore(tmp_path / "one-memory.sqlite3")
+    secrets = InMemorySecretStore()
+    unnamed = ModelProfileInput(
+        name="Remote",
+        provider_kind=ProviderKind.OPENAI_COMPATIBLE,
+        base_url="https://models.example.com/v1",
+        model="qwen",
+        embedding_model="text-embedding-3-small",
+        api_key=SecretStr(API_KEY),
+    )
+    with pytest.raises(EmbeddingBaseUrlRequiredError):
+        store.create(unnamed, secrets)
+    assert store.list(secrets) == []
+
+    named = store.create(
+        unnamed.model_copy(update={"embedding_base_url": "https://models.example.com/v1"}),
+        secrets,
+    )
+    with pytest.raises(EmbeddingBaseUrlRequiredError):
+        store.update(
+            named.profile_id,
+            ModelProfilePatch(
+                expected_revision=named.revision,
+                embedding_provider_kind=ProviderKind.OLLAMA,
+                embedding_base_url="",
+            ),
+            secrets,
+        )
+
+
+def test_stored_profiles_get_the_endpoint_they_resolved_to_and_keep_their_space(
+    tmp_path,
+) -> None:
+    """An older build stored remote embeddings with a NULL endpoint and resolved
+    it to the generation `base_url` at runtime. Opening the store writes that
+    endpoint in once, so the collection, and the memories in it, stay put; a
+    later edit of the generation endpoint no longer moves them.
+    """
+    from one_memory.hy_sdk_runtime import vector_collection_bases
+
+    database = tmp_path / "one-memory.sqlite3"
+    store = ModelProfileStore(database)
+    secrets = InMemorySecretStore()
+    profile = store.create(
+        ModelProfileInput(
+            name="Remote",
+            provider_kind=ProviderKind.OPENAI_COMPATIBLE,
+            base_url="https://gateway.example.com/v1",
+            model="qwen",
+            embedding_model="text-embedding-3-small",
+            embedding_base_url="https://gateway.example.com/v1",
+            embedding_dimensions=1536,
+            api_key=SecretStr(API_KEY),
+        ),
+        secrets,
+    )
+    identity = (profile.profile_id, profile.revision)
+    # The space the older build's Runtime resolved: `embedding_base_url or base_url`.
+    before = vector_collection_bases(store.snapshot(*identity))[0]
+    with sqlite3.connect(database) as db:
+        # The shape an older build stored.
+        db.execute("UPDATE model_profiles SET embedding_base_url=NULL")
+        db.execute("UPDATE model_profile_versions SET embedding_base_url=NULL")
+
+    migrated = ModelProfileStore(database)
+    assert migrated.snapshot(*identity).embedding_base_url == "https://gateway.example.com/v1"
+    assert _space(migrated, profile.profile_id) == before
+    # Idempotent: opening again changes nothing.
+    assert _space(ModelProfileStore(database), profile.profile_id) == before
+    with sqlite3.connect(database) as db:
+        assert db.execute(
+            "SELECT COUNT(*) FROM model_profile_versions WHERE embedding_base_url IS NULL"
+        ).fetchone() == (0,)
+
+    edited = migrated.update(
+        profile.profile_id,
+        ModelProfilePatch(
+            expected_revision=profile.revision,
+            base_url="https://gateway.example.com/eu/v1",
+        ),
+        secrets,
+    )
+    assert edited.base_url == "https://gateway.example.com/eu/v1"
+    assert edited.embedding_base_url == "https://gateway.example.com/v1"
+    assert _space(migrated, profile.profile_id) == before

memhub/apps/core/tests/test_multi_user_runtime.py

@@ -0,0 +1,858 @@
+"""One Runtime, many users: `user_id` on the memory API, per-user Hermes profiles.
+
+A user is the root that is read (a Hermes profile's own `state.db`), never a
+session or platform field. The App never names a user and keeps `local`;
+every other harness stays on `local`. These tests drive the production app
+against a user-aware stand-in for the public SDK client, whose `get`,
+`update` and `delete` take no user exactly like hy-memory 2.0.9's.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import sqlite3
+from contextlib import asynccontextmanager
+from datetime import UTC, datetime, timedelta
+from pathlib import Path
+from uuid import uuid4
+
+import pytest
+from httpx import ASGITransport, AsyncClient
+from pydantic import SecretStr
+
+from one_memory.api import _sdk_recall_target, create_app
+from one_memory.config import Settings
+from one_memory.imports.access import AuthorizedSourceRoot
+from one_memory.imports.adapters import ScannedEvidence, ScanOptions
+from one_memory.imports.models import (
+    EvidenceInput,
+    EvidenceOrigin,
+    OriginMetadata,
+    WrittenBy,
+    materialize_write_input,
+)
+from one_memory.imports.sources.hermes import HERMES_DESCRIPTOR, create_hermes_adapter
+from one_memory.imports.sources.registry import build_source_catalog
+from one_memory.memory_server.contracts import internal_session_id_for
+from one_memory.memory_server.durable_queue import (
+    DurableWriteDrain,
+    DurableWriteQueue,
+    execute_job,
+)
+from one_memory.sdk_write_processor import SdkWriteProcessor
+from one_memory.secrets import InMemorySecretStore
+from one_memory.user_scope import lane_key, split_lane, valid_user_id
+
+TOKEN = "multi-user-test"
+
+
+class UserAwareSdk:
+    """The public client surface MemHub calls, holding memories per user."""
+
+    def __init__(self) -> None:
+        self.memories: dict[str, dict[str, object]] = {}
+        self.add_calls: list[dict[str, object]] = []
+
+    def _store(self, user_id: str, content: str, *, source: str, session_id: str) -> str:
+        memory_id = f"memory-{len(self.memories) + 1}"
+        self.memories[memory_id] = {
+            "memory_id": memory_id,
+            "content": content,
+            "layer": "l2_fact",
+            "status": "active",
+            "user_id": user_id,
+            "agent_id": "all-apps",
+            "session_id": session_id,
+            "source": source,
+            "tags": [],
+            "gmt_created": 1_790_000_000,
+        }
+        return memory_id
+
+    def add(self, messages, *, user_id, session_id, source, request_id, **_kwargs):
+        self.add_calls.append({"user_id": user_id, "request_id": request_id})
+        text = " ".join(item["content"] for item in messages if item["role"] == "user")
+        memory_id = self._store(user_id, text, source=source, session_id=session_id)
+        return {"success": True, "memory_ids": [memory_id]}
+
+    def add_extracted(self, contents, *, user_id, session_id, source, request_id, **_kwargs):
+        self.add_calls.append({"user_id": user_id, "request_id": request_id})
+        memory_id = self._store(user_id, " ".join(contents), source=source, session_id=session_id)
+        return {"success": True, "memory_ids": [memory_id]}
+
+    def _rows(self, user_ids) -> list[dict[str, object]]:
+        return [dict(row) for row in self.memories.values() if row["user_id"] in user_ids]
+
+    def list_memories(self, *, user_id, scene=None, offset=0, limit=100, **_kwargs):
+        rows = [] if _kwargs.get("layers") == ["l1_raw"] else self._rows({user_id})
+        page = rows[offset : offset + limit]
+        if scene == "all":
+            return {
+                "scene": "all",
+                "memories": [{**row, "scene": "chat", "kind": "chat"} for row in page],
+                "total": len(rows),
+                "totals": {"chat": len(rows), "working": 0, "coding": 0},
+                "errors": {},
+                "limit": limit,
+                "offset": offset,
+                "request_id": "list",
+                "elapsed_ms": 1.0,
+            }
+        return {"vdb": {"memories": page, "total": len(rows), "limit": limit, "offset": offset}}
+
+    def memory_stats(self, *, user_id, statuses=None):
+        return {"total": len(self._rows({user_id})), "request_id": "stats", "elapsed_ms": 1.0}
+
+    def search(self, query, *, user_ids, scene=None, **kwargs):
+        rows = [] if kwargs.get("layers") == ["l1_raw"] else self._rows(set(user_ids))
+        items = [{**row, "scene": "chat", "kind": "chat", "score": 0.9} for row in rows]
+        if scene == "all":
+            return {
+                "request_id": "search",
+                "scene": "all",
+                "memories": items,
+                "lanes": {"chat": len(items), "working": 0, "coding": 0},
+                "errors": {},
+                "elapsed_ms": 1.0,
+            }
+        return {"request_id": "search", "scene": scene or "normal", "memories": items}
+
+    def get(self, memory_id, kind=None, variant=None):
+        row = self.memories.get(memory_id)
+        return None if row is None else {**row, "kind": "chat", "variant": None}
+
+    def update(self, memory_id, content, kind=None, variant=None, task=None):
+        if memory_id not in self.memories:
+            return {"success": False, "error_message": f"memory_id '{memory_id}' not found"}
+        self.memories[memory_id]["content"] = content
+        return {"success": True, "memory_id": memory_id}
+
+    def delete(self, memory_id, kind=None, variant=None):
+        if self.memories.pop(memory_id, None) is None:
+            return {"success": False, "deleted_count": 0}
+        return {"success": True, "deleted_count": 1}
+
+    def get_user_profile(self, *, user_id):
+        return {
+            "request_id": f"profile-{user_id}",
+            "profile": {
+                "memory_id": f"l0-{user_id}",
+                "content": f"name is {user_id}",
+                "layer": "L0_BASIC_INFO",
+                "basic_info": {"name": user_id},
+            },
+            "elapsed_ms": 1.0,
+        }
+
+
+class Handle:
+    def __init__(self, sdk: UserAwareSdk) -> None:
+        self.sdk = sdk
+
+    async def client(self) -> UserAwareSdk:
+        return self.sdk
+
+    async def close(self) -> None:
+        return None
+
+
+class Coordinator:
+    """The production write processor over the stand-in client."""
+
+    def __init__(self, sdk: UserAwareSdk) -> None:
+        self.processor = SdkWriteProcessor(Handle(sdk))  # type: ignore[arg-type]
+
+    async def process_sdk_job(self, records, *, request_id, before_hy_invocation):
+        return await self.processor.process_sdk_job(
+            records, request_id=request_id, before_hy_invocation=before_hy_invocation
+        )
+
+
+async def _never(_job_id: str) -> None:
+    return None
+
+
+async def drain_all(queue: DurableWriteQueue, sdk: UserAwareSdk) -> None:
+    """Claim and run every pending job, ungated, the way the drain would."""
+    drain = DurableWriteDrain(queue, execute=_never, lease=timedelta(seconds=30))
+    while claimed := drain.claim(limit=8):
+        for job_id in claimed:
+            await execute_job(
+                job_id,
+                queue=queue,
+                supervisor=Coordinator(sdk),
+                evidence=None,  # type: ignore[arg-type]
+            )
+
+
+def application(tmp_path: Path, *, home: Path | None = None):
+    """The production app; with `home`, imports run over that home's sources."""
+    settings = Settings(
+        data_dir=tmp_path / "data",
+        source_home=home or tmp_path,
+        api_token=SecretStr(TOKEN),
+        imports_enabled=home is not None,
+        automatic_memory_processing=False,
+        live_sync_enabled=False,
+    )
+    return create_app(
+        settings,
+        secret_store=InMemorySecretStore(),
+        source_catalog=(
+            build_source_catalog(home=home, environment={}, platform="darwin")
+            if home is not None
+            else None
+        ),
+    )
+
+
+def client_for(app) -> AsyncClient:
+    return AsyncClient(
+        transport=ASGITransport(app=app),
+        base_url="http://test",
+        headers={"Authorization": f"Bearer {TOKEN}"},
+    )
+
+
+def run_with_sdk(tmp_path, monkeypatch, scenario) -> None:
+    async def main() -> None:
+        app = application(tmp_path)
+        sdk = UserAwareSdk()
+
+        @asynccontextmanager
+        async def write_client():
+            yield sdk
+
+        async with app.router.lifespan_context(app):
+            app.state.sdk_client_handle = Handle(sdk)
+            monkeypatch.setattr(app.state.sdk_write_coordinator, "use_client", write_client)
+            async with client_for(app) as client:
+                await scenario(app, sdk, client)
+
+    asyncio.run(main())
+
+
+def _messages(text: str) -> list[dict[str, str]]:
+    return [{"role": "user", "content": text}, {"role": "assistant", "content": "noted"}]
+
+
+async def _write_both_users(app, sdk, client) -> dict[str, str]:
+    """`local` and `wife` each add one memory and submit one conversation."""
+    receipts = {}
+    for user, word in (("local", "tea"), ("wife", "coffee")):
+        request_id = str(uuid4())
+        body = {"request_id": request_id, "text": f"{user} drinks {word}", "scope": "all-apps"}
+        if user != "local":
+            body["user_id"] = user
+        added = await client.post("/api/v2/memories", json=body)
+        assert added.status_code == 201, added.text
+        receipts[f"add:{user}"] = request_id
+        write = {
+            "request_id": "same-request-id",
+            "session_id": "same-session",
+            "source_id": "one-memory-manual",
+            "messages": _messages(f"{user} walks the {word} route"),
+        }
+        if user != "local":
+            write["user_id"] = user
+        written = await client.post("/api/v2/memory-writes", json=write)
+        assert written.status_code == 202, written.text
+        assert written.json()["request_id"] == "same-request-id"
+    await drain_all(app.state.memory_write_queue, sdk)
+    return receipts
+
+
+def _contents(page: dict) -> set[str]:
+    return {memory["content"] for memory in page["memories"]}
+
+
+@pytest.mark.fresh_sdk
+def test_each_user_reads_only_their_own_library_search_profile_and_graph(
+    tmp_path, monkeypatch
+) -> None:
+    async def scenario(app, sdk, client):
+        await _write_both_users(app, sdk, client)
+        assert {call["user_id"] for call in sdk.add_calls} == {"local", "wife"}
+        # The same caller request id from two users is two SDK requests.
+        assert len({call["request_id"] for call in sdk.add_calls}) == len(sdk.add_calls) == 4
+
+        local = (await client.get("/api/v2/memory-library")).json()
+        wife = (await client.get("/api/v2/memory-library", params={"user_id": "wife"})).json()
+        assert _contents(local) == {"local drinks tea", "local walks the tea route"}
+        assert _contents(wife) == {"wife drinks coffee", "wife walks the coffee route"}
+        nobody = await client.get("/api/v2/memory-library", params={"user_id": "__nobody__"})
+        assert nobody.status_code == 200 and nobody.json()["memories"] == []
+
+        searched = (
+            await client.get("/api/v2/memory-library", params={"query": "route", "user_id": "wife"})
+        ).json()
+        assert _contents(searched) == {"wife walks the coffee route"}
+
+        context = await client.post(
+            "/api/v2/context/search",
+            json={"query": "drinks", "kinds": ["memory"], "user_id": "wife"},
+        )
+        assert context.status_code == 200, context.text
+        assert {item["content"] for item in context.json()["items"]} == {
+            "wife drinks coffee",
+            "wife walks the coffee route",
+        }
+        local_context = await client.post(
+            "/api/v2/context/search", json={"query": "drinks", "kinds": ["memory"]}
+        )
+        assert all("wife" not in item["content"] for item in local_context.json()["items"])
+
+        profile = (await client.get("/api/v2/user-profile", params={"user_id": "wife"})).json()
+        assert profile["profile"]["basic_info"] == {"name": "wife"}
+        assert (await client.get("/api/v2/user-profile")).json()["profile"]["basic_info"] == {
+            "name": "local"
+        }
+
+        graph = (await client.get("/api/v2/memories/graph", params={"user_id": "wife"})).json()
+        labels = {node["label"] for node in graph["nodes"] if node["node_type"] == "memory"}
+        assert labels == {"wife drinks coffee", "wife walks the coffee route"}
+
+        summary = (await client.get("/api/v2/home-summary", params={"user_id": "wife"})).json()
+        assert summary["memory_count"] == 2
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_another_users_memory_id_answers_404_for_read_update_and_delete(
+    tmp_path, monkeypatch
+) -> None:
+    async def scenario(app, sdk, client):
+        await _write_both_users(app, sdk, client)
+        [wife_memory] = [
+            memory_id
+            for memory_id, row in sdk.memories.items()
+            if row["user_id"] == "wife" and row["content"] == "wife drinks coffee"
+        ]
+        path = f"/api/v2/memories/{wife_memory}"
+
+        assert (await client.get(path)).status_code == 404
+        assert (await client.put(path, json={"content": "stolen"})).status_code == 404
+        assert (await client.delete(path)).status_code == 204
+        assert sdk.memories[wife_memory]["content"] == "wife drinks coffee"
+
+        own = await client.get(path, params={"user_id": "wife"})
+        assert own.status_code == 200 and own.json()["user_id"] == "wife"
+        updated = await client.put(path, json={"content": "wife drinks tea", "user_id": "wife"})
+        assert updated.status_code == 200
+        assert (await client.delete(path, params={"user_id": "wife"})).status_code == 204
+        assert wife_memory not in sdk.memories
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_receipts_belong_to_the_user_that_submitted_them(tmp_path, monkeypatch) -> None:
+    async def scenario(app, sdk, client):
+        receipts = await _write_both_users(app, sdk, client)
+        wife_add = f"/api/v2/memories/additions/{receipts['add:wife']}"
+        assert (await client.get(wife_add)).status_code == 404
+        own = await client.get(wife_add, params={"user_id": "wife"})
+        assert own.status_code == 200 and own.json()["state"] == "ready"
+        local_add = f"/api/v2/memories/additions/{receipts['add:local']}"
+        assert (await client.get(local_add, params={"user_id": "wife"})).status_code == 404
+
+        # Both users used `same-request-id`; each reads only their own receipt.
+        local_write = await client.get("/api/v2/memory-writes/same-request-id")
+        wife_write = await client.get(
+            "/api/v2/memory-writes/same-request-id", params={"user_id": "wife"}
+        )
+        assert local_write.status_code == wife_write.status_code == 200
+        assert local_write.json()["memory_ids"] != wife_write.json()["memory_ids"]
+        assert (
+            await client.get("/api/v2/memory-writes/same-request-id", params={"user_id": "kid"})
+        ).status_code == 404
+        # `local` cannot name another user's namespaced queue key directly.
+        with sqlite3.connect(app.state.settings.data_dir / "one-memory.sqlite3") as db:
+            keys = [row[0] for row in db.execute("SELECT idempotency_key FROM memory_write_jobs")]
+        wife_key = next(key for key in keys if key.startswith("user:wife:"))
+        assert (await client.get(f"/api/v2/memory-writes/{wife_key}")).status_code == 404
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_omitting_user_id_keeps_local_and_recall_and_ask_pass_it_through(
+    tmp_path, monkeypatch
+) -> None:
+    async def scenario(app, sdk, client):
+        await _write_both_users(app, sdk, client)
+        seen: list[str] = []
+        original = sdk.list_memories
+
+        def recording(**kwargs):
+            seen.append(kwargs["user_id"])
+            return original(**kwargs)
+
+        sdk.list_memories = recording  # type: ignore[method-assign]
+        # Built at startup over the real handle; rebuilt here over the stand-in.
+        app.state.recall_target = _sdk_recall_target(app, "local")
+        await client.get("/api/v2/memory-library")
+        await client.get("/api/v2/raw")
+        await client.get("/api/v2/home-summary")
+        assert set(seen) == {"local"}
+
+        monkeypatch.setattr(
+            "one_memory.api._recall_model_ready", lambda _app: _true(), raising=True
+        )
+        recalled = await client.post("/api/v1/recall", json={"query": "drinks", "user_id": "wife"})
+        assert recalled.status_code == 200, recalled.text
+        assert {hit["content"] for hit in recalled.json()["hits"]} == {
+            "wife drinks coffee",
+            "wife walks the coffee route",
+        }
+        local = await client.post("/api/v1/recall", json={"query": "drinks"})
+        assert all("wife" not in hit["content"] for hit in local.json()["hits"])
+
+        answered = await client.post(
+            "/api/v2/memories/ask", json={"question": "what does she drink", "user_id": "wife"}
+        )
+        assert answered.status_code == 200, answered.text
+        assert {memory["content"] for memory in answered.json()["memories"]} <= {
+            "wife drinks coffee",
+            "wife walks the coffee route",
+        }
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+async def _true() -> bool:
+    return True
+
+
+@pytest.mark.fresh_sdk
+@pytest.mark.parametrize(
+    ("method", "path", "kwargs"),
+    [
+        ("get", "/api/v2/memory-library", {"params": {"nobody": "1"}}),
+        ("get", "/api/v2/home-summary", {"params": {"userid": "wife"}}),
+        ("get", "/api/v2/user-profile", {"params": {"user": "wife"}}),
+        ("get", "/api/v2/raw", {"params": {"scope": "x"}}),
+        ("get", "/api/v2/memories/graph", {"params": {"scene": "all"}}),
+        ("get", "/api/v2/memories/some-id", {"params": {"owner": "wife"}}),
+        ("delete", "/api/v2/memories/some-id", {"params": {"owner": "wife"}}),
+        ("get", "/api/v2/memory-writes/some-id", {"params": {"owner": "wife"}}),
+        ("get", f"/api/v2/memories/additions/{uuid4()}", {"params": {"owner": "wife"}}),
+        ("post", "/api/v2/memories/ask", {"json": {"question": "q", "user": "wife"}}),
+        ("post", "/api/v1/recall", {"json": {"query": "q", "user": "wife"}}),
+        ("post", "/api/v1/recall", {"json": {"query": "q"}, "params": {"user_id": "wife"}}),
+        ("post", "/api/v2/context/search", {"json": {"query": "q", "owner": "wife"}}),
+        (
+            "post",
+            "/api/v2/memories",
+            {"json": {"request_id": str(uuid4()), "text": "t", "scope": "all-apps", "u": 1}},
+        ),
+        ("put", "/api/v2/memories/some-id", {"json": {"content": "c", "owner": "wife"}}),
+    ],
+)
+def test_unknown_query_and_body_parameters_answer_422(tmp_path, method, path, kwargs) -> None:
+    async def scenario():
+        app = application(tmp_path)
+        async with app.router.lifespan_context(app), client_for(app) as client:
+            response = await client.request(method.upper(), path, **kwargs)
+            assert response.status_code == 422, response.text
+            assert response.json()["detail"] == "request_validation_failed"
+
+    asyncio.run(scenario())
+
+
+@pytest.mark.fresh_sdk
+@pytest.mark.parametrize("bad", ["a:b", "", "x" * 65, "white space", "ü", "a/b"])
+def test_a_malformed_user_id_answers_422(tmp_path, bad) -> None:
+    async def scenario():
+        app = application(tmp_path)
+        async with app.router.lifespan_context(app), client_for(app) as client:
+            responses = [
+                await client.get("/api/v2/memory-library", params={"user_id": bad}),
+                await client.get("/api/v2/user-profile", params={"user_id": bad}),
+                await client.post(
+                    "/api/v2/memories",
+                    json={
+                        "request_id": str(uuid4()),
+                        "text": "t",
+                        "scope": "all-apps",
+                        "user_id": bad,
+                    },
+                ),
+                await client.post(
+                    "/api/v2/memory-writes",
+                    json={
+                        "request_id": "r",
+                        "session_id": "s",
+                        "messages": _messages("hello"),
+                        "user_id": bad,
+                    },
+                ),
+                await client.post("/api/v1/recall", json={"query": "q", "user_id": bad}),
+            ]
+            assert [response.status_code for response in responses] == [422] * len(responses)
+
+    asyncio.run(scenario())
+
+
+def test_user_ids_and_lanes() -> None:
+    assert valid_user_id("wife") and valid_user_id("A.b_c-9")
+    assert not valid_user_id("a:b") and not valid_user_id("wife\n")
+    assert lane_key("local", "hermes") == "hermes"
+    assert lane_key("wife", "hermes") == "hermes@wife"
+    assert split_lane("hermes@wife") == ("wife", "hermes")
+    assert split_lane("hermes") == ("local", "hermes")
+    # `local` keeps the session identity it always had; another user's never equals it.
+    assert internal_session_id_for("s-1", "hermes") == "s-1"
+    assert internal_session_id_for("s-1", "hermes", "wife") != "s-1"
+    assert internal_session_id_for("s-1", "hermes", "wife") != internal_session_id_for(
+        "s-1", "hermes", "kid"
+    )
+
+
+def _hermes_state_db(root: Path, rows: list[tuple[int, str, str, str]]) -> None:
+    root.mkdir(parents=True, exist_ok=True)
+    connection = sqlite3.connect(root / "state.db")
+    connection.execute(
+        "CREATE TABLE messages (id INTEGER PRIMARY KEY, session_id TEXT, role TEXT,"
+        " content TEXT, timestamp REAL, tool_name TEXT, active INTEGER)"
+    )
+    connection.executemany(
+        "INSERT INTO messages VALUES (?, ?, ?, ?, 1790000000, NULL, 1)",
+        rows,
+    )
+    connection.commit()
+    connection.close()
+
+
+def _hermes_homes(home: Path) -> tuple[Path, Path]:
+    hermes = home / ".hermes"
+    wife = hermes / "profiles" / "wife"
+    # The same Hermes session id in both profiles: identities must not merge.
+    _hermes_state_db(
+        hermes,
+        [(1, "session-1", "user", "I am the husband"), (2, "session-1", "assistant", "hi")],
+    )
+    _hermes_state_db(
+        wife,
+        [(1, "session-1", "user", "I am the wife"), (2, "session-1", "assistant", "hello")],
+    )
+    return hermes, wife
+
+
+def _scan_texts(root: Path, home: Path) -> list[str]:
+    adapter = create_hermes_adapter(
+        AuthorizedSourceRoot.grant(
+            source_id=HERMES_DESCRIPTOR.source_id, root=root, user_confirmed=True, home=home
+        )
+    )
+
+    async def collect() -> list[str]:
+        return [
+            event.evidence.verbatim
+            async for event in adapter.scan(ScanOptions(cursors={}))
+            if isinstance(event, ScannedEvidence)
+        ]
+
+    return asyncio.run(collect())
+
+
+def test_a_hermes_root_scan_never_reads_its_profiles(tmp_path) -> None:
+    hermes, wife = _hermes_homes(tmp_path)
+    (hermes / "profiles" / "wife" / "sessions").mkdir()
+    (hermes / "profiles" / "wife" / "sessions" / "s.jsonl").write_text(
+        json.dumps({"type": "message", "role": "user", "content": "wife's transcript"}) + "\n"
+    )
+
+    assert _scan_texts(hermes, tmp_path) == ["I am the husband", "hi"]
+    assert _scan_texts(wife, tmp_path) == ["I am the wife", "hello", "wife's transcript"]
+
+
+def _jobs(database: Path) -> list[sqlite3.Row]:
+    with sqlite3.connect(database) as db:
+        db.row_factory = sqlite3.Row
+        return db.execute(
+            "SELECT session_key,source_key,idempotency_key,state,payload_json "
+            "FROM memory_write_jobs ORDER BY queue_seq"
+        ).fetchall()
+
+
+@pytest.mark.fresh_sdk
+def test_two_hermes_grants_each_read_their_own_root_into_separate_batches(tmp_path) -> None:
+    async def scenario():
+        home = tmp_path / "home"
+        _hermes, wife = _hermes_homes(home)
+        app = application(tmp_path, home=home)
+        async with app.router.lifespan_context(app), client_for(app) as client:
+            location = (await client.post("/api/v1/sources/hermes/location")).json()
+            local = await client.post(
+                "/api/v1/sources/connect",
+                json={
+                    "source_ids": ["hermes"],
+                    "request_id": str(uuid4()),
+                    "source_roots": {"hermes": location["root_id"]},
+                    "enable_sync": True,
+                },
+            )
+            assert local.status_code == 200, local.text
+            profile = await client.post(
+                "/api/v1/sources/connect",
+                json={
+                    "source_ids": ["hermes"],
+                    "request_id": str(uuid4()),
+                    "user_id": "wife",
+                    "root": str(wife),
+                },
+            )
+            assert profile.status_code == 200, profile.text
+            [wife_job] = profile.json()["jobs"]
+            assert wife_job["user_id"] == "wife"
+            jobs = [*local.json()["jobs"], wife_job]
+            for job in jobs:
+                finished = await app.state.import_supervisor.wait(job["job_id"])
+                assert finished.status == "completed", finished
+
+            rows = [row for row in _jobs(app.state.settings.data_dir / "one-memory.sqlite3")]
+            assert rows, "history import wrote no batches"
+            by_user: dict[str, list[str]] = {}
+            for row in rows:
+                records = json.loads(row["payload_json"])["input_records"]
+                users = {record["user_id"] for record in records}
+                assert len(users) == 1, "a batch mixed users"
+                [user] = users
+                by_user.setdefault(user, []).extend(record["verbatim"] for record in records)
+                assert row["source_key"] == lane_key(user, "hermes")
+                assert row["session_key"] == internal_session_id_for("session-1", "hermes", user)
+            assert sorted(by_user["local"]) == ["I am the husband", "hi"]
+            assert sorted(by_user["wife"]) == ["I am the wife", "hello"]
+
+            grants = (await client.get("/api/v1/source-grants")).json()["grants"]
+            assert {(grant["user_id"], grant["source_id"]) for grant in grants} >= {
+                ("local", "hermes"),
+                ("wife", "hermes"),
+            }
+            # The App's view stays `local`'s.
+            listed = (await client.get("/api/v1/import-jobs")).json()
+            assert {job["user_id"] for job in listed} == {"local"}
+            wife_jobs = (await client.get("/api/v1/import-jobs", params={"user_id": "wife"})).json()
+            assert {job["user_id"] for job in wife_jobs} == {"wife"}
+            assert all(
+                item["source_id"] != "hermes@wife"
+                for item in (await client.get("/api/v1/live-sync")).json()
+            )
+
+            revoked = await client.post("/api/v1/sources/hermes/revoke", params={"user_id": "wife"})
+            assert revoked.status_code == 200 and revoked.json()["revoked_at"] is not None
+            local_grant = app.state.source_grant_store.receipt(user_id="local", source_id="hermes")
+            assert local_grant.revoked_at is None
+
+    asyncio.run(scenario())
+
+
+@pytest.mark.fresh_sdk
+def test_a_migrated_home_with_the_old_grant_triggers_can_connect_and_revoke(tmp_path) -> None:
+    """#233: a home an earlier build migrated kept two `source_grants` triggers
+    whose body reads the dropped `governed_memory_evidence`. Every grant write
+    then failed, so connecting and revoking answered 503. Startup drops them.
+    """
+    stale = """
+        CREATE TRIGGER memory_visibility_dirty_on_grants{suffix}
+        AFTER {event} ON source_grants
+        BEGIN
+            INSERT INTO memory_visibility_dirty (memory_id)
+            SELECT DISTINCT link.memory_id
+            FROM governed_memory_evidence link
+            JOIN evidence e ON e.evidence_id = link.evidence_id
+            WHERE e.source_id = NEW.source_id
+            ON CONFLICT(memory_id) DO NOTHING;
+        END;"""
+
+    async def scenario():
+        home = tmp_path / "home"
+        _hermes, wife = _hermes_homes(home)
+        app = application(tmp_path, home=home)
+        async with app.router.lifespan_context(app):
+            database = app.state.settings.data_dir / "one-memory.sqlite3"
+        with sqlite3.connect(database) as db:
+            db.execute("CREATE TABLE IF NOT EXISTS memory_visibility_dirty (memory_id TEXT)")
+            db.execute(stale.format(suffix="", event="INSERT"))
+            db.execute(stale.format(suffix="_update", event="UPDATE"))
+
+        app = application(tmp_path, home=home)
+        async with app.router.lifespan_context(app), client_for(app) as client:
+            location = (await client.post("/api/v1/sources/hermes/location")).json()
+            local = await client.post(
+                "/api/v1/sources/connect",
+                json={
+                    "source_ids": ["hermes"],
+                    "request_id": str(uuid4()),
+                    "source_roots": {"hermes": location["root_id"]},
+                },
+            )
+            assert local.status_code == 200, local.text
+            profile = await client.post(
+                "/api/v1/sources/connect",
+                json={
+                    "source_ids": ["hermes"],
+                    "request_id": str(uuid4()),
+                    "user_id": "wife",
+                    "root": str(wife),
+                },
+            )
+            assert profile.status_code == 200, profile.text
+            for user_id in ("wife", "local"):
+                revoked = await client.post(
+                    "/api/v1/sources/hermes/revoke", params={"user_id": user_id}
+                )
+                assert revoked.status_code == 200, revoked.text
+                assert revoked.json()["revoked_at"] is not None
+        with sqlite3.connect(database) as db:
+            assert (
+                db.execute(
+                    "SELECT COUNT(*) FROM sqlite_master "
+                    "WHERE type='trigger' AND tbl_name='source_grants'"
+                ).fetchone()[0]
+                == 0
+            )
+
+    asyncio.run(scenario())
+
+
+@pytest.mark.fresh_sdk
+def test_only_hermes_takes_another_user_or_a_custom_root(tmp_path) -> None:
+    async def scenario():
+        home = tmp_path / "home"
+        (home / ".claude" / "projects").mkdir(parents=True)
+        app = application(tmp_path, home=home)
+        async with app.router.lifespan_context(app), client_for(app) as client:
+            for extra in ({"user_id": "wife"}, {"root": str(home / ".claude" / "projects")}):
+                response = await client.post(
+                    "/api/v1/sources/connect",
+                    json={
+                        "source_ids": ["claude-code"],
+                        "request_id": str(uuid4()),
+                        "source_roots": {},
+                        **extra,
+                    },
+                )
+                assert response.status_code == 422, response.text
+                assert response.json()["detail"] == "user_id_requires_multi_user_source"
+            revoke = await client.post(
+                "/api/v1/sources/claude-code/revoke", params={"user_id": "wife"}
+            )
+            assert revoke.status_code == 422
+            # Hermes too: another user must name their own root.
+            rootless = await client.post(
+                "/api/v1/sources/connect",
+                json={
+                    "source_ids": ["hermes"],
+                    "request_id": str(uuid4()),
+                    "source_roots": {},
+                    "user_id": "wife",
+                },
+            )
+            assert rootless.status_code == 422
+            assert rootless.json()["detail"] == "source_root_required_for_user"
+
+    asyncio.run(scenario())
+
+
+def _record(user_id: str, message_id: str, role: str, text: str):
+    return materialize_write_input(
+        EvidenceInput(
+            user_id=user_id,
+            source_id="hermes",
+            conversation_id="session-1",
+            message_id=message_id,
+            role=role,
+            written_by=WrittenBy.PERSON if role == "user" else WrittenBy.AGENT,
+            created_at=datetime(2026, 10, 1, tzinfo=UTC),
+            verbatim=text,
+            origin_meta={EvidenceOrigin.FILE: OriginMetadata()},
+        )
+    )
+
+
+def test_assembling_rows_and_batches_are_keyed_by_user(tmp_path) -> None:
+    """Two users' `state.db` share a relative file key and a session id.
+
+    Finishing one user's file must neither take nor drop the other user's
+    staged turns, and the batches it writes hold one user each.
+    """
+    path = tmp_path / "one-memory.sqlite3"
+    queue = DurableWriteQueue(path, sdk_authority=True)
+    file_key = "f" * 64
+    from one_memory.database_access import main_database
+
+    with main_database(path).write(row_factory=sqlite3.Row) as connection, connection:
+        for user in ("local", "wife"):
+            for index, role in enumerate(("user", "assistant")):
+                queue.stage_sdk_input(
+                    _record(user, f"m{index}", role, f"{user} {role}"),
+                    file_key=file_key,
+                    priority="normal",
+                    connection=connection,
+                )
+        queue.finalize_sdk_file(
+            user_id="local", source_id="hermes", file_key=file_key, connection=connection
+        )
+
+    rows = _jobs(path)
+    pending = [row for row in rows if row["state"] == "pending"]
+    assembling = [row for row in rows if row["state"] == "assembling"]
+    assert [
+        {record["user_id"] for record in json.loads(row["payload_json"])["input_records"]}
+        for row in pending
+    ] == [{"local"}]
+    assert [split_lane(row["source_key"]) for row in assembling] == [("wife", "hermes")]
+    assert pending[0]["session_key"] != assembling[0]["session_key"]
+
+
+def _observed(response) -> tuple[int, bytes]:
+    return response.status_code, response.content
+
+
+@pytest.mark.fresh_sdk
+def test_another_users_id_answers_exactly_like_an_id_that_does_not_exist(
+    tmp_path, monkeypatch
+) -> None:
+    """Comparing responses must not tell whether another user's id exists."""
+
+    async def scenario(app, sdk, client):
+        receipts = await _write_both_users(app, sdk, client)
+        [wife_memory] = [
+            memory_id
+            for memory_id, row in sdk.memories.items()
+            if row["user_id"] == "wife" and row["content"] == "wife drinks coffee"
+        ]
+        before = {key: dict(row) for key, row in sdk.memories.items()}
+        with sqlite3.connect(app.state.settings.data_dir / "one-memory.sqlite3") as db:
+            keys = [row[0] for row in db.execute("SELECT idempotency_key FROM memory_write_jobs")]
+        wife_key = next(key for key in keys if key.startswith("user:wife:"))
+
+        pairs = [
+            ("GET", f"/api/v2/memories/{wife_memory}", "/api/v2/memories/no-such-memory", {}),
+            (
+                "PUT",
+                f"/api/v2/memories/{wife_memory}",
+                "/api/v2/memories/no-such-memory",
+                {"json": {"content": "stolen"}},
+            ),
+            ("DELETE", f"/api/v2/memories/{wife_memory}", "/api/v2/memories/no-such-memory", {}),
+            (
+                "GET",
+                f"/api/v2/memories/additions/{receipts['add:wife']}",
+                f"/api/v2/memories/additions/{uuid4()}",
+                {},
+            ),
+            ("GET", f"/api/v2/memory-writes/{wife_key}", "/api/v2/memory-writes/no-such-id", {}),
+        ]
+        for method, theirs, missing, kwargs in pairs:
+            other = await client.request(method, theirs, **kwargs)
+            absent = await client.request(method, missing, **kwargs)
+            assert _observed(other) == _observed(absent), (method, theirs)
+            assert other.status_code == (204 if method == "DELETE" else 404)
+        assert sdk.memories == before
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)

memhub/apps/core/tests/test_policy_api_wiring.py

@@ -173,12 +173,14 @@ def test_first_import_gate_waits_for_every_selected_history_job() -> None:
     jobs = [
         SimpleNamespace(
             operation="history",
+            user_id="local",
             source_id="cursor",
             created_at=datetime(2026, 10, 18, tzinfo=UTC),
             status="completed",
         ),
         SimpleNamespace(
             operation="history",
+            user_id="local",
             source_id="codex",
             created_at=datetime(2026, 10, 18, tzinfo=UTC),
             status="running",

memhub/apps/core/tests/test_read_path_freshness_batch.py

@@ -40,6 +40,7 @@ def library(tmp_path):
             provider_kind=ProviderKind.OLLAMA,
             model="hy-bot-phi4",
             embedding_model="nomic-embed-text",
+            embedding_base_url="http://127.0.0.1:11434",
         ),
         secrets,
     )

memhub/apps/core/tests/test_recall_raw_fallback.py

@@ -210,7 +210,7 @@ def test_recall_route_falls_back_and_keeps_the_raw_grant(tmp_path, monkeypatch):
     async def no_model(_app):
         return False
 
-    async def raw(_app, query, *, limit, sources):
+    async def raw(_app, query, *, limit, sources, user_id="local"):
         searched.append(sources)
         return (_raw("r1", "上周决定用 Postgres", source="cursor"),)
 

memhub/apps/core/tests/test_sdk_api_surface.py

@@ -0,0 +1,744 @@
+"""The SDK surface MemHub exposes: search pass-through, layers, raw, v2 aliases.
+
+P2-1 (R8): `context/search`, `recall` and the v2 ask routes pass the public
+`client.search` knobs through, validated, and keep a caller's floor on what
+they return. P2-2 (R9): the library honours `layers`, `statuses`, `scene` and
+the source/session/agent filters, and counts memories per layer. P2-4: raw
+lists the SDK's L1 raw turns, active and shadow. P2-3: v2 aliases of the
+source and import routes, per user. Profile versions read the L0 chain.
+
+Every route stays scoped to its one `user_id`; the stand-in client records
+each call so a test can see exactly what reached the SDK.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import hashlib
+import sqlite3
+from contextlib import asynccontextmanager
+from pathlib import Path
+from uuid import uuid4
+
+import pytest
+from httpx import ASGITransport, AsyncClient
+from pydantic import SecretStr
+
+from one_memory.api import create_app
+from one_memory.config import Settings
+from one_memory.imports.sources.registry import build_source_catalog
+from one_memory.secrets import InMemorySecretStore
+
+TOKEN = "sdk-surface-test"
+
+CHAT_ROWS = {
+    "local": [
+        ("m-high", "local likes green tea", 0.95, "l2_fact", "hermes"),
+        ("m-mid", "local walks at dawn", 0.6, "l2_fact", "codex"),
+    ],
+    "wife": [("w-high", "wife likes coffee", 0.92, "l2_fact", "hermes")],
+}
+RAW_ROWS = {
+    "local": [
+        {
+            "memory_id": "raw-1",
+            "content": "[user]: 我叫费天翔\n[assistant]: 好的，记住了。",
+            "status": "shadow",
+            "session_id": "s-1",
+            "source": "hermes",
+            "memory_at": 1_790_000_000,
+            "gmt_created": 1_790_000_100,
+        },
+        {
+            "memory_id": "raw-2",
+            "content": "[user]: 你好",
+            "status": "active",
+            "session_id": "s-2",
+            "source": "codex",
+            "memory_at": 1_790_100_000,
+            "gmt_created": 1_790_100_100,
+        },
+    ],
+    "wife": [
+        {
+            "memory_id": "raw-w",
+            "content": "[user]: I am the wife",
+            "status": "active",
+            "session_id": "s-w",
+            "source": "hermes",
+            "memory_at": 1_790_000_000,
+            "gmt_created": 1_790_000_000,
+        }
+    ],
+}
+
+
+def _chat(user_id: str, memory_id: str, content: str, layer: str, source: str) -> dict:
+    return {
+        "memory_id": memory_id,
+        "content": content,
+        "layer": layer,
+        "status": "active",
+        "user_id": user_id,
+        "agent_id": "all-apps",
+        "session_id": "s-1",
+        "source": source,
+        "tags": [],
+        "gmt_created": 1_790_000_000,
+    }
+
+
+class RecordingSdk:
+    """The public client surface, recording every call's keyword arguments."""
+
+    def __init__(self) -> None:
+        self.calls: list[tuple[str, dict]] = []
+
+    def _record(self, name: str, kwargs: dict) -> None:
+        self.calls.append((name, dict(kwargs)))
+
+    def last(self, name: str) -> dict:
+        return [kwargs for called, kwargs in self.calls if called == name][-1]
+
+    def all(self, name: str) -> list[dict]:
+        return [kwargs for called, kwargs in self.calls if called == name]
+
+    # The SDK answers rows below a caller's floor in some lanes (a schema
+    # floor, a fused rank score), so Core must filter what it returns.
+    def search(self, query, **kwargs):
+        self._record("search", {"query": query, **kwargs})
+        [user_id] = kwargs["user_ids"]
+        if kwargs.get("layers") == ["l1_raw"]:
+            hits = [
+                {**row, "user_id": user_id, "agent_id": "all-apps", "score": score}
+                for row, score in zip(RAW_ROWS.get(user_id, []), (0.97, 0.3), strict=False)
+            ]
+            return {"request_id": "s", "memories": hits}
+        rows = [
+            {**_chat(user_id, mid, text, layer, source), "score": score}
+            for mid, text, score, layer, source in CHAT_ROWS.get(user_id, [])
+        ]
+        scene = kwargs.get("scene")
+        if scene == "all":
+            fused = [
+                {**row, "scene": "chat", "kind": "chat", "lane_score": row["score"], "score": 0.016}
+                for row in rows
+            ]
+            return {"request_id": "s", "scene": "all", "memories": fused, "lanes": {}, "errors": {}}
+        if scene in {"working", "coding"}:
+            return {
+                "request_id": "s",
+                "scene": "productivity",
+                "variant": scene,
+                "facts": [
+                    {"fact_id": "f-1", "statement": "use uv", "score": 0.91, "user_id": user_id}
+                ],
+                "experiences": [],
+            }
+        return {
+            "request_id": "s",
+            "scene": "normal",
+            "memories": {"normal": rows, "schema": [], "proactive": []},
+        }
+
+    def get(self, memory_id, kind=None, variant=None):
+        self._record("get", {"memory_id": memory_id, "kind": kind, "variant": variant})
+        if kind == "fact":
+            return {
+                "kind": "fact",
+                "variant": variant,
+                "fact_id": memory_id,
+                "statement": "use uv",
+                "status": "active",
+                "user_id": "local",
+            }
+        for user_id, rows in CHAT_ROWS.items():
+            for mid, text, _score, layer, source in rows:
+                if mid == memory_id:
+                    return {**_chat(user_id, mid, text, layer, source), "kind": "chat"}
+        return None
+
+    def list_memories(self, **kwargs):
+        self._record("list_memories", kwargs)
+        user_id = kwargs["user_id"]
+        layers = kwargs.get("layers")
+        if layers == ["l1_raw"]:
+            statuses = set(kwargs.get("statuses") or ["active"])
+            rows = [
+                {**row, "layer": "l1_raw", "user_id": user_id, "agent_id": "all-apps"}
+                for row in RAW_ROWS.get(user_id, [])
+                if row["status"] in statuses
+                and (
+                    not kwargs.get("include_sources") or row["source"] in kwargs["include_sources"]
+                )
+                and (not kwargs.get("session_ids") or row["session_id"] in kwargs["session_ids"])
+            ]
+            return {"vdb": {"memories": rows, "total": len(rows)}}
+        rows = [_chat(user_id, *row[:2], *row[3:]) for row in CHAT_ROWS.get(user_id, [])]
+        if layers:
+            rows = [row for row in rows if row["layer"] in layers]
+            if "l0_basic_info" in layers:
+                rows.append(_chat(user_id, "l0-1", "name is Fei", "l0_basic_info", "hermes"))
+        if kwargs.get("scene") == "all":
+            return {
+                "scene": "all",
+                "memories": [{**row, "scene": "chat", "kind": "chat"} for row in rows],
+                "total": len(rows),
+                "totals": {"chat": len(rows), "working": 0, "coding": 0},
+                "errors": {},
+            }
+        return {"vdb": {"memories": rows, "total": len(rows)}}
+
+    def memory_stats(self, *, user_id, statuses=None):
+        return {"total": 0, "request_id": "stats", "elapsed_ms": 1.0}
+
+    def get_user_basic_info(self, *, user_id):
+        self._record("get_user_basic_info", {"user_id": user_id})
+        return {
+            "request_id": f"basic-{user_id}",
+            "elapsed_ms": 1.5,
+            "memories": [
+                {
+                    "memory_id": f"{user_id}-l0-head",
+                    "content": "The user's name is Fei.",
+                    "layer": "l0_basic_info",
+                    "status": "active",
+                    "memory_at": 1_790_000_100,
+                    "gmt_created": 1_790_000_100,
+                    "tags": [],
+                    "basic_info_kv": {"name": "Fei"},
+                    "evolution_chain": [
+                        {
+                            "node_id": f"{user_id}-l0-head",
+                            "content": "The user's name is Fei.",
+                            "memory_at": "2026-10-31 10:00",
+                            "gmt_created": 1_790_000_100,
+                            "layer": "l0_basic_info",
+                        },
+                        {
+                            "node_id": f"{user_id}-l0-old",
+                            "content": "The user's name is 费天翔.",
+                            "memory_at": "2026-10-30 10:00",
+                            "gmt_created": 1_790_000_000,
+                            "layer": "l0_basic_info",
+                        },
+                    ],
+                }
+            ],
+        }
+
+
+class Handle:
+    def __init__(self, sdk: RecordingSdk) -> None:
+        self.sdk = sdk
+
+    async def client(self) -> RecordingSdk:
+        return self.sdk
+
+    async def close(self) -> None:
+        return None
+
+
+def application(tmp_path: Path, *, home: Path | None = None):
+    settings = Settings(
+        data_dir=tmp_path / "data",
+        source_home=home or tmp_path,
+        api_token=SecretStr(TOKEN),
+        imports_enabled=home is not None,
+        automatic_memory_processing=False,
+        live_sync_enabled=False,
+    )
+    return create_app(
+        settings,
+        secret_store=InMemorySecretStore(),
+        source_catalog=(
+            build_source_catalog(home=home, environment={}, platform="darwin")
+            if home is not None
+            else None
+        ),
+    )
+
+
+def client_for(app) -> AsyncClient:
+    return AsyncClient(
+        transport=ASGITransport(app=app),
+        base_url="http://test",
+        headers={"Authorization": f"Bearer {TOKEN}"},
+    )
+
+
+async def _model_ready(_app) -> bool:
+    return True
+
+
+def run_with_sdk(tmp_path, monkeypatch, scenario) -> None:
+    async def main() -> None:
+        app = application(tmp_path)
+        sdk = RecordingSdk()
+
+        @asynccontextmanager
+        async def write_client():
+            yield sdk
+
+        async with app.router.lifespan_context(app):
+            app.state.sdk_client_handle = Handle(sdk)
+            monkeypatch.setattr(app.state.sdk_write_coordinator, "use_client", write_client)
+            monkeypatch.setattr("one_memory.api._recall_model_ready", _model_ready)
+            async with client_for(app) as client:
+                await scenario(app, sdk, client)
+
+    asyncio.run(main())
+
+
+PASS_THROUGH = {
+    "min_score": 0.5,
+    "schema_min_score": 0.6,
+    "working_fact_min_score": 0.3,
+    "working_experience_min_score": 0.45,
+    "schema_limit": 4,
+    "intention_limit": 2,
+    "rerank": False,
+    "reader": "legacy",
+    "scene": "chat",
+    "search_mode": "agentic",
+    "include_sources": ["hermes", "codex"],
+    "exclude_sources": ["cursor"],
+    "created_after": 1_780_000_000,
+    "session_ids": ["s-1"],
+    "agent_ids": ["all-apps"],
+    "workspace_id": "ws-1",
+    "branch": "main",
+    "layers": ["l2_fact", "l0_basic_info"],
+}
+#: How each field reaches `client.search` (lists stay lists).
+EXPECTED_SEARCH = {**PASS_THROUGH}
+
+
+def _memory_search(sdk: RecordingSdk) -> dict:
+    return [call for call in sdk.all("search") if call.get("layers") != ["l1_raw"]][-1]
+
+
+@pytest.mark.fresh_sdk
+def test_every_context_search_option_reaches_the_sdk_and_absent_ones_keep_the_old_call(
+    tmp_path, monkeypatch
+) -> None:
+    async def scenario(app, sdk, client):
+        baseline = await client.post(
+            "/api/v2/context/search", json={"query": "tea", "kinds": ["memory"]}
+        )
+        assert baseline.status_code == 200, baseline.text
+        before = _memory_search(sdk)
+        assert before["min_score"] == before["schema_min_score"] == 0.35
+        assert before["scene"] == "normal" and before["reader"] == "hybrid"
+        assert before["schema_limit"] == 3 and before["intention_limit"] == 1
+        for absent in ("rerank", "search_mode", "layers", "session_ids", "created_after"):
+            assert absent not in before
+
+        response = await client.post(
+            "/api/v2/context/search",
+            json={"query": "tea", "kinds": ["memory"], **PASS_THROUGH},
+        )
+        assert response.status_code == 200, response.text
+        sent = _memory_search(sdk)
+        for name, value in EXPECTED_SEARCH.items():
+            assert sent[name] == value, name
+        assert sent["user_ids"] == ["local"]
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_every_recall_and_ask_option_reaches_the_sdk(tmp_path, monkeypatch) -> None:
+    async def scenario(app, sdk, client):
+        recalled = await client.post(
+            "/api/v1/recall", json={"query": "tea", "user_id": "wife", **PASS_THROUGH}
+        )
+        assert recalled.status_code == 200, recalled.text
+        sent = _memory_search(sdk)
+        for name, value in EXPECTED_SEARCH.items():
+            assert sent[name] == value, name
+        assert sent["user_ids"] == ["wife"]
+
+        ask_options = {
+            name: value
+            for name, value in PASS_THROUGH.items()
+            if name
+            not in {
+                "working_fact_min_score",
+                "working_experience_min_score",
+                "schema_limit",
+                "intention_limit",
+                "reader",
+                "scene",
+                "search_mode",
+                "workspace_id",
+                "branch",
+            }
+        }
+        asked = await client.post(
+            "/api/v2/memories/ask", json={"question": "tea?", "user_id": "wife", **ask_options}
+        )
+        assert asked.status_code == 200, asked.text
+        sent = _memory_search(sdk)
+        for name, value in ask_options.items():
+            assert sent[name] == value, name
+        assert sent["user_ids"] == ["wife"] and sent["scene"] == "normal"
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_min_score_is_enforced_on_what_core_returns(tmp_path, monkeypatch) -> None:
+    """Acceptance item 9: `min_score=0.9` answers only results scoring ≥ 0.9."""
+
+    async def scenario(app, sdk, client):
+        context = await client.post(
+            "/api/v2/context/search", json={"query": "tea", "min_score": 0.9}
+        )
+        assert context.status_code == 200, context.text
+        items = context.json()["items"]
+        assert items and all(item["score"] >= 0.9 for item in items)
+        assert {item.get("memory_id") or item.get("raw_id") for item in items} == {
+            "m-high",
+            "raw-1",
+        }
+        raw_call = sdk.last("search") if sdk.last("search").get("layers") else None
+        assert raw_call is None or raw_call["min_score"] == 0.9
+
+        # The fused scene ranks by 1/(60+rank); the lane's own score is the floor's.
+        fused = await client.post(
+            "/api/v2/context/search",
+            json={"query": "tea", "kinds": ["memory"], "scene": "all", "min_score": 0.9},
+        )
+        assert [item["memory_id"] for item in fused.json()["items"]] == ["m-high"]
+        assert fused.json()["items"][0]["score"] == 0.95
+
+        recalled = await client.post("/api/v1/recall", json={"query": "tea", "min_score": 0.9})
+        assert recalled.status_code == 200, recalled.text
+        assert [hit["memory_id"] for hit in recalled.json()["hits"]] == ["m-high"]
+
+        # A working-scene fact is read back from its own namespace.
+        working = await client.post(
+            "/api/v2/context/search",
+            json={"query": "uv", "kinds": ["memory"], "scene": "working", "min_score": 0.9},
+        )
+        assert [item["memory_id"] for item in working.json()["items"]] == ["f-1"]
+        assert sdk.last("get") == {"memory_id": "f-1", "kind": "fact", "variant": "working"}
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+@pytest.mark.parametrize(
+    "extra",
+    [
+        {"min_score": 1.5},
+        {"min_score": -0.1},
+        {"schema_min_score": "high"},
+        {"scene": "normal"},
+        {"search_mode": "fast"},
+        {"reader": "vector"},
+        {"layers": ["l9_nothing"]},
+        {"layers": []},
+        {"branch": "main"},
+        {"rerank": "no"},
+        {"created_after": -1},
+        {"user_ids": ["wife"]},
+        {"embed_override": {"model": "x"}},
+        {"query_embedding": [0.1]},
+        {"request_id": "r-1"},
+    ],
+)
+def test_out_of_range_and_unexposed_search_fields_answer_422(tmp_path, extra) -> None:
+    async def scenario():
+        app = application(tmp_path)
+        async with app.router.lifespan_context(app), client_for(app) as client:
+            for path, body in (
+                ("/api/v2/context/search", {"query": "q"}),
+                ("/api/v1/recall", {"query": "q"}),
+                ("/api/v2/memories/ask", {"question": "q"}),
+            ):
+                response = await client.post(path, json={**body, **extra})
+                assert response.status_code == 422, (path, response.text)
+
+    asyncio.run(scenario())
+
+
+@pytest.mark.fresh_sdk
+def test_raw_lane_takes_the_callers_sources_only_inside_its_authority(
+    tmp_path, monkeypatch
+) -> None:
+    async def scenario(app, sdk, client):
+        response = await client.post(
+            "/api/v2/context/search",
+            json={
+                "query": "tea",
+                "kinds": ["raw"],
+                "include_sources": ["hermes"],
+                "session_ids": ["s-1"],
+                "created_after": 1_700_000_000,
+            },
+        )
+        assert response.status_code == 200, response.text
+        raw_call = sdk.last("search")
+        assert raw_call["layers"] == ["l1_raw"]
+        assert raw_call["include_sources"] == ["hermes"]
+        assert raw_call["session_ids"] == ["s-1"]
+        assert raw_call["created_after"] == 1_700_000_000
+        # Memory-lane knobs never reach the raw search.
+        assert raw_call["scene"] == "chat" and raw_call["schema_limit"] == 0
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_library_honours_layers_statuses_and_filters(tmp_path, monkeypatch) -> None:
+    """Acceptance item 10: `layer=l0_basic_info` returns only L0 records."""
+
+    async def scenario(app, sdk, client):
+        page = await client.get("/api/v2/memory-library", params={"layer": "l0_basic_info"})
+        assert page.status_code == 200, page.text
+        assert {memory["layer"] for memory in page.json()["memories"]} == {"l0_basic_info"}
+        sent = sdk.last("list_memories")
+        assert sent["layers"] == ["l0_basic_info"] and sent["scene"] is None
+
+        await client.get(
+            "/api/v2/memory-library",
+            params=[
+                ("layers", "l2_fact,l6_schema"),
+                ("layers", "l7_intention"),
+                ("statuses", "active,superseded"),
+                ("session_ids", "s-1"),
+                ("agent_ids", "all-apps"),
+                ("exclude_sources", "cursor"),
+                ("user_id", "wife"),
+            ],
+        )
+        sent = sdk.last("list_memories")
+        assert sent["layers"] == ["l2_fact", "l6_schema", "l7_intention"]
+        assert sent["statuses"] == ["active", "superseded"]
+        assert sent["session_ids"] == ["s-1"] and sent["agent_ids"] == ["all-apps"]
+        assert sent["exclude_sources"] == ["cursor"] and sent["user_id"] == "wife"
+
+        # Without a layer the page is what it always was.
+        await client.get("/api/v2/memory-library")
+        sent = sdk.last("list_memories")
+        assert sent["scene"] == "all" and "layers" not in sent
+
+        searched = await client.get(
+            "/api/v2/memory-library", params={"query": "tea", "layers": "l2_fact"}
+        )
+        assert searched.status_code == 200, searched.text
+        assert sdk.last("search")["layers"] == ["l2_fact"]
+
+        for params in (
+            {"layers": "l9_nothing"},
+            {"statuses": "gone"},
+            {"layers": "l2_fact", "scene": "working"},
+        ):
+            refused = await client.get("/api/v2/memory-library", params=params)
+            assert refused.status_code == 422, (params, refused.text)
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_layer_counts_read_each_layer_for_one_user(tmp_path, monkeypatch) -> None:
+    async def scenario(app, sdk, client):
+        counts = await client.get("/api/v2/memory-library/layers", params={"user_id": "wife"})
+        assert counts.status_code == 200, counts.text
+        body = counts.json()
+        assert body["layers"]["l1_raw"] == 1
+        assert body["layers"]["l2_fact"] == 1
+        assert body["layers"]["l0_basic_info"] == 1
+        assert body["total"] == sum(body["layers"].values())
+        calls = sdk.all("list_memories")
+        assert {call["user_id"] for call in calls} == {"wife"}
+        [raw_call] = [call for call in calls if call["layers"] == ["l1_raw"]]
+        assert raw_call["statuses"] == ["active", "shadow"]
+        assert (
+            await client.get("/api/v2/memory-library/layers", params={"layer": "x"})
+        ).status_code == 422
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_raw_lists_active_and_shadow_turns_with_roles_and_hashes(tmp_path, monkeypatch) -> None:
+    async def scenario(app, sdk, client):
+        page = await client.get("/api/v2/raw")
+        assert page.status_code == 200, page.text
+        body = page.json()
+        assert body["total"] == 2 and len(body["items"]) == 2
+        sent = sdk.last("list_memories")
+        assert sent["layers"] == ["l1_raw"] and sent["statuses"] == ["active", "shadow"]
+        first = next(item for item in body["items"] if item["raw_id"] == "raw-1")
+        assert first["status"] == "shadow"
+        assert [message["role"] for message in first["messages"]] == ["user", "assistant"]
+        assert first["messages"][0]["content"] == "我叫费天翔"
+        assert first["role"] is None
+        expected = hashlib.sha256(first["content"].encode()).hexdigest()
+        assert first["content_hash"] == f"sha256:{expected}"
+        assert first["session_id"] == "s-1" and first["source"] == "hermes"
+        assert first["created"].startswith("2026-")
+        second = next(item for item in body["items"] if item["raw_id"] == "raw-2")
+        assert second["role"] == "user"
+
+        await client.get(
+            "/api/v2/raw",
+            params=[("source", "hermes,codex"), ("session_id", "s-1"), ("order", "asc")],
+        )
+        sent = sdk.last("list_memories")
+        assert sent["include_sources"] == ["hermes", "codex"]
+        assert sent["session_ids"] == ["s-1"] and sent["order"] == "asc"
+
+        windowed = (await client.get("/api/v2/raw", params={"after": 1_790_050_000})).json()
+        assert windowed["time_filtered"] is True
+        assert [item["raw_id"] for item in windowed["items"]] == ["raw-2"]
+
+        wife = (await client.get("/api/v2/raw", params={"user_id": "wife"})).json()
+        assert [item["raw_id"] for item in wife["items"]] == ["raw-w"]
+        assert sdk.last("list_memories")["user_id"] == "wife"
+        assert (await client.get("/api/v2/raw", params={"statuses": "bogus"})).status_code == 422
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_profile_versions_read_the_l0_chain_for_one_user(tmp_path, monkeypatch) -> None:
+    async def scenario(app, sdk, client):
+        response = await client.get("/api/v2/user-profile/versions", params={"user_id": "wife"})
+        assert response.status_code == 200, response.text
+        body = response.json()
+        assert body["user_id"] == "wife"
+        assert sdk.last("get_user_basic_info") == {"user_id": "wife"}
+        [chain] = body["chains"]
+        assert chain["memory_id"] == "wife-l0-head"
+        assert chain["basic_info"] == {"name": "Fei"}
+        assert [(v["memory_id"], v["current"]) for v in chain["versions"]] == [
+            ("wife-l0-head", True),
+            ("wife-l0-old", False),
+        ]
+        local = (await client.get("/api/v2/user-profile/versions")).json()
+        assert local["user_id"] == "local"
+        assert sdk.last("get_user_basic_info") == {"user_id": "local"}
+        refused = await client.get("/api/v2/user-profile/versions", params={"owner": "wife"})
+        assert refused.status_code == 422
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+@pytest.mark.fresh_sdk
+def test_graph_takes_layers_and_session_filters(tmp_path, monkeypatch) -> None:
+    async def scenario(app, sdk, client):
+        response = await client.get(
+            "/api/v2/memories/graph",
+            params={"layers": "l2_fact", "session_ids": "s-1", "user_id": "wife"},
+        )
+        assert response.status_code == 200, response.text
+        sent = sdk.last("list_memories")
+        assert sent["layers"] == ["l2_fact"] and sent["scene"] is None
+        assert sent["session_ids"] == ["s-1"] and sent["user_id"] == "wife"
+
+    run_with_sdk(tmp_path, monkeypatch, scenario)
+
+
+def _hermes_state_db(root: Path, text: str) -> None:
+    root.mkdir(parents=True, exist_ok=True)
+    connection = sqlite3.connect(root / "state.db")
+    connection.execute(
+        "CREATE TABLE messages (id INTEGER PRIMARY KEY, session_id TEXT, role TEXT,"
+        " content TEXT, timestamp REAL, tool_name TEXT, active INTEGER)"
+    )
+    connection.executemany(
+        "INSERT INTO messages VALUES (?, ?, ?, ?, 1790000000, NULL, 1)",
+        [(1, "session-1", "user", text), (2, "session-1", "assistant", "ok")],
+    )
+    connection.commit()
+    connection.close()
+
+
+@pytest.mark.fresh_sdk
+def test_v2_source_and_import_aliases_behave_like_v1_per_user(tmp_path) -> None:
+    async def scenario():
+        home = tmp_path / "home"
+        hermes = home / ".hermes"
+        wife = hermes / "profiles" / "wife"
+        _hermes_state_db(hermes, "I am the husband")
+        _hermes_state_db(wife, "I am the wife")
+        app = application(tmp_path, home=home)
+        async with app.router.lifespan_context(app), client_for(app) as client:
+            location = (await client.post("/api/v1/sources/hermes/location")).json()
+            local = await client.post(
+                "/api/v1/sources/connect",
+                json={
+                    "source_ids": ["hermes"],
+                    "request_id": str(uuid4()),
+                    "source_roots": {"hermes": location["root_id"]},
+                    "enable_sync": True,
+                },
+            )
+            assert local.status_code == 200, local.text
+            connected = await client.post(
+                "/api/v1/sources/connect",
+                json={
+                    "source_ids": ["hermes"],
+                    "request_id": str(uuid4()),
+                    "user_id": "wife",
+                    "root": str(wife),
+                },
+            )
+            assert connected.status_code == 200, connected.text
+            for job in (*local.json()["jobs"], *connected.json()["jobs"]):
+                await app.state.import_supervisor.wait(job["job_id"])
+
+            v1 = (await client.get("/api/v1/sources")).json()
+            v2 = (await client.get("/api/v2/sources")).json()
+            assert v1 == v2
+            wife_sources = (await client.get("/api/v2/sources", params={"user_id": "wife"})).json()
+            assert [source["source_id"] for source in wife_sources] == ["hermes"]
+            assert wife_sources[0]["authorized"] is True
+            assert (await client.get("/api/v2/sources", params={"u": "x"})).status_code == 422
+
+            created = await client.post(
+                "/api/v2/import-jobs",
+                json={"source_id": "hermes", "operation": "sync", "user_id": "wife"},
+            )
+            assert created.status_code == 200, created.text
+            wife_job = created.json()
+            assert wife_job["user_id"] == "wife" and wife_job["source_id"] == "hermes"
+            await app.state.import_supervisor.wait(wife_job["job_id"])
+            refused = await client.post(
+                "/api/v2/import-jobs",
+                json={"source_id": "claude-code", "user_id": "wife"},
+            )
+            assert refused.status_code in {404, 422}
+            unknown = await client.post(
+                "/api/v2/import-jobs",
+                json={"source_id": "hermes", "from_beginning": True},
+            )
+            assert unknown.status_code == 422
+
+            listed = (await client.get("/api/v2/import-jobs", params={"user_id": "wife"})).json()
+            assert {job["user_id"] for job in listed} == {"wife"}
+            assert (await client.get("/api/v2/import-jobs")).json() == (
+                await client.get("/api/v1/import-jobs")
+            ).json()
+            path = f"/api/v2/import-jobs/{wife_job['job_id']}"
+            assert (await client.get(path)).status_code == 404
+            own = await client.get(path, params={"user_id": "wife"})
+            assert own.status_code == 200 and own.json()["job_id"] == wife_job["job_id"]
+
+            synced = await client.post("/api/v2/sources/hermes/sync", json={"user_id": "wife"})
+            v1_synced = await client.post("/api/v1/sources/hermes/sync")
+            assert synced.status_code == v1_synced.status_code, (synced.text, v1_synced.text)
+            assert synced.status_code == 200
+            assert synced.json()["user_id"] == "wife"
+            assert v1_synced.json()["user_id"] == "local"
+            other = await client.post("/api/v2/sources/claude-code/sync", json={"user_id": "wife"})
+            assert other.status_code in {404, 422}
+            assert (
+                await client.post("/api/v2/sources/hermes/sync", json={"owner": "wife"})
+            ).status_code == 422
+
+    asyncio.run(scenario())

memhub/apps/core/tests/test_sdk_data_plane_migration.py

@@ -6,7 +6,7 @@ import io
 import json
 import sqlite3
 from contextlib import closing
-from datetime import UTC, datetime, timedelta
+from datetime import UTC, datetime
 from pathlib import Path
 
 import pytest
@@ -30,7 +30,6 @@ from one_memory.sdk_data_plane_migration import (
     MemoryAuthority,
     SdkDataPlaneMigrationStore,
 )
-from one_memory.sdk_memory_runtime import SdkMemoryRuntime
 from one_memory.sdk_migration_bundle import create_sdk_migration_bundle
 from one_memory.sdk_migration_runner import (
     PREPARED_PHASE,
@@ -708,6 +707,7 @@ def test_cutover_waits_until_the_new_runtime_reads_every_chat_memory(store, acti
         "chat_readable": 1,
         "productivity_expected": 0,
         "productivity_readable": 0,
+        "chat_missing_before_migration": 0,
         "passed": False,
     }
     assert store.authority() is MemoryAuthority.LEGACY
@@ -774,18 +774,186 @@ def test_retired_coding_experiences_do_not_block_but_an_empty_store_does(store,
     assert (result.readback.productivity_expected, result.readback.productivity_readable) == (2, 1)
 
 
-def test_memories_without_a_model_profile_keep_legacy_authority(store, monkeypatch):
+def test_a_home_without_a_model_profile_migrates_on_the_builtin_embedding(store, monkeypatch):
+    """#234: no generation model was ever configured, and the upgrade stopped.
+
+    No profile means no LLM, not no embedding: the vectors are the built-in
+    model's, the same space the Runtime reads with. They are reattached and
+    read back through it, and the client is opened without resolving a key.
+    """
+    from one_memory.hy_sdk_runtime import builtin_retrieval_snapshot, vector_collection_bases
+
     monkeypatch.setattr("one_memory.sdk_migration_runner.read_active_profile", lambda _db: None)
+    home = store.database_path.parent
+    builtin = builtin_retrieval_snapshot()
+    _current, legacy_bases = vector_collection_bases(builtin)
+    client = _chroma(home)
+    _fill(client, f"{legacy_bases[0]}_512", ["m1", "m2"])
+    del client
+    _confirmed(store, "m1", "m2")
+    opened: list[str] = []
+
+    def open_client(profile):
+        opened.append(profile.profile_id)
+        return _reading({"m1", "m2"})(profile)
+
+    result = run_startup_migration(
+        data_dir=home,
+        runtime_build="21",
+        sdk_version="2.0.5",
+        namespace="governed-v2",
+        open_client=open_client,
+    )
+
+    assert result.state == "cutover"
+    assert opened == [builtin.profile_id]
+    assert [item.legacy for item in result.reattached] == [f"{legacy_bases[0]}_512"]
+    assert result.readback is not None and result.readback.passed
+    assert store.authority() is MemoryAuthority.SDK
+
+
+def test_the_builtin_readback_client_resolves_no_secret(tmp_path, monkeypatch):
+    """The built-in snapshot has no stored version and no key to look up."""
+    from one_memory import sdk_migration_verify
+    from one_memory.hy_sdk_runtime import builtin_retrieval_snapshot
+
+    leased: list[tuple[str, object]] = []
+
+    class Lease:
+        client = object()
+
+        def close(self) -> None:
+            leased.append(("closed", None))
+
+    def acquire(profile, *, api_key, data_dir):
+        leased.append((profile.profile_id, api_key))
+        return Lease()
+
+    def no_lookup(*_args, **_kwargs):
+        raise AssertionError("a key-less profile must not touch the secret store")
+
+    monkeypatch.setattr("one_memory.hy_sdk_runtime.acquire_client", acquire)
+    monkeypatch.setattr(
+        "one_memory.model_profiles.ModelProfileStore.resolve_snapshot_secret", no_lookup
+    )
+    with sdk_migration_verify.runtime_sdk_client(
+        tmp_path / "missing.sqlite3", tmp_path, builtin_retrieval_snapshot()
+    ) as client:
+        assert client is Lease.client
+    assert leased == [("builtin-retrieval", None), ("closed", None)]
+
+
+def test_vectors_left_by_a_deleted_profile_stop_the_upgrade_by_name(store, monkeypatch):
+    """No active profile, and the vectors sit in another profile's space.
+
+    The built-in model cannot read them, so the old abort was right for this
+    one shape; it now says what is actually wrong, before anything changes.
+    """
+    from one_memory.hy_sdk_runtime import vector_collection_bases
+    from one_memory.model_profiles import ModelProfileSnapshot, ProviderKind
+
+    monkeypatch.setattr("one_memory.sdk_migration_runner.read_active_profile", lambda _db: None)
+    deleted = ModelProfileSnapshot(
+        profile_id="deleted",
+        revision=3,
+        provider_kind=ProviderKind.OPENAI_COMPATIBLE,
+        base_url="https://models.example/v1",
+        model="chat",
+        embedding_provider_kind=ProviderKind.OPENAI_COMPATIBLE,
+        embedding_base_url="https://models.example/v1",
+        embedding_model="text-embedding-3-large",
+        embedding_dimensions=1024,
+        timeout_seconds=60,
+        max_concurrency=1,
+        context_window=None,
+        has_api_key=True,
+    )
+    current, _legacy = vector_collection_bases(deleted)
+    home = store.database_path.parent
+    client = _chroma(home)
+    _fill(client, f"{current}_1024", ["m1"])
+    del client
     _confirmed(store, "m1")
 
-    with pytest.raises(DataPlaneMigrationError, match="sdk_migration_model_profile_required"):
+    with pytest.raises(DataPlaneMigrationError, match="sdk_migration_vector_profile_missing"):
         run_startup_migration(
-            data_dir=store.database_path.parent,
+            data_dir=home,
             runtime_build="21",
             sdk_version="2.0.5",
             namespace="governed-v2",
+            open_client=_reading({"m1"}),
         )
     assert store.authority() is MemoryAuthority.LEGACY
+    assert not (home / "backups").exists()
+
+
+def test_confirmed_chats_the_sdk_had_already_merged_do_not_fail_the_upgrade(store, active_profile):
+    """#229/#240: 658 confirmed chats, 657 in the store, and the upgrade failed.
+
+    A 1.0.4 library can hold confirmed rows whose SDK node had been merged or
+    deleted before the upgrade; no migration can read those back. The check
+    holds chat to what the store held before this run and counts the rest.
+    """
+    from one_memory.hy_sdk_runtime import vector_collection_bases
+
+    home = store.database_path.parent
+    stored = [f"chat-{index:03d}" for index in range(657)]
+    current, _legacy = vector_collection_bases(active_profile)
+    client = _chroma(home)
+    _fill(client, f"{current}_512", stored)
+    del client
+    _confirmed(store, *stored, "chat-merged-before-upgrade")
+
+    result = run_startup_migration(
+        data_dir=home,
+        runtime_build="21",
+        sdk_version="2.0.5",
+        namespace="governed-v2",
+        open_client=_reading(set(stored)),
+    )
+
+    assert result.state == "cutover"
+    assert result.readback is not None
+    assert result.readback.as_document() == {
+        "chat_expected": 657,
+        "chat_readable": 657,
+        "productivity_expected": 0,
+        "productivity_readable": 0,
+        "chat_missing_before_migration": 1,
+        "passed": True,
+    }
+    assert store.authority() is MemoryAuthority.SDK
+
+
+def test_a_chat_in_the_store_before_the_upgrade_must_still_read_back(store, active_profile):
+    """Only what was already gone is excused; losing a stored memory still fails."""
+    from one_memory.hy_sdk_runtime import vector_collection_bases
+    from one_memory.sdk_migration_verify import MemoryReadbackFailed
+
+    home = store.database_path.parent
+    current, _legacy = vector_collection_bases(active_profile)
+    client = _chroma(home)
+    _fill(client, f"{current}_512", ["m1", "m2"])
+    del client
+    _confirmed(store, "m1", "m2", "merged")
+
+    with pytest.raises(MemoryReadbackFailed) as caught:
+        run_startup_migration(
+            data_dir=home,
+            runtime_build="21",
+            sdk_version="2.0.5",
+            namespace="governed-v2",
+            open_client=_reading({"m1"}),
+        )
+    assert caught.value.report.as_document() == {
+        "chat_expected": 2,
+        "chat_readable": 1,
+        "productivity_expected": 0,
+        "productivity_readable": 0,
+        "chat_missing_before_migration": 1,
+        "passed": False,
+    }
+    assert store.authority() is MemoryAuthority.LEGACY
 
 
 def test_a_build_20_home_gets_its_vectors_back_and_is_checked_against_its_backup(
@@ -1072,7 +1240,13 @@ def test_host_capture_is_fenced_for_the_whole_migration(store):
     )
 
     assert result.state == "cutover"
-    assert fenced == {"bundle": True, "prepare": True, "cutover": True, "cleanup": True}
+    assert fenced == {
+        "bundle": True,
+        "prepare": True,
+        "vectors": True,
+        "cutover": True,
+        "cleanup": True,
+    }
     assert not sdk_migration_capture_fenced(home)
 
 
@@ -1175,57 +1349,6 @@ def test_raw_http_dto_uses_the_unified_memory_list_shape():
     assert page.raw[0].custom == {"role": "user"}
 
 
-def test_sdk_raw_retention_keeps_only_the_latest_three_days():
-    now = datetime(2026, 10, 23, tzinfo=UTC)
-
-    class Client:
-        def __init__(self):
-            self.rows = [
-                {
-                    "memory_id": "old",
-                    "memory_at": int((now - timedelta(days=4)).timestamp()),
-                },
-                {
-                    "memory_id": "recent",
-                    "memory_at": int((now - timedelta(days=2)).timestamp()),
-                },
-            ]
-
-        def list_memories(self, **_kwargs):
-            return {
-                "vdb": {
-                    "memories": list(self.rows),
-                    "total": len(self.rows),
-                    "limit": 500,
-                    "offset": 0,
-                }
-            }
-
-        def delete(self, memory_id):
-            before = len(self.rows)
-            self.rows = [row for row in self.rows if row["memory_id"] != memory_id]
-            return {
-                "success": True,
-                "deleted_count": before - len(self.rows),
-            }
-
-    client = Client()
-
-    class Handle:
-        async def client(self):
-            return client
-
-    runtime = SdkMemoryRuntime(
-        coordinator=None,
-        read_handle=Handle(),
-        write_queue=None,
-        extraction_settings=None,
-        data_dir=Path("."),
-    )
-    assert asyncio.run(runtime.prune_expired_raw(now=now)) == 1
-    assert [row["memory_id"] for row in client.rows] == ["recent"]
-
-
 # The UI asks which memory data plane is authoritative and must get the
 # answer the API gates use.
 
@@ -1353,7 +1476,7 @@ def test_the_sdk_memory_map_draws_sdk_memories_without_legacy_tables(tmp_path) -
             original = app.state.sdk_client_handle
             app.state.sdk_client_handle = Handle()
 
-            async def graph_snapshot(*, limit):
+            async def graph_snapshot(*, limit, user_id="local"):
                 del limit
                 return HyGraphSnapshot(
                     nodes=(HyGraphNode(node_id="t1", content="topic", layer="TOPIC"),),

memhub/apps/core/tests/test_sdk_migration_bundle.py

@@ -91,7 +91,7 @@ def test_bundle_contains_management_sdk_config_and_verified_manifest(tmp_path):
     assert manifest["bundle_id"] == "migration-1"
     assert manifest["sdk_namespace"] == "governed-v2"
     assert manifest["runtime_build"] == "build-13"
-    assert manifest["snapshot_sdk_version"] == "2.0.8"
+    assert manifest["snapshot_sdk_version"] == "2.0.9"
     paths = {item["path"]: item for item in manifest["files"]}
     assert {
         "management/one-memory.sqlite3",

memhub/apps/core/tests/test_sdk_migration_log.py

@@ -306,7 +306,7 @@ def test_the_runner_names_the_step_and_sub_step_that_failed(tmp_path, monkeypatc
         ("step.finished", "preflight.profile_read"),
         ("step.finished", "preflight.resumable_identity"),
         ("step.finished", "preflight.legacy_ids_scan"),
-        ("step.finished", "preflight.model_profile_check"),
+        ("step.finished", "preflight.vector_space_check"),
         ("step.finished", "prepare.recover_abandoned_staging"),
         ("step.finished", "bundle.ensure"),
         ("step.finished", "bundle.restore_point"),

memhub/apps/core/tests/test_sdk_migration_restore.py

@@ -463,11 +463,17 @@ def test_a_run_killed_after_its_restore_point_is_rolled_back_on_the_next_start(h
 
 
 def test_a_preflight_failure_changes_nothing_and_needs_no_restore(home, monkeypatch):
+    def unreadable(_data_dir, _profile):
+        raise DataPlaneMigrationError("sdk_migration_vector_profile_missing")
+
     monkeypatch.setattr("one_memory.sdk_migration_runner.read_active_profile", lambda _db: None)
+    monkeypatch.setattr(
+        "one_memory.sdk_migration_runner.vectors_belong_to_another_space", unreadable
+    )
     before = _state(home)
     report: dict[str, object] = {}
 
-    with pytest.raises(DataPlaneMigrationError, match="sdk_migration_model_profile_required"):
+    with pytest.raises(DataPlaneMigrationError, match="sdk_migration_vector_profile_missing"):
         _run(home, report=report)
 
     assert report == {}

memhub/apps/core/tests/test_search_without_profile.py

@@ -1,5 +1,6 @@
 """Model-free SDK search remains available without an active write profile."""
 
+import asyncio
 from pathlib import Path
 from types import SimpleNamespace
 
@@ -95,3 +96,158 @@ async def test_search_without_profile_uses_public_sdk_search() -> None:
     assert options["user_ids"] == ["local"]
     assert options["agent_ids"] == []
     assert options["scene"] == "normal"
+
+
+def _activate(profiles, secrets, *, dimensions: int, model: str = "chat"):
+    from pydantic import SecretStr
+
+    from one_memory.model_profiles import ModelCapabilities, ModelProfileInput
+
+    profile = profiles.create(
+        ModelProfileInput(
+            name=f"embedding-{dimensions}",
+            provider_kind=ProviderKind.OPENAI_COMPATIBLE,
+            base_url="https://models.example/v1",
+            model=model,
+            embedding_provider_kind=ProviderKind.OPENAI_COMPATIBLE,
+            embedding_base_url="https://models.example/v1",
+            embedding_model="qwen3-embedding-4b",
+            embedding_dimensions=dimensions,
+            api_key=SecretStr("sk-test"),
+        ),
+        secrets,
+    )
+    profiles.record_capabilities(
+        profile.profile_id,
+        expected_revision=profile.revision,
+        capabilities=ModelCapabilities(
+            endpoint_reachable=True,
+            model_available=True,
+            structured_output=True,
+            embedding_available=True,
+        ),
+        secrets=secrets,
+    )
+    profiles.activate(profile.profile_id, secrets)
+    return profiles.snapshot(profile.profile_id, profiles.active_identity()[1])
+
+
+class _Recorder:
+    """Stands in for `acquire_client`: records the collection each client opens."""
+
+    def __init__(self, data_dir: Path) -> None:
+        self.data_dir = data_dir
+        self.opened: list[tuple[str, int | None, str | None]] = []
+        self.closed: list[str] = []
+
+    def __call__(self, profile, *, api_key, data_dir, http_transport_factory=None, mode="pro"):
+        from one_memory.hy_sdk_runtime import build_sdk_config
+
+        config = build_sdk_config(profile, api_key=api_key, data_dir=data_dir)
+        name = config.vector_store.collection_name
+        self.opened.append((name, config.vector_store.embedding_dims, api_key))
+        return SimpleNamespace(
+            client=SimpleNamespace(collection=name),
+            close=lambda: self.closed.append(name),
+        )
+
+
+@pytest.mark.anyio
+async def test_reads_open_the_collection_the_active_profile_writes(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """#242: a 2560-dim profile wrote `memhub_<sha>_2560`, reads opened `_512`."""
+    from one_memory.hy_sdk_runtime import build_sdk_config
+    from one_memory.model_profiles import ModelProfileStore
+    from one_memory.sdk_read_handle import ActiveProfileReadHandle
+    from one_memory.secrets import InMemorySecretStore
+
+    recorder = _Recorder(tmp_path)
+    monkeypatch.setattr("one_memory.hy_sdk_runtime.acquire_client", recorder)
+    profiles = ModelProfileStore(tmp_path / "one-memory.sqlite3")
+    secrets = InMemorySecretStore()
+    handle = ActiveProfileReadHandle(profiles=profiles, secrets=secrets, data_dir=tmp_path)
+
+    builtin = build_sdk_config(builtin_retrieval_snapshot(), api_key=None, data_dir=tmp_path)
+    assert (await handle.client()).collection == builtin.vector_store.collection_name
+
+    written = _activate(profiles, secrets, dimensions=2560)
+    writes = build_sdk_config(written, api_key="sk-test", data_dir=tmp_path)
+
+    async def read() -> str:
+        return str((await handle.client()).collection)
+
+    assert await asyncio.create_task(read()) == writes.vector_store.collection_name
+    assert recorder.opened[-1] == (writes.vector_store.collection_name, 2560, "sk-test")
+    assert writes.vector_store.collection_name != builtin.vector_store.collection_name
+    await handle.close()
+
+
+@pytest.mark.anyio
+async def test_a_profile_change_rebuilds_reads_without_closing_one_in_flight(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    from one_memory.model_profiles import ModelProfileStore
+    from one_memory.sdk_read_handle import ActiveProfileReadHandle
+    from one_memory.secrets import InMemorySecretStore
+
+    recorder = _Recorder(tmp_path)
+    monkeypatch.setattr("one_memory.hy_sdk_runtime.acquire_client", recorder)
+    profiles = ModelProfileStore(tmp_path / "one-memory.sqlite3")
+    secrets = InMemorySecretStore()
+    handle = ActiveProfileReadHandle(profiles=profiles, secrets=secrets, data_dir=tmp_path)
+    _activate(profiles, secrets, dimensions=1024)
+
+    holding = asyncio.Event()
+    finish = asyncio.Event()
+
+    async def slow_read() -> str:
+        client = await handle.client()
+        holding.set()
+        await finish.wait()
+        return str(client.collection)
+
+    in_flight = asyncio.create_task(slow_read())
+    await holding.wait()
+    [(old, _, _)] = recorder.opened
+
+    _activate(profiles, secrets, dimensions=2560, model="chat-2")
+
+    async def read() -> str:
+        return str((await handle.client()).collection)
+
+    new = await asyncio.create_task(read())
+    assert new != old
+    # The read that started on the old space still has its client.
+    assert recorder.closed == []
+
+    finish.set()
+    assert await in_flight == old
+    for _ in range(5):
+        await asyncio.sleep(0)
+    assert recorder.closed == [old]
+    await handle.close()
+    assert recorder.closed == [old, new]
+
+
+def test_every_api_read_shares_the_active_profile_handle(tmp_path: Path) -> None:
+    """Library, search, recall, graph, profile, raw, ask and context all read
+    through `_sdk_client_handle`, so it must be the active-profile handle."""
+    from one_memory.api import _sdk_client_handle
+    from one_memory.config import Settings
+    from one_memory.model_profiles import ModelProfileStore
+    from one_memory.sdk_read_handle import ActiveProfileReadHandle
+    from one_memory.secrets import InMemorySecretStore
+
+    state = SimpleNamespace(
+        sdk_client_handle=None,
+        model_profile_store=ModelProfileStore(tmp_path / "one-memory.sqlite3"),
+        secret_store=InMemorySecretStore(),
+        settings=Settings(_env_file=None, data_dir=tmp_path),
+    )
+    app = SimpleNamespace(state=state)
+
+    handle = _sdk_client_handle(app)  # type: ignore[arg-type]
+
+    assert isinstance(handle, ActiveProfileReadHandle)
+    assert _sdk_client_handle(app) is handle is state.sdk_client_handle  # type: ignore[arg-type]

memhub/apps/core/tests/test_v2_authority_and_call_events.py

@@ -57,10 +57,10 @@ def client_for(app, headers=None):
 
 
 class Answerer:
-    async def answer(self, question):
+    async def answer(self, question, *, user_id="local"):
         return SdkMemoryAnswer(answer="真实数据。", memories=(MEMORY,), mode="extractive")
 
-    async def answer_stream(self, question):
+    async def answer_stream(self, question, *, user_id="local"):
         yield {"type": "recall", "memories": (MEMORY,)}
         yield {
             "type": "final",

memhub/apps/core/uv.lock

@@ -991,7 +991,7 @@ wheels = [
 
 [[package]]
 name = "hy-memory"
-version = "2.0.8"
+version = "2.0.9"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "chromadb" },
@@ -1008,9 +1008,9 @@ dependencies = [
     { name = "scikit-learn" },
     { name = "tiktoken" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/21/15/83ca83b649ead238f186fdb33efdc73615fbba279efbbdca3857c1a86beb/hy_memory-2.0.8.tar.gz", hash = "sha256:0fa0ed1a1160b389a3d7c13a9fb6528b140d7ec23cf49e65f1be5974820a576b", size = 856025, upload-time = "2026-10-28T07:35:47.783Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/48/28/e01fd9a92cd090c8236df0dd06dae049c5de66345e5bd9653497ed56125e/hy_memory-2.0.9.tar.gz", hash = "sha256:5f55ea92044ced19a00b4e56c565215adedd3439a1bdb9cfe9ffe42387c8f1a1", size = 858323, upload-time = "2026-11-01T00:29:49.914Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/11/6a/c699cac0ff06bd089fceeff90446130763d8187e649dbdb0313014491a10/hy_memory-2.0.8-py3-none-any.whl", hash = "sha256:68bd789df24e51e64bb231e97a8d24ee89b4aa5f9ba14c3e0824589e9213bf7c", size = 921119, upload-time = "2026-10-28T07:35:45.571Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/44/4368555a408a58daf902c81b62184a3c99f12abd45d4f1144bf79f171a6f/hy_memory-2.0.9-py3-none-any.whl", hash = "sha256:95dbea232d2041f9b0ea3d08dd751603c2a5c0e4422b97e74f664e6ca913549e", size = 923250, upload-time = "2026-11-01T00:29:47.814Z" },
 ]
 
 [[package]]
@@ -1837,7 +1837,7 @@ requires-dist = [
     { name = "fastapi", specifier = ">=0.141.1" },
     { name = "filelock", specifier = ">=3.32.4" },
     { name = "httpx", specifier = ">=0.28.1" },
-    { name = "hy-memory", specifier = "==2.0.8" },
+    { name = "hy-memory", specifier = "==2.0.9" },
     { name = "keyring", specifier = ">=25.7.0" },
     { name = "numpy", specifier = "==2.5.2" },
     { name = "onnxruntime", specifier = "==1.23.2" },

memhub/apps/desktop/src/stable-runtime.ts

@@ -133,6 +133,7 @@ const SAFE_STABLE_RUNTIME_ERROR_CODES = new Set([
   "sdk_migration_preflight_failed",
   "sdk_migration_memory_unreadable",
   "sdk_migration_model_profile_required",
+  "sdk_migration_vector_profile_missing",
   "sdk_migration_model_secret_unavailable",
   "sdk_migration_secure_storage_unavailable",
   "sdk_migration_disk_space_insufficient",
@@ -219,6 +220,8 @@ export const STABLE_RUNTIME_USER_MESSAGES: Record<string, string> = {
     "升级前核对发现，新版本还读不到你的全部记忆，所以没有切换数据源，旧数据保持原样可用。请把这个提示反馈给我们；原有记忆数据不会丢失。",
   sdk_migration_model_profile_required:
     "升级记忆数据格式需要先有可用的模型配置，用来核对新版本能否读回你的记忆。请在旧版本中完成模型配置后重启；你的记忆数据不会受到影响。",
+  sdk_migration_vector_profile_missing:
+    "你的记忆是用一个已删除的模型配置生成向量的，当前没有可用的模型配置能读回它们，所以没有升级。请在旧版本中重新添加那个模型配置并启用后重启；你的记忆数据不会受到影响。",
   sdk_migration_model_secret_unavailable:
     "模型密钥读不到（可能被系统钥匙串拒绝），无法核对新版本能否读回你的记忆。请允许 MemHub 访问钥匙串后重新启动；你的记忆数据不会受到影响。",
   sdk_migration_secure_storage_unavailable: