This PR fixes two issues with MinerU PDF parser where table and image chunks were not properly classified or associated:
1. **Table chunks** were incorrectly classified as `text` instead of `table`
2. **Image chunks** were missing image resource association and classified as `text`
## Summary
The TitleChunker (both `hierarchy` and `group` methods) had no
token-size ceiling, so a long section without sub-headings became one
giant chunk. This adds a configurable `chunk_token_cap` that guarantees
every text chunk stays within a token budget.
When `TokenChunker` starts a fresh chunk with an overlap prefix (Go `computeOverlapPrefix` / Python visible-text cut), the previous chunk's **tail PDF coordinates were dropped**. As a result, the overlap head of a PDF chunk is displayed but **not highlighted** — the highlight box is shifted/truncated relative to the displayed span (infiniflow/ragflow#18148).
Unifies the Go TokenChunker merge path on a single `mergeUnits` core and
fixes coordinate-tag drift in the Python JSON merge at `overlap > 0`.
Rebased on top of #17979 (delimiter_mode convergence).
Converge `TokenChunker.delimiter_mode` from three values (`token_size`,
`delimiter`, `one`) to two (`delimiter`, `one`). The unified `delimiter`
mode now carries the old `token_size` semantics: when no active
(backtick) delimiter is present, text/JSON chunks are merged up to
`chunk_token_size`; when a backtick delimiter is present, the text is
split by it and not merged. `one` continues to be handled by the
separate `OneChunker`.
Ports the dataset knowledge compilation (wiki/graph/tree/mindmap) to the
Go scheduler with a status contract, aligns wiki storage/retrieval with
Python, and sizes prompts by content_length.
## Summary
Fixes a regression introduced by #17203 (strict-cap atom-split) and a
secondary delimiter-handling bug from #17723.
**Root cause:**
- #17203 added `_split_oversized_unit` / `_compute_chunk_update`, which
split oversize units into ≤ token_size pieces. This collapsed
`token_size=1` into 1-token chunks and set the cap at 512, mismatching
the model-layer truncation boundary (embedding ~8191 / rerank
500/4096/8192/2048). Atom-split is unnecessary: oversize units stay
whole and the model layer truncates.
- #17723's delimiter handling dropped consecutive delimiters (`A####B`
-> `A##B`), glued JSON items with `"".join`, ignored
`children_delimiters`, and stripped whitespace delimiters.
## Changes
- New pure helper `merge_paragraphs(paragraphs, token_size, strategy)`
with a `MergeStrategy` enum (`UNDER_CAP` / `OVER_CAP`); **default
`OVER_CAP`**. `UNDER_CAP` is a strict cap (never overflows
`token_size`); `OVER_CAP` greedily accumulates adjacent paragraphs while
the projected total stays within `token_size`, merging one
boundary-overflow paragraph before closing. Oversize paragraphs stand
alone.
- `naive_merge` / `naive_merge_with_images` /
`RAGFlowTxtParser.parser_txt` now use `merge_paragraphs`; atom-split
removed. `naive_merge` / `naive_merge_with_images` always split a
section on the delimiter whenever one is present (even when the section
already fits `token_size`), so delimiter text never leaks into a chunk.
Only the empty-delimiter (size-only) mode skips splitting.
- `token_chunker`: delimiter text is dropped (not stripped); JSON flush
joins buffered items with `"\n"`; `children_delimiters` and
`PDF_POSITIONS_KEY` are preserved on the delimiter path. PDF positions
are now attributed **per segment** — each split chunk carries only the
positions of the item(s) that contributed to it — fixing a leak where
page-N coordinates were attached to page-M chunks and all segments
shared one preview image.
- `test_txt_parser.py` rewritten to assert the new contract (not the old
strict cap); `naive_merge` and delimiter-case-sensitive matrices
updated.
## Contract (refs #17799)
- user specified delimiter = chunk boundary; user specified delimiter
text never enters a chunk.
- `token_size` = soft target + merge strategy; no atom-split.
- Default strategy = `OVER_CAP`; migration can switch to `UNDER_CAP`
(strict cap).
- `OVER_CAP` has no hard cap; the model layer truncates oversize units.
`UNDER_CAP` enforces a strict cap.
## Notes
- Closes the wrong-object revert in #17774 (revert #17723 would
re-introduce delimiter-in-chunk and the strict cap).
- Go-side alignment (`internal/ingestion/component/chunker/token.go`) is
a follow-up PR.
---------
Co-authored-by: CodeBuddy <noreply@tencent.com>
## Summary
`TokenChunker._invoke` discarded the user's configured `delimiters`
whenever no backtick-wrapped delimiter was present: it passed a
hardcoded `""` to `naive_merge`, so the configured delimiters (including
the default `["\n"]`) were never forwarded. `naive_merge` then ignored
the newline sentence boundary and cut chunks **mid-sentence** once the
token budget was exceeded.
## Root cause
`token_chunker.py:326` called `naive_merge(payload, chunk_token_size,
"", overlapped_percent)`. `_compile_delimiter_pattern` intentionally
returns `""` for bare (non-backtick) delimiters — that return value is a
*path selector* (empty → token-budget merge; non-empty → hard
`_split_text_by_pattern` split). The bug was not in that selector but in
the `else` branch, which threw away `self._param.delimiters` instead of
forwarding it.
## Fix
Forward the configured delimiters as a soft boundary:
```python
else naive_merge(
payload,
self._param.chunk_token_size,
"".join(self._param.delimiters),
overlapped_percent,
)
```
`naive_merge` already parses the string via the canonical
`parse_delimiter_field`, so bare and backtick-wrapped delimiters are
honored as soft boundaries while the token budget is still respected.
The path-selection role of `_compile_delimiter_pattern` is untouched:
backtick-wrapped delimiters still select the hard
`_split_text_by_pattern` path; bare delimiters still take the
token-budget merge path. No regression for any previously-working
(wrapped-delimiter) configuration.
## Test
Adds `rag/flow/tests/test_token_chunker_delimiter.py`:
- `test_token_chunker_token_size_mode_does_not_split_sentences` — fails
on the old code (3 sentences cut mid-stream), passes after the fix.
- `test_naive_merge_empty_delimiter_ignores_newline_break` — root-cause
companion asserting `naive_merge("")` cuts while `naive_merge("\n")`
preserves boundaries.
Both tests skip when the tokenizer is unavailable (dead-tokenizer
guard).
## Note
This branch contains only this one-line fix on top of `main`; it is
intentionally independent of the unrelated tokenizer/offline-BPE work.
Co-authored-by: CodeBuddy <noreply@codebuddy.ai>
## Summary
Refactor the Go `KnowledgeCompilerComponent` so its parameter is a
**single string template id** instead of a DSL-level `variant` (or
plural group id list). The `variant` is no longer in the DSL — it is now
**derived at runtime from the resolved compilation template's `kind`
field**.
This aligns the Go ingestion port with the frontend Compiler operator,
which emits a singular `compilation_template_group_id` and does not
write `variant` into the generated `compiler.json`.
### What problem does this PR solve?
Adds first-class support for **Mistral OCR** (`POST /v1/ocr`) as a
document parser, and fixes the long-standing bug where selecting
`mistral-ocr-latest` fails with `Can't find model for
<tenant>/image2text/mistral-ocr-latest`.
`mistral-ocr-latest` is Mistral's dedicated document-OCR endpoint, not a
vision-chat (`image2text`) model, but the catalog tagged it `image2text`
— so it resolved to the `CvModel` registry, which has no `Mistral`
entry, and there was no `OcrModel` entry either. This PR registers it
correctly and wires it end to end.
Closes#17056Closes#5782Closes#7075
**What it does**
1. **`MistralParser` + `MistralOcrModel`**
(`deepdoc/parser/mistral_parser.py`, `rag/llm/ocr_model.py`) — a proper
`OcrModel` factory `Mistral OCR`, mirroring the SoMark cloud-OCR
template. Tables stay inline as HTML; the page range maps to Mistral's
native `pages` selector (absolute page indices, billed per selected
page, so multi-task documents do not re-OCR the whole file); documents
over the inline limit go through the `/v1/files` signed-URL flow with
cleanup.
2. **Removes the `image2text` mis-tag** for `mistral-ocr-latest` from
the `Mistral` factory in `conf/llm_factories.json` (it now lives only in
the `Mistral OCR` factory, typed `ocr`). This is what closes the `Can't
find model` path.
3. **`MistralCV`** (`rag/llm/cv_model.py`) — a thin `GptV4` subclass
over Mistral's OpenAI-compatible endpoint, registering a `Mistral` entry
in the `CvModel` registry so Mistral vision models (`pixtral-*`) become
usable as `image2text` at all.
4. **Figure description** — Mistral-OCR-extracted figures are captioned
using the tenant's configured `image2text` model (any provider),
matching MinerU/deepdoc behaviour.
5. **Wires the parser into every chunking method** (`naive`, `paper`,
`book`, `laws`, `manual`, `one`, `presentation`) and the `rag/flow` DAG
path. This also fixes a related latent gap where those chunkers
forwarded only `mineru_llm_name`, so any model-based OCR provider
selected on a non-`naive` method silently fell through.
**Notes on the API contract** (verified against the live Mistral API):
`pages` is a selector (returns absolute `index`, bills only the
requested pages); `include_blocks: true` returns per-block bounding
boxes usable for chunk highlighting and figure cropping; large files use
`POST /v1/files` → signed URL → OCR → `DELETE`.
**Testing**: new unit tests cover the response→sections contract (both
the 2-tuple `naive` path and the typed 3-tuple DAG path), the
position-tag rescale, the HTTP client incl. upload failure/cleanup
paths, `parse_pdf` page-range threading, registry registration, env
config, the suffix normalization, the factory catalog entry, `MistralCV`
registration, and figure-description injection. Verified end to end
against the live Mistral API on real PDFs (table extraction,
page-selector cost avoidance, figure captioning).
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
- [x] New Feature (non-breaking change which adds functionality)
Fixes#16812
### Problem
In the `rag/flow` ingestion pipeline, when `TitleChunker` feeds
`TokenChunker`, the chapter-aware chunks are silently discarded and the
parser's raw flat json is re-chunked instead.
`TitleChunker` emits `output_format="chunks"` and writes its
chapter-aware output to the `chunks` field
(`rag/flow/chunker/title_chunker/common.py`,
`set_output("output_format", "chunks")`). But `TokenChunker._invoke`
only handles `output_format` in `["markdown", "text", "html"]`, then
falls through to the `# json` path which reads
`from_upstream.json_result`. There is no branch for `"chunks"`, so
`from_upstream.chunks` is never read.
Downstream effects reported in #16812: PageIndex/TOC extraction receives
flat line-level text instead of structured chapter blocks
(incorrect/duplicate/missing chapters), and retrieval quality degrades
because chunks are no longer aligned to document structure.
### Fix
Select the source list based on `output_format`, mirroring the exact
pattern already used in `title_chunker/common.py`:
```python
json_result = (from_upstream.chunks if from_upstream.output_format == "chunks" else from_upstream.json_result) or []
```
`chunks` items share the same dict shape as `json_result` items (both
consumed via `.get("text")`, `.get("doc_type_kwd")`, etc.), so they flow
through the existing token-sizing path unchanged. One-line change, no
behavior change for the `json`/`markdown`/`text`/`html` paths.
### Test
Adds `rag/flow/tests/test_token_chunker.py`, an isolated unit test that
runs the real `TokenChunker._invoke` (heavy deps stubbed; real pydantic
schema used when available) and asserts that with
`output_format="chunks"` the upstream `chunks` are consumed rather than
the raw parser `json`.
Verified RED -> GREEN: the test fails against the current code (reads
the raw json) and passes with the fix.
Signed-off-by: Yash Raj Pandey <yashpn62@gmail.com>
### What problem does this PR solve?
- Clear stale pipeline IDs and generated data when updating documents
without `pipeline_id`.
- Support tree compilation results in pipeline workflows.
- Update compilation templates in place while preserving existing
template IDs.
- Improve duplicate-template validation messages.
### Type of change
- [x] Bug Fix (non-breaking change which fixes an issue)
- [x] New Feature (non-breaking change which adds functionality)
Co-authored-by: Jin Hai <haijin.chn@gmail.com>
xugangqiang
Yingfeng
balibabu
MarMar Labs
buua436
euvre
Wang Qi
天海蒼灆
Aeovy
qinling0210
Ziyang Guo
Jin Hai
Zhichang Yu
taek105
Xavierando
Lynn
Kevin Hu
Yash Raj Pandey