Initial open source release

tests/test_write_preprocessing_cache.py

@@ -0,0 +1,91 @@
+import pytest
+
+
+def test_bm25_preprocessing_is_cached(monkeypatch):
+    from hy_memory.pipelines._retrieval import lemmatize
+
+    calls = 0
+
+    def fake_english(text):
+        nonlocal calls
+        calls += 1
+        return text.lower()
+
+    monkeypatch.setattr(lemmatize, "_lemmatize_english", fake_english)
+    lemmatize._lemmatize_cached.cache_clear()
+
+    assert lemmatize.lemmatize_for_bm25("Repeated Candidate") == "repeated candidate"
+    assert lemmatize.lemmatize_for_bm25("Repeated Candidate") == "repeated candidate"
+    assert calls == 1
+
+
+def test_ner_preprocessing_is_cached(monkeypatch):
+    from hy_memory.pipelines._retrieval import ner
+
+    calls = 0
+
+    class Entity:
+        text = "Tokyo"
+        label_ = "GPE"
+
+    class Doc:
+        ents = [Entity()]
+
+    class NLP:
+        def __call__(self, text):
+            nonlocal calls
+            calls += 1
+            return Doc()
+
+    monkeypatch.setattr(ner, "_get_nlp", lambda: NLP())
+    ner._extract_entities_cached.cache_clear()
+
+    assert ner.extract_entities("Tokyo", for_boost=True) == [("Tokyo", "GPE")]
+    assert ner.extract_entities("Tokyo", for_boost=True) == [("Tokyo", "GPE")]
+    assert calls == 1
+
+
+def test_ner_result_cache_isolated_across_model_reload(monkeypatch):
+    from types import SimpleNamespace
+    from hy_memory.pipelines._retrieval import ner
+
+    def model(label):
+        return lambda text: SimpleNamespace(
+            ents=[SimpleNamespace(text=text, label_=label)]
+        )
+
+    monkeypatch.setattr(ner, "_ner_nlp", None)
+    monkeypatch.setattr(ner, "_ner_load_attempted", False)
+    monkeypatch.setattr(ner, "_load_nlp", lambda: model("GPE"))
+    ner._extract_entities_cached.cache_clear()
+    assert ner.extract_entities("London") == [("London", "GPE")]
+
+    monkeypatch.setattr(ner, "_ner_nlp", None)
+    monkeypatch.setattr(ner, "_ner_load_attempted", False)
+    monkeypatch.setattr(ner, "_load_nlp", lambda: model("ORG"))
+    ner._extract_entities_cached.cache_clear()
+    assert ner.extract_entities("London") == [("London", "ORG")]
+
+
+@pytest.mark.asyncio
+async def test_existing_tags_cache_can_be_disabled(monkeypatch):
+    from hy_memory.models.memory import MemoryNode
+    from hy_memory.pipelines.writer import MemoryWriter
+
+    class Store:
+        def __init__(self):
+            self.calls = 0
+
+        async def list_by_user(self, **kwargs):
+            self.calls += 1
+            node = MemoryNode(node_id="n", user_id="u", content="c")
+            node.tags = ["tag"]
+            return [node]
+
+    monkeypatch.setenv("MEMORY_EXISTING_TAGS_CACHE_TTL_SEC", "0")
+    writer = MemoryWriter.__new__(MemoryWriter)
+    store = Store()
+
+    await writer._collect_existing_tags(store, "u", "a")
+    await writer._collect_existing_tags(store, "u", "a")
+    assert store.calls == 2