Simplify AI memory pipeline

api/tests/test_memory_enrichment_baseline.py

@@ -7,7 +7,7 @@ from types import SimpleNamespace
 import pytest
 
 from app.features.memory.enrichment import enrich_memory_after_ingest_async
-from app.features.memory.llm_schemas import EnrichmentPayload, parse_json_payload
+from app.features.memory.llm_schemas import EnrichmentPayload
 from app.features.memory.models import MemorySource
 from app.features.user.models import User
 
@@ -19,8 +19,7 @@ def test_enrichment_payload_roundtrip() -> None:
         '"object_json":{"value":"北京","approximate_era":"1990年代"},'
         '"confidence":0.85,"source_chunk_id":"ch-1"}]}'
     )
-    p = parse_json_payload(raw, EnrichmentPayload)
-    assert p is not None
+    p = EnrichmentPayload.model_validate_json(raw)
     assert p.summary == "要点摘要"
     assert len(p.facts) == 1
     assert p.facts[0].subject == "王伟"
@@ -36,16 +35,27 @@ async def test_enrich_memory_after_ingest_async_single_llm_call(
 
     invoke_count = {"n": 0}
 
-    async def fake_invoke(llm, prompt, max_tokens, agent):
+    async def fake_run(llm, numbered, narrator_label):
         invoke_count["n"] += 1
-        assert agent == "memory.enrichment"
-        return (
-            '{"summary":"本轮要点",'
-            '"facts":[{"fact_type":"event","subject":"王伟","predicate":"住",'
-            '"object_json":{"value":"上海"},"confidence":0.8,"source_chunk_id":"ch1"}]}'
+        assert "[chunk_id=ch1]" in numbered
+        assert narrator_label == "老王"
+        return EnrichmentPayload.model_validate(
+            {
+                "summary": "本轮要点",
+                "facts": [
+                    {
+                        "fact_type": "event",
+                        "subject": "王伟",
+                        "predicate": "住",
+                        "object_json": {"value": "上海"},
+                        "confidence": 0.8,
+                        "source_chunk_id": "ch1",
+                    }
+                ],
+            }
         )
 
-    monkeypatch.setattr(mod, "ainvoke_json_object", fake_invoke)
+    monkeypatch.setattr(mod, "_run_enrichment_llm_async", fake_run)
 
     summaries: list[dict] = []
     facts: list[dict] = []
@@ -74,7 +84,7 @@ async def test_enrich_memory_after_ingest_async_single_llm_call(
             if model is User and key == "u1":
                 return SimpleNamespace(nickname="老王")
             if model is MemorySource and key == "src-1":
-                return SimpleNamespace(lineage_json=None)
+                return SimpleNamespace(user_id="u1", lineage_json=None)
             return None
 
         async def execute(self, _stmt):
@@ -105,10 +115,10 @@ async def test_enrich_memory_skips_when_parse_returns_none(
 
     monkeypatch.setattr("app.core.config.settings.memory_enrichment_enabled", True)
 
-    async def fake_invoke(*_args, **_kwargs):
-        return "{not json"
+    async def fake_run(*_args, **_kwargs):
+        return None
 
-    monkeypatch.setattr(mod, "ainvoke_json_object", fake_invoke)
+    monkeypatch.setattr(mod, "_run_enrichment_llm_async", fake_run)
     called = {"summary": False, "fact": False}
 
     async def capture_summary(*_args, **_kwargs):
@@ -135,7 +145,7 @@ async def test_enrich_memory_skips_when_parse_returns_none(
             if model is User and key == "u":
                 return None
             if model is MemorySource and key == "s":
-                return SimpleNamespace(lineage_json=None)
+                return SimpleNamespace(user_id="u", lineage_json=None)
             return None
 
         async def execute(self, _stmt):