feat(api)!: memory single chain — async MemoryService, strict eval closure

Route all memory ingest/retrieve/enrichment/compaction through async MemoryService.
Remove legacy sync memory implementations (ingest/retrieve/compaction); Celery and
memoir Phase2 call asyncio.run into MemoryService-backed helpers.

Memoir Phase1 batch ingest uses MemoryService.ingest_transcripts_batch; drop chapters.
evidence_bundle_json mirror (Alembic 0015). Evaluation uses snapshot/link-only bundles;
raise EvidenceClosureMissing instead of partial/fallback lineage tiers.

Split memoir state into NarrativeCoverageState and InterviewControlState; delete the
_interview_meta_store adapter layer. Remove rolling-query and recent-fact fallback
settings from config and evidence assembly.

Update judges, docs, tests, and PlaygroundPage alignment.

Made-with: Cursor

api/tests/test_eval_judge_llm_spec.py

@@ -42,6 +42,20 @@ def test_build_eval_judge_deepseek_requires_key(
     assert build_eval_judge_llm_spec("deepseek", None) is None
 
 
+def test_build_eval_judge_deepseek_v4_flash_non_thinking_default_path(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    """默认 deepseek-v4-flash 且关闭 thinking 时显式传 disabled（避免 API 默认 enabled）。"""
+    monkeypatch.setattr(settings, "deepseek_api_key", "d-test")
+    monkeypatch.setattr(settings, "eval_judge_deepseek_model", "deepseek-v4-flash")
+    monkeypatch.setattr(settings, "eval_judge_deepseek_thinking_enabled", False)
+    spec = build_eval_judge_llm_spec("deepseek", None)
+    assert spec is not None
+    assert spec.resolved_model == "deepseek-v4-flash"
+    assert spec.llm.extra_body == {"thinking": {"type": "disabled"}}
+    assert spec.llm.reasoning_effort is None
+
+
 def test_build_eval_judge_deepseek_context_budget(
     monkeypatch: pytest.MonkeyPatch,
 ) -> None: