feat(memory,conversation): 记忆富化/证据包、时间线幂等字段与对话分段全链路

数据库
- 新增迁移 0003：timeline_events.memory_source_id 外键 → memory_sources，便于按 ingest 源做时间线幂等

后端 - 记忆
- 新增 ingest 后 LLM 富化（摘要/事实/时间线），可配置开关与最大字符数
- 新增证据包组装：合并 chunk、摘要、事实、时间线、故事等检索结果；支持空 query 时是否仍带 rolling 等开关
- repo/retriever/service/router/schemas/summarizer/timeline/extractor 等扩展；文档 memory-retrieval.md 更新

后端 - 对话 WS
- 增加 PING/PONG；分段 ASR 日志与空音频处理；转写失败与「无助手回复」错误提示更明确
- 助手多段回复持久化使用统一分隔符，与分段逻辑一致

后端 - Agent
- reply_limits：按 [SPLIT] 与段落拆段，并保证非空 fallback，供 WS 与 TTS 多段下发

后端 - 回忆录任务
- transcript ingest 记录 source_id；任务成功结?

api/app/features/memory/retriever.py

@@ -2,12 +2,8 @@
 
 from sqlalchemy.ext.asyncio import AsyncSession
 
-from app.features.memory.repo import (
-    get_facts_for_user,
-    get_timeline_events_for_user,
-    search_chunks_fts,
-    search_chunks_vector,
-)
+from app.features.memory.evidence import retrieve_evidence_bundle_async
+from app.features.memory.repo import search_chunks_fts, search_chunks_vector
 from app.ports.embedding import EmbeddingProvider
 
 
@@ -44,24 +40,31 @@ class HybridRetriever:
         """
         Return evidence bundle:
         {relevant_chunks, relevant_summaries, relevant_facts, timeline_hints, relevant_stories}
-
-        `relevant_summaries` / `relevant_stories` 当前多为占位空列表；叙事 prompt 仅应依赖
-        已实现填充的字段（见 `format_evidence_chunks_for_prompt`）。
         """
+        if not query.strip():
+            return await retrieve_evidence_bundle_async(
+                self._db,
+                user_id,
+                query,
+                top_k=top_k,
+                merged_chunk_dicts=[],
+            )
+
+        q = query.strip()
         fts_chunks = await search_chunks_fts(
             self._db, user_id=user_id, query=query, limit=top_k * 2
         )
 
         vector_chunks: list[dict] = []
-        if self._embedding and query.strip():
-            q_emb = await self._embedding.embed_text(query.strip())
+        if self._embedding and q:
+            q_emb = await self._embedding.embed_text(q)
             if q_emb:
                 vector_chunks = await search_chunks_vector(
                     self._db, user_id=user_id, query_embedding=q_emb, limit=top_k * 2
                 )
 
         merged = _rrf_merge(fts_chunks, vector_chunks)[:top_k]
-        relevant_chunks = [
+        merged_chunk_dicts = [
             {
                 "id": c["id"],
                 "content": c["content"],
@@ -70,36 +73,10 @@ class HybridRetriever:
             for c in merged
         ]
 
-        facts = await get_facts_for_user(self._db, user_id=user_id, limit=top_k)
-        relevant_facts = [
-            {
-                "id": f.id,
-                "fact_type": f.fact_type,
-                "subject": f.subject,
-                "predicate": f.predicate,
-                "object_json": f.object_json,
-            }
-            for f in facts
-        ]
-
-        events = await get_timeline_events_for_user(
-            self._db, user_id=user_id, limit=top_k
+        return await retrieve_evidence_bundle_async(
+            self._db,
+            user_id,
+            query,
+            top_k=top_k,
+            merged_chunk_dicts=merged_chunk_dicts,
         )
-        timeline_hints = [
-            {
-                "id": e.id,
-                "event_year": e.event_year,
-                "event_date": e.event_date,
-                "title": e.title,
-                "description": e.description,
-            }
-            for e in events
-        ]
-
-        return {
-            "relevant_chunks": relevant_chunks,
-            "relevant_summaries": [],
-            "relevant_facts": relevant_facts,
-            "timeline_hints": timeline_hints,
-            "relevant_stories": [],
-        }