feat(api): 收敛对话与记忆流程边界，引入 LLM 网关与专用服务

- MemoryService 异步路径委托 MemoryIngestService / MemoryRetrievalService；富化派发经 MemoryEnrichmentScheduler
- WebSocket pipeline 经 ChatTurnService 与显式 DTO 编排单轮对话；回忆录片段入队由 MemoirIngestScheduler 封装
- 新增 LlmGateway（LlmUseCase），各 agent、任务与适配器对齐 ports
- 补充 memory 提示适配、runtime 类型、memory-retrieval 文档、ai-touchpoints 说明与扫描脚本及配套测试

Made-with: Cursor

api/app/features/evaluation/replay_service.py

@@ -17,7 +17,7 @@ from app.features.auth import repo as auth_repo
 from app.features.conversation.models import Conversation
 from app.features.conversation.service import ConversationService
 from app.features.conversation.ws.pipeline import (
-    background_runner,
+    memoir_ingest_scheduler,
     process_user_message,
 )
 from app.features.evaluation.errors import (
@@ -160,10 +160,11 @@ class ReplayConversationService:
             segment_ids.append(segment.id)
             ts = segment.created_at or conv.last_message_at
             if not skip_memoir:
-                await background_runner.queue_message(
+                await memoir_ingest_scheduler.queue_segment(
                     conv.user_id,
                     segment.id,
                     text_char_count=len(text),
+                    trigger="evaluation_replay",
                 )
             await process_user_message(
                 conversation_id=cid,
@@ -178,7 +179,10 @@ class ReplayConversationService:
             count += 1
 
         if flush_memoir_after and conv.user_id and (not skip_memoir):
-            await background_runner.flush_pending(conv.user_id)
+            await memoir_ingest_scheduler.flush_pending(
+                conv.user_id,
+                trigger="evaluation_replay",
+            )
 
         logger.info(
             "eval replay done conversation_id={} turns={} flush={} skip_memoir={} skip_tts={}",