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/memoir/ingest_scheduler.py

@@ -0,0 +1,77 @@
+"""Memoir ingest scheduling boundary.
+
+The real batching logic still lives in ``BackgroundTaskRunner``. This adapter
+keeps conversation code from depending on that implementation directly.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Literal, Sequence
+
+from app.features.memoir.background_runner import BackgroundTaskRunner
+
+MemoirTrigger = Literal[
+    "turn",
+    "conversation_end",
+    "manual_flush",
+    "evaluation_replay",
+]
+
+
+@dataclass(frozen=True)
+class MemoirPhasePlan:
+    """A visible plan for submitting segments into the memoir pipeline."""
+
+    user_id: str
+    segment_ids: tuple[str, ...]
+    trigger: MemoirTrigger
+
+
+class MemoirIngestScheduler:
+    """Small facade over debounce batching and Phase2 flush dispatch."""
+
+    def __init__(self, runner: BackgroundTaskRunner | None = None) -> None:
+        self._runner = runner or BackgroundTaskRunner()
+
+    @property
+    def runner(self) -> BackgroundTaskRunner:
+        """Compatibility escape hatch for existing tests and eval utilities."""
+
+        return self._runner
+
+    async def queue_segment(
+        self,
+        user_id: str,
+        segment_id: str,
+        *,
+        text_char_count: int = 0,
+        trigger: MemoirTrigger = "turn",
+    ) -> MemoirPhasePlan:
+        await self._runner.queue_message(
+            user_id,
+            segment_id,
+            text_char_count=text_char_count,
+        )
+        return MemoirPhasePlan(
+            user_id=user_id,
+            segment_ids=(segment_id,),
+            trigger=trigger,
+        )
+
+    async def flush_pending(
+        self,
+        user_id: str,
+        *,
+        extra_segment_ids: Sequence[str] | None = None,
+        trigger: MemoirTrigger = "manual_flush",
+    ) -> tuple[MemoirPhasePlan, str | None]:
+        ids = tuple(str(x) for x in (extra_segment_ids or ()) if str(x).strip())
+        task_id = await self._runner.flush_pending(
+            user_id,
+            extra_segment_ids=list(ids),
+        )
+        return MemoirPhasePlan(user_id=user_id, segment_ids=ids, trigger=trigger), task_id
+
+
+__all__ = ["MemoirIngestScheduler", "MemoirPhasePlan", "MemoirTrigger"]