Feature/followup latency

ai/src/ai_server/config/settings.py

@@ -38,9 +38,8 @@ class Settings(BaseSettings):
     ai_realtime_exchange: str = "stackup.realtime"
     ai_realtime_routing_user: str = "realtime.user.notify"
     feedback_rag_top_k: int = 5
-    # 리랭킹: 하이브리드 검색으로 후보 N개(candidate_k)를 가져와 LLM 으로 재정렬 후 top_k 주입.
-    rerank_enabled: bool = True
-    rerank_candidate_k: int = 20
+    # RAG 컨텍스트 구성(임베딩+검색) 전체 상한. 초과 시 (none) 으로 폴백해 첫 토큰을 막지 않는다.
+    followup_rag_timeout_sec: float = 1.5
     # 질문 풀 초기 크기. Core 의 applyPool 이 questions[0] 만 INSERT 하므로 1 로 고정해 토큰 낭비 차단.
     # 후속 작업에서 풀 저장 도입 시 늘리기 (예: 5).
     questions_initial_pool_size: int = 1

ai/src/ai_server/messaging/consumers/feedback_consumer.py

@@ -15,7 +15,6 @@
     VoiceAnalysisSummary,
 )
 from ai_server.rag.embedder import EmbeddingProvider
-from ai_server.rag.reranker import NoopReranker, Reranker, rerank_hits
 
 log = structlog.get_logger(__name__)
 
@@ -41,8 +40,6 @@ def __init__(
         core_client: CoreClient,
         embedder: EmbeddingProvider | None = None,
         rag_top_k: int = 5,
-        reranker: Reranker | None = None,
-        candidate_k: int = 20,
     ) -> None:
         self._generator = generator
         self._publisher = publisher
@@ -51,8 +48,6 @@ def __init__(
         self._core = core_client
         self._embedder = embedder
         self._rag_top_k = rag_top_k
-        self._reranker = reranker or NoopReranker()
-        self._candidate_k = max(candidate_k, rag_top_k)
 
     async def handle(self, message: AbstractIncomingMessage) -> None:
         async with message.process(requeue=False):
@@ -148,16 +143,13 @@ async def _build_rag_context(self, req: GenerateFeedbackRequest) -> str:
                 query_embedding=query_vec,
                 query_text=last_answer,
                 document_ids=req.context_document_ids,
-                top_k=self._candidate_k,
+                top_k=self._rag_top_k,
             )
         except Exception as exc:
             log.warn("feedback.rag.failed", error=str(exc), session_id=req.session_id)
             return "(none)"
         if not hits:
             return "(none)"
-        hits = await rerank_hits(
-            self._reranker, query=last_answer, hits=hits, top_k=self._rag_top_k
-        )
         return "\n---\n".join(
             f"[doc#{h.document_id} chunk#{h.chunk_index}] {h.chunk_text}" for h in hits
         )

ai/src/ai_server/messaging/consumers/followup_consumer.py

@@ -20,7 +20,6 @@
     GenerateFollowupRequest,
 )
 from ai_server.rag.embedder import EmbeddingProvider
-from ai_server.rag.reranker import NoopReranker, Reranker, rerank_hits
 
 log = structlog.get_logger(__name__)
 
@@ -43,13 +42,12 @@ def __init__(
         core_client: CoreClient | None = None,
         embedder: EmbeddingProvider | None = None,
         rag_top_k: int = 5,
-        reranker: Reranker | None = None,
-        candidate_k: int = 20,
         streaming_generator: StreamingFollowupGenerator | None = None,
         session_notifier: SessionRealtimeNotifier | None = None,
         tts=None,
         storage=None,
         tts_voice: str = "",
+        rag_timeout_sec: float = 1.5,
     ) -> None:
         self._generator = generator
         self._publisher = publisher
@@ -58,13 +56,12 @@ def __init__(
         self._core = core_client
         self._embedder = embedder
         self._rag_top_k = rag_top_k
-        self._reranker = reranker or NoopReranker()
-        self._candidate_k = max(candidate_k, rag_top_k)
         self._streaming = streaming_generator
         self._notifier = session_notifier
         self._tts = tts
         self._storage = storage
         self._tts_voice = tts_voice
+        self._rag_timeout_sec = rag_timeout_sec
 
     async def handle(self, message: AbstractIncomingMessage) -> None:
         async with message.process(requeue=False):
@@ -233,7 +230,19 @@ async def _synth_segment(
     async def _build_rag_context(self, req: GenerateFollowupRequest) -> str:
         if not self._core or not self._embedder or not req.context_document_ids:
             return "(none)"
+        try:
+            return await asyncio.wait_for(
+                self._do_build_rag_context(req), timeout=self._rag_timeout_sec
+            )
+        except asyncio.TimeoutError:
+            log.warning(
+                "followup.rag.timeout",
+                session_id=req.session_id,
+                timeout_sec=self._rag_timeout_sec,
+            )
+            return "(none)"
 
+    async def _do_build_rag_context(self, req: GenerateFollowupRequest) -> str:
         query = f"{req.previous_question}\n\n{req.answer_text}"
         try:
             query_vec = (
@@ -243,17 +252,14 @@ async def _build_rag_context(self, req: GenerateFollowupRequest) -> str:
                 query_embedding=query_vec,
                 query_text=query,
                 document_ids=req.context_document_ids,
-                top_k=self._candidate_k,
+                top_k=self._rag_top_k,
             )
         except Exception as exc:
             log.warn("followup.rag.failed", error=str(exc), session_id=req.session_id)
             return "(none)"
 
         if not hits:
             return "(none)"
-        hits = await rerank_hits(
-            self._reranker, query=query, hits=hits, top_k=self._rag_top_k
-        )
         return "\n---\n".join(
             f"[doc#{h.document_id} chunk#{h.chunk_index}] {h.chunk_text}" for h in hits
         )

ai/src/ai_server/messaging/consumers/questions_consumer.py

@@ -14,7 +14,6 @@
     QuestionPoolCallbackPayload,
 )
 from ai_server.rag.embedder import EmbeddingProvider
-from ai_server.rag.reranker import NoopReranker, Reranker, rerank_hits
 
 log = structlog.get_logger(__name__)
 
@@ -31,8 +30,6 @@ def __init__(
         core_client: CoreClient | None = None,
         embedder: EmbeddingProvider | None = None,
         rag_top_k: int = 5,
-        reranker: Reranker | None = None,
-        candidate_k: int = 20,
     ) -> None:
         self._generator = generator
         self._publisher = publisher
@@ -44,8 +41,6 @@ def __init__(
         self._core = core_client
         self._embedder = embedder
         self._rag_top_k = rag_top_k
-        self._reranker = reranker or NoopReranker()
-        self._candidate_k = max(candidate_k, rag_top_k)
 
     async def handle(self, message: AbstractIncomingMessage) -> None:
         async with message.process(requeue=False):
@@ -138,17 +133,14 @@ async def _build_context(self, req: GenerateQuestionsRequest) -> str:
                 query_embedding=query_vec,
                 query_text=query,
                 document_ids=document_ids,
-                top_k=self._candidate_k,
+                top_k=self._rag_top_k,
             )
         except Exception as exc:
             log.warn("questions.rag.failed", error=str(exc), session_id=req.session_id)
             return base_context
 
         if not hits:
             return base_context
-        hits = await rerank_hits(
-            self._reranker, query=query, hits=hits, top_k=self._rag_top_k
-        )
         rag_context = "\n---\n".join(
             f"[doc#{h.document_id} chunk#{h.chunk_index}] {h.chunk_text}" for h in hits
         )

ai/src/ai_server/messaging/runner.py

@@ -32,7 +32,6 @@
 from ai_server.messaging.connection import RabbitConnection
 from ai_server.rag.chunker import MarkdownChunker
 from ai_server.rag.embedder import build_embedding_provider
-from ai_server.rag.reranker import build_reranker
 from ai_server.messaging.consumers.feedback_consumer import FeedbackConsumer
 from ai_server.messaging.consumers.followup_consumer import FollowupConsumer
 from ai_server.messaging.consumers.questions_consumer import QuestionsConsumer
@@ -102,7 +101,6 @@ def __init__(self, settings: Settings) -> None:
             max_retries=settings.embedding_max_retries,
             retry_base_delay_sec=settings.embedding_retry_base_delay_sec,
         )
-        reranker = build_reranker(settings, core_client=core_client)
         vision_pdf_reader = build_vision_pdf_reader(settings, core_client=core_client)
 
         # 이력서 PDF
@@ -182,8 +180,6 @@ def __init__(self, settings: Settings) -> None:
             initial_pool_size=settings.questions_initial_pool_size,
             core_client=core_client,
             embedder=embedder,
-            reranker=reranker,
-            candidate_k=settings.rerank_candidate_k,
         )
 
         # 꼬리질문 생성 (US-19)
@@ -206,13 +202,12 @@ def __init__(self, settings: Settings) -> None:
             callback_routing_key=settings.ai_callback_routing_questions,
             core_client=core_client,
             embedder=embedder,
-            reranker=reranker,
-            candidate_k=settings.rerank_candidate_k,
             streaming_generator=streaming_followup_generator,
             session_notifier=session_notifier,
             tts=tts,
             storage=storage,
             tts_voice=settings.openai_tts_voice,
+            rag_timeout_sec=settings.followup_rag_timeout_sec,
         )
 
         # 종합 피드백 생성 (US-24)
@@ -226,8 +221,6 @@ def __init__(self, settings: Settings) -> None:
             callback_routing_key=settings.ai_callback_routing_feedback,
             core_client=core_client,
             embedder=embedder,
-            reranker=reranker,
-            candidate_k=settings.rerank_candidate_k,
             rag_top_k=settings.feedback_rag_top_k,
         )