From d7afc218b1ab2a5279e35de52b536efda10fc34b Mon Sep 17 00:00:00 2001 From: Claude-51124 Date: Tue, 23 Dec 2025 18:34:47 +0900 Subject: [PATCH] =?UTF-8?q?docs:=20=EA=B3=84=ED=9A=8D=20=EB=AC=B8=EC=84=9C?= =?UTF-8?q?=20=EC=88=98=EC=A0=95=20-=20=EA=B5=AC=ED=98=84=20=EB=B0=A9?= =?UTF-8?q?=EB=B2=95=20=EC=A0=9C=EA=B1=B0,=20=ED=95=84=EC=9A=94=20?= =?UTF-8?q?=EC=9E=91=EC=97=85=EB=A7=8C=20=EB=AA=85=EC=8B=9C?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- ...1223_짧은_후속_질문_LLM_우선_해결_계획.md | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/journey/plans/251223_짧은_후속_질문_LLM_우선_해결_계획.md b/journey/plans/251223_짧은_후속_질문_LLM_우선_해결_계획.md index 040baba..3f1b78e 100644 --- a/journey/plans/251223_짧은_후속_질문_LLM_우선_해결_계획.md +++ b/journey/plans/251223_짧은_후속_질문_LLM_우선_해결_계획.md @@ -27,20 +27,19 @@ ### 1. LLM 질문 확장 - 짧은 질문(len <= 10 또는 확신도 < 0.7) → LLM이 맥락 포함 완전한 질문으로 확장 -- 구현: `llm_service.py:expand_question()`, `message_service.py:206-228` 수정 -- Few-shot 예시로 정확도 향상 +- 필요 작업: `llm_service.py`에 질문 확장 메서드 추가, `message_service.py`에서 호출 ### 2. LLM 의도 분류 강화 - 확신도 < 0.7 → LLM이 맥락 포함하여 재분류 -- 구현: `intent_graph.py`, `intent_parser.py` 맥락 포함 분류 추가 +- 필요 작업: `intent_graph.py`, `intent_parser.py` 맥락 포함 분류 추가 ### 3. LangGraph 워크플로우 - 질문 확장 → 의도 분류 → 엔티티 추출 → 스킬 선택 -- 구현: `intent_graph_langgraph.py` (신규) +- 필요 작업: LangGraph 기반 의도 분류 워크플로우 구현 ### 4. 세션 맥락 통합 - 세션 상태를 LLM에 전달하여 맥락 유지 -- 구현: `session_manager.py`, `message_service.py` 세션 맥락 포함 +- 필요 작업: `session_manager.py`, `message_service.py` 세션 맥락 포함 ---