Merge remote-tracking branch 'origin/CODEX' into sub_laptop_3

B01_Dashboard/B01_Dashboard_Knowledge_Chat.py

@@ -1,10 +1,13 @@
 """지식DB Markdown 검색과 Gemini 설명. 제외 폴더는 읽지 않는다."""
 
 import asyncio
+import logging
 import os
 import re
+import time
 from functools import lru_cache
 from pathlib import Path
+from uuid import uuid4
 
 import httpx
 from fastapi import APIRouter, Depends, HTTPException
@@ -16,6 +19,7 @@ router = APIRouter(prefix="/knowledge-chat", tags=["B01_KnowledgeChat"])
 KNOWLEDGE = Path(__file__).resolve().parents[1] / "resources" / "knowledge"
 EXCLUDED = ("실무문서", "stmate", "다산소프트", "technical_info")
 MODEL = "gemini-3.8-flash"
+logger = logging.getLogger(__name__)
 
 
 class ChatRequest(BaseModel):
@@ -149,21 +153,36 @@ async def ask_knowledge_chat(payload: ChatRequest, _session=Depends(verify_sessi
             f"질문: {payload.question}"
         )
     )
+    request_id = uuid4().hex[:8]
+    started = time.monotonic()
     try:
         async with httpx.AsyncClient(timeout=30) as client:
             for attempt in range(3):
                 response = await client.post(
                     f"https://generativelanguage.googleapis.com/v1beta/models/{MODEL}:generateContent",
                     headers={"x-goog-api-key": key},
-                    json={"contents": [{"parts": [{"text": prompt}]}]},
+                    json={
+                        "contents": [{"parts": [{"text": prompt}]}],
+                        "generationConfig": {"thinkingConfig": {"thinkingLevel": "low"}},
+                    },
                 )
+                if response.status_code >= 400:
+                    logger.warning(
+                        "Gemini request failed id=%s model=%s attempt=%d status=%d elapsed=%.1fs prompt_chars=%d",
+                        request_id,
+                        MODEL,
+                        attempt + 1,
+                        response.status_code,
+                        time.monotonic() - started,
+                        len(prompt),
+                    )
                 if response.status_code != 503 or attempt == 2:
                     break
                 await asyncio.sleep(attempt + 1)
             if response.status_code == 503:
                 raise HTTPException(
                     status_code=503,
-                    detail="Gemini 모델에 요청이 몰려 있습니다. 잠시 후 다시 질문해 주세요.",
+                    detail=f"Gemini 모델에 요청이 몰려 있습니다. 잠시 후 다시 질문해 주세요. (문의번호 {request_id})",
                 )
             response.raise_for_status()
             answer = "".join(
@@ -180,12 +199,21 @@ async def ask_knowledge_chat(payload: ChatRequest, _session=Depends(verify_sessi
         }.get(exc.response.status_code, "Gemini 요청에 실패했습니다. 잠시 후 다시 시도해 주세요.")
         raise HTTPException(status_code=502, detail=detail) from exc
     except httpx.TimeoutException as exc:
+        logger.warning(
+            "Gemini request timed out id=%s elapsed=%.1fs", request_id, time.monotonic() - started
+        )
         raise HTTPException(
             status_code=504,
             detail="Gemini 3.8 Flash 응답이 지연되어 시간 초과됐습니다. "
             "잠시 후 다시 시도해 주세요.",
         ) from exc
     except (httpx.RequestError, ValueError) as exc:
+        logger.warning(
+            "Gemini request error id=%s type=%s elapsed=%.1fs",
+            request_id,
+            type(exc).__name__,
+            time.monotonic() - started,
+        )
         raise HTTPException(
             status_code=502,
             detail="Gemini 서버에 연결하지 못했습니다. 잠시 후 다시 시도해 주세요.",

resources/tester/test_b01_knowledge_chat.py

@@ -36,7 +36,7 @@ def test_missing_api_key_is_reported_before_search(monkeypatch):
     assert error.value.status_code == 503
 
 
-def test_busy_gemini_is_retried_and_reported(monkeypatch):
+def test_busy_gemini_is_retried_and_reported(monkeypatch, caplog):
     monkeypatch.setenv("GEMINI_API_KEY", "test-key")
     monkeypatch.setattr(chat, "search", lambda _question: [])
     calls = []
@@ -64,6 +64,9 @@ def test_busy_gemini_is_retried_and_reported(monkeypatch):
         asyncio.run(chat.ask_knowledge_chat(chat.ChatRequest(question="일위대가가 뭐야"), {}))
     assert len(calls) == 3
     assert "요청이 몰려" in error.value.detail
+    assert [f"attempt={n} status=503" in caplog.text for n in (1, 2, 3)] == [True] * 3
+    assert "test-key" not in caplog.text
+    assert "일위대가가 뭐야" not in caplog.text
 
 
 def test_gemini_timeout_is_reported(monkeypatch):
@@ -132,11 +135,13 @@ def test_chat_searches_all_knowledge_markdown_except_excluded_folders(monkeypatc
 
         async def post(self, url, *, headers, json):
             sent["prompt"] = json["contents"][0]["parts"][0]["text"]
+            sent["thinking_level"] = json["generationConfig"]["thinkingConfig"]["thinkingLevel"]
             return Response()
 
     monkeypatch.setattr(chat.httpx, "AsyncClient", Client)
     result = asyncio.run(chat.ask_knowledge_chat(chat.ChatRequest(question="임도 기준"), {}))
     assert result["answer"] == "설명 [1]"
+    assert sent["thinking_level"] == "low"
     paths = {source["path"] for source in result["sources"]}
     assert paths == {"original/법률/현행.md"}
     assert "실무문서" not in sent["prompt"]