feat: stabilize rag chat conversations and auto title sync

2026-05-19 15:52:05 +08:00
parent 564f2bebc8
commit afaba4dd99
19 changed files with 1988 additions and 93 deletions
@@ -38,6 +38,7 @@ from fastapi_admin.config import (
    LEAUDIT_LLM_RETRY_BACKOFF_BASE_SECONDS,
 )
 from fastapi_modules.fastapi_leaudit.govdoc_engine.llm.cache import LlmCache, make_key
+from fastapi_modules.fastapi_leaudit.rag_engine.config import normalize_openai_base_url

 _log = logging.getLogger(__name__)

@@ -153,8 +154,9 @@ class LlmClient:
                "LLM_API_KEY is not configured. Set LLM_API_KEY in platform config."
            )
        else:
-            self._client = OpenAI(api_key=key, base_url=base_url or LLM_BASE_URL)
-            self._aclient = AsyncOpenAI(api_key=key, base_url=base_url or LLM_BASE_URL)
+            normalized_base_url = normalize_openai_base_url(base_url or LLM_BASE_URL)
+            self._client = OpenAI(api_key=key, base_url=normalized_base_url)
+            self._aclient = AsyncOpenAI(api_key=key, base_url=normalized_base_url)
        self.model = model or LLM_MODEL
        self.timeout = timeout_seconds if timeout_seconds is not None else LEAUDIT_LLM_REQUEST_TIMEOUT
        self.max_retries = max_retries if max_retries is not None else LEAUDIT_LLM_RETRY_MAX_ATTEMPTS