feat: stabilize rag chat conversations and auto title sync

fastapi_modules/fastapi_leaudit/rag_engine/config.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from fastapi_admin.config._settings import llm
+from fastapi_admin.config._settings import embedding, llm
 
 
 def _get_str(name: str, default: str = "") -> str:
@@ -36,11 +36,23 @@ RAG_CONFIG = {
     "CHROMA_PORT": _get_int("RAG_CHROMA_PORT", 8010),
     "CHROMA_TOKEN": _get_str("RAG_CHROMA_TOKEN", ""),
     "CHROMA_AUTH_HEADER": _get_str("RAG_CHROMA_AUTH_HEADER", "X-Chroma-Token"),
-    "EMBED_URL": _get_str("RAG_EMBED_URL", _get_str("GRAPH_RAG_EMBED_URL", "")),
-    "EMBED_KEY": _get_str("RAG_EMBED_KEY", _get_str("GRAPH_RAG_EMBED_KEY", "")),
-    "EMBED_MODEL": _get_str("RAG_EMBED_MODEL", _get_str("GRAPH_RAG_EMBED_MODEL", "")),
-    "EMBED_DIM": _get_int("RAG_EMBED_DIM", 1024),
-    "EMBED_BATCH_SIZE": _get_int("RAG_EMBED_BATCH_SIZE", 10),
+    "EMBED_URL": _get_str(
+        "RAG_EMBED_URL",
+        _get_str("GRAPH_RAG_EMBED_URL", _get_str("EMBEDDING_BASE_URL", embedding.EMBEDDING_BASE_URL)),
+    ),
+    "EMBED_KEY": _get_str(
+        "RAG_EMBED_KEY",
+        _get_str("GRAPH_RAG_EMBED_KEY", _get_str("EMBEDDING_API_KEY", embedding.EMBEDDING_API_KEY)),
+    ),
+    "EMBED_MODEL": _get_str(
+        "RAG_EMBED_MODEL",
+        _get_str("GRAPH_RAG_EMBED_MODEL", _get_str("EMBEDDING_MODEL", embedding.EMBEDDING_MODEL)),
+    ),
+    "EMBED_DIM": _get_int("RAG_EMBED_DIM", _get_int("EMBEDDING_DIM", embedding.EMBEDDING_DIM)),
+    "EMBED_BATCH_SIZE": _get_int(
+        "RAG_EMBED_BATCH_SIZE",
+        _get_int("EMBEDDING_BATCH_SIZE", embedding.EMBEDDING_BATCH_SIZE),
+    ),
     "RERANKER_URL": _get_str("RAG_RERANKER_URL", _get_str("GRAPH_RAG_RERANKER_URL", "")),
     "RERANKER_KEY": _get_str("RAG_RERANKER_KEY", _get_str("GRAPH_RAG_RERANKER_KEY", "")),
     "RERANKER_MODEL": _get_str("RAG_RERANKER_MODEL", _get_str("GRAPH_RAG_RERANKER_MODEL", "")),
@@ -58,3 +70,34 @@ RAG_CONFIG = {
     "HYBRID_SEARCH": _get_bool("RAG_HYBRID_SEARCH", True),
     "RERANKING": _get_bool("RAG_RERANKING", True),
 }
+
+
+def build_openai_chat_completions_url(base_url: str) -> str:
+    normalized = (base_url or "").strip().rstrip("/")
+    if not normalized:
+        return "/chat/completions"
+    if normalized.endswith("/chat/completions"):
+        return normalized
+    return f"{normalized}/chat/completions"
+
+
+def build_openai_embeddings_url(base_url: str) -> str:
+    normalized = (base_url or "").strip().rstrip("/")
+    if not normalized:
+        return "/embeddings"
+    if normalized.endswith("/chat/completions"):
+        normalized = normalized[:-len("/chat/completions")]
+    if normalized.endswith("/embeddings"):
+        return normalized
+    return f"{normalized}/embeddings"
+
+
+def normalize_openai_base_url(base_url: str) -> str:
+    normalized = (base_url or "").strip().rstrip("/")
+    if not normalized:
+        return ""
+    if normalized.endswith("/chat/completions"):
+        return normalized[:-len("/chat/completions")]
+    if normalized.endswith("/embeddings"):
+        return normalized[:-len("/embeddings")]
+    return normalized

fastapi_modules/fastapi_leaudit/rag_engine/generator.py

@@ -7,7 +7,7 @@ from typing import AsyncGenerator
 
 import httpx
 
-from fastapi_modules.fastapi_leaudit.rag_engine.config import RAG_CONFIG
+from fastapi_modules.fastapi_leaudit.rag_engine.config import RAG_CONFIG, build_openai_chat_completions_url
 
 DEFAULT_SYSTEM_PROMPT = """你是烟草行业智慧法务小助手，专注于烟草专卖法规、合同管理、行政处罚等相关法律法规。\n\n回答要求：\n- 先用一句话直接回答，再展开详细说明\n- 多个要点用编号列表\n- 关键法条和数字用 **加粗**\n- 分类信息用表格\n- 层级结构用缩进子列表\n- 不要加标题，直接输出正文"""
 
@@ -17,13 +17,14 @@ async def generate_stream(
     context_chunks: list[dict],
     conversation_id: str,
     message_id: str,
+    task_id: str | None = None,
     system_prompt: str = "",
     model: str = "",
     temperature: float | None = None,
     max_tokens: int | None = None,
     dataset_name: str = "",
 ) -> AsyncGenerator[str, None]:
-    task_id = str(uuid.uuid4())
+    task_id = task_id or str(uuid.uuid4())
     created_at = int(time.time())
     _model = model or RAG_CONFIG["LLM_MODEL"]
     _temp = temperature if temperature is not None else RAG_CONFIG["LLM_TEMPERATURE"]
@@ -55,7 +56,7 @@ async def generate_stream(
         async with httpx.AsyncClient(timeout=RAG_CONFIG["LLM_TIMEOUT"]) as client:
             async with client.stream(
                 "POST",
-                f"{RAG_CONFIG['LLM_BASE_URL'].rstrip('/')}" + "/chat/completions",
+                build_openai_chat_completions_url(RAG_CONFIG["LLM_BASE_URL"]),
                 json={
                     "model": _model,
                     "messages": messages,

fastapi_modules/fastapi_leaudit/rag_engine/question_chains.py

@@ -4,7 +4,7 @@ import json
 
 import httpx
 
-from fastapi_modules.fastapi_leaudit.rag_engine.config import RAG_CONFIG
+from fastapi_modules.fastapi_leaudit.rag_engine.config import RAG_CONFIG, build_openai_chat_completions_url
 
 
 async def generate_followups(query: str, answer: str) -> list[str]:
@@ -15,7 +15,7 @@ async def generate_followups(query: str, answer: str) -> list[str]:
     )
     async with httpx.AsyncClient(timeout=30.0) as client:
         resp = await client.post(
-            f"{RAG_CONFIG['LLM_BASE_URL'].rstrip('/')}" + "/chat/completions",
+            build_openai_chat_completions_url(RAG_CONFIG["LLM_BASE_URL"]),
             json={
                 "model": RAG_CONFIG["LLM_MODEL"],
                 "messages": [{"role": "user", "content": prompt}],