feat: 重构知识库系统，移除Hermes集成，增强RAG和同步功能

主要变更:
- 移除Hermes智能体及相关回调服务
- 新增知识库RAG、同步、调度、规范化和索引任务服务
- 重构orchestrator服务，增强运行时聊天功能
- 更新前端聊天、政策制度、设置等页面样式和逻辑
- 更新expense_claims和document_intelligence服务
- 删除llm_wiki相关服务和测试文件
- 更新docker-compose配置和启动脚本

server/src/app/services/model_connectivity.py

@@ -59,9 +59,18 @@ def _probe_openai_compatible(payload: ModelConnectivityTestRequest) -> int:
     normalized_endpoint = _normalize_endpoint(payload.endpoint)
     headers = _build_headers(api_key=payload.api_key, use_bearer=True)
 
-    if payload.capability == "embedding":
+    if payload.capability == "reranker" and payload.provider == "Ali":
+        url, body = _build_ali_reranker_request(payload.model, normalized_endpoint)
+    elif payload.capability == "embedding":
         url = _ensure_path(normalized_endpoint, "embeddings")
         body = {"model": payload.model, "input": "connectivity test"}
+    elif payload.capability == "reranker":
+        url = _ensure_path(normalized_endpoint, "rerank")
+        body = {
+            "model": payload.model,
+            "query": "connectivity test",
+            "documents": ["sample document"],
+        }
     else:
         url = _ensure_path(normalized_endpoint, "chat/completions")
         body = {
@@ -74,6 +83,35 @@ def _probe_openai_compatible(payload: ModelConnectivityTestRequest) -> int:
     return status_code
 
 
+def _build_ali_reranker_request(model: str, endpoint: str) -> tuple[str, dict[str, Any]]:
+    normalized_model = str(model or "").strip()
+    if normalized_model == "qwen3-rerank":
+        return (
+            "https://dashscope.aliyuncs.com/compatible-api/v1/reranks",
+            {
+                "model": normalized_model,
+                "query": "connectivity test",
+                "documents": ["sample document"],
+                "top_n": 1,
+            },
+        )
+
+    return (
+        "https://dashscope.aliyuncs.com/api/v1/services/rerank/text-rerank/text-rerank",
+        {
+            "model": normalized_model,
+            "input": {
+                "query": "connectivity test",
+                "documents": ["sample document"],
+            },
+            "parameters": {
+                "return_documents": False,
+                "top_n": 1,
+            },
+        },
+    )
+
+
 def _probe_ollama(payload: ModelConnectivityTestRequest) -> int:
     normalized_endpoint = _normalize_endpoint(payload.endpoint)
     headers = _build_headers(api_key=payload.api_key, use_bearer=False)
@@ -81,6 +119,8 @@ def _probe_ollama(payload: ModelConnectivityTestRequest) -> int:
     if payload.capability == "embedding":
         url = _ensure_path(normalized_endpoint, "api/embed")
         body = {"model": payload.model, "input": "connectivity test"}
+    elif payload.capability == "reranker":
+        raise ConnectivityCheckError("Ollama 暂不支持 reranker 连通性探测。", status_code=HTTPStatus.BAD_REQUEST)
     else:
         url = _ensure_path(normalized_endpoint, "api/chat")
         body = {
@@ -100,6 +140,12 @@ def _probe_azure_openai(payload: ModelConnectivityTestRequest) -> int:
     if payload.capability == "embedding":
         url = f"{deployment_base}/embeddings?api-version={AZURE_API_VERSION}"
         body = {"input": "connectivity test"}
+    elif payload.capability == "reranker":
+        url = f"{deployment_base}/rerank?api-version={AZURE_API_VERSION}"
+        body = {
+            "query": "connectivity test",
+            "documents": ["sample document"],
+        }
     else:
         url = f"{deployment_base}/chat/completions?api-version={AZURE_API_VERSION}"
         body = {
@@ -168,12 +214,13 @@ def _send_json_request(
     *,
     headers: dict[str, str],
     payload: dict[str, Any],
+    timeout_seconds: int = DEFAULT_TIMEOUT_SECONDS,
 ) -> tuple[int, Any]:
     data = json.dumps(payload).encode("utf-8")
     request = Request(url=url, data=data, headers=headers, method=method)
 
     try:
-        with urlopen(request, timeout=DEFAULT_TIMEOUT_SECONDS) as response:
+        with urlopen(request, timeout=timeout_seconds) as response:
             body = response.read().decode("utf-8") if response.length != 0 else ""
             return response.status, _parse_json_body(body)
     except HTTPError as exc: