feat: enhance error logging and upstream response tracking for provider failures

- Introduce ProviderCompatibilityException for unsupported parameter/feature errors
- Add COMPATIBILITY_ERROR_PATTERNS to detect provider-specific limitations
- Implement _is_compatibility_error() method in ErrorClassifier
- Prioritize compatibility error checking before client error validation
- Remove 'max_tokens' from CLIENT_ERROR_PATTERNS as it can indicate compatibility issues
- Enable automatic failover when provider doesn't support requested features
- Improve error classification accuracy with pattern matching for common compatibility issues

src/api/handlers/base/chat_handler_base.py

@@ -639,6 +639,8 @@ class ChatHandlerBase(BaseMessageHandler, ABC):
 
             logger.info(f"  [{self.request_id}] 发送非流式请求: Provider={provider.name}, "
                 f"模型={model} -> {mapped_model or '无映射'}")
+            logger.debug(f"  [{self.request_id}] 请求URL: {url}")
+            logger.debug(f"  [{self.request_id}] 请求体stream字段: {provider_payload.get('stream', 'N/A')}")
 
             # 创建 HTTP 客户端（支持代理配置）
             from src.clients.http_client import HTTPClientPool
@@ -662,10 +664,32 @@ class ChatHandlerBase(BaseMessageHandler, ABC):
                         response_headers=response_headers,
                     )
                 elif resp.status_code >= 500:
-                    raise ProviderNotAvailableException(f"提供商服务不可用: {provider.name}")
-                elif resp.status_code != 200:
+                    # 记录响应体以便调试
+                    error_body = ""
+                    try:
+                        error_body = resp.text[:1000]
+                        logger.error(f"  [{self.request_id}] 上游返回5xx错误: status={resp.status_code}, body={error_body[:500]}")
+                    except Exception:
+                        pass
                     raise ProviderNotAvailableException(
-                        f"提供商返回错误: {provider.name}, 状态: {resp.status_code}"
+                        f"提供商服务不可用: {provider.name}",
+                        provider_name=str(provider.name),
+                        upstream_status=resp.status_code,
+                        upstream_response=error_body,
+                    )
+                elif resp.status_code != 200:
+                    # 记录非200响应以便调试
+                    error_body = ""
+                    try:
+                        error_body = resp.text[:1000]
+                        logger.warning(f"  [{self.request_id}] 上游返回非200: status={resp.status_code}, body={error_body[:500]}")
+                    except Exception:
+                        pass
+                    raise ProviderNotAvailableException(
+                        f"提供商返回错误: {provider.name}, 状态: {resp.status_code}",
+                        provider_name=str(provider.name),
+                        upstream_status=resp.status_code,
+                        upstream_response=error_body,
                     )
 
                 response_json = resp.json()

src/core/exceptions.py

@@ -188,12 +188,16 @@ class ProviderNotAvailableException(ProviderException):
         message: str,
         provider_name: Optional[str] = None,
         request_metadata: Optional[Any] = None,
+        upstream_status: Optional[int] = None,
+        upstream_response: Optional[str] = None,
     ):
         super().__init__(
             message=message,
             provider_name=provider_name,
             request_metadata=request_metadata,
         )
+        self.upstream_status = upstream_status
+        self.upstream_response = upstream_response
 
 
 class ProviderTimeoutException(ProviderException):
@@ -442,6 +446,36 @@ class EmbeddedErrorException(ProviderException):
         self.error_status = error_status
 
 
+class ProviderCompatibilityException(ProviderException):
+    """Provider 兼容性错误异常 - 应该触发故障转移
+
+    用于处理因 Provider 不支持某些参数或功能导致的错误。
+    这类错误不是用户请求本身的问题，换一个 Provider 可能就能成功，应该触发故障转移。
+
+    常见场景：
+    - Unsupported parameter（不支持的参数）
+    - Unsupported model（不支持的模型）
+    - Unsupported feature（不支持的功能）
+    """
+
+    def __init__(
+        self,
+        message: str,
+        provider_name: Optional[str] = None,
+        status_code: int = 400,
+        upstream_error: Optional[str] = None,
+        request_metadata: Optional[Any] = None,
+    ):
+        self.upstream_error = upstream_error
+        super().__init__(
+            message=message,
+            provider_name=provider_name,
+            request_metadata=request_metadata,
+        )
+        # 覆盖状态码为 400（保持与上游一致）
+        self.status_code = status_code
+
+
 class UpstreamClientException(ProxyException):
     """上游返回的客户端错误异常 - HTTP 4xx 错误，不应该重试
 

src/main.py

@@ -4,13 +4,10 @@
 """
 
 from contextlib import asynccontextmanager
-from pathlib import Path
 
 import uvicorn
-from fastapi import FastAPI, HTTPException, Request
+from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import FileResponse
-from fastapi.staticfiles import StaticFiles
 
 from src.api.admin import router as admin_router
 from src.api.announcements import router as announcement_router
@@ -299,33 +296,6 @@ app.include_router(dashboard_router)  # 仪表盘端点
 app.include_router(public_router)  # 公开API端点（用户可查看提供商和模型）
 app.include_router(monitoring_router)  # 监控端点
 
-# 静态文件服务（前端构建产物）
-# 检查前端构建目录是否存在
-frontend_dist = Path(__file__).parent.parent / "frontend" / "dist"
-if frontend_dist.exists():
-    # 挂载静态资源目录
-    app.mount("/assets", StaticFiles(directory=str(frontend_dist / "assets")), name="assets")
-
-    # SPA catch-all路由 - 必须放在最后
-    @app.get("/{full_path:path}")
-    async def serve_spa(request: Request, full_path: str):
-        """
-        处理所有未匹配的GET请求，返回index.html供前端路由处理
-        仅对非API路径生效
-        """
-        # 如果是API路径，不处理
-        if full_path in {"api", "v1"} or full_path.startswith(("api/", "v1/")):
-            raise HTTPException(status_code=404, detail="Not Found")
-
-        # 返回index.html，让前端路由处理
-        index_file = frontend_dist / "index.html"
-        if index_file.exists():
-            return FileResponse(str(index_file))
-        else:
-            raise HTTPException(status_code=404, detail="Frontend not built")
-
-else:
-    logger.warning("前端构建目录不存在，前端路由将无法使用")
 
 
 def main():

src/services/orchestration/error_classifier.py

@@ -15,6 +15,7 @@ from src.core.enums import APIFormat
 from src.core.exceptions import (
     ConcurrencyLimitError,
     ProviderAuthException,
+    ProviderCompatibilityException,
     ProviderException,
     ProviderNotAvailableException,
     ProviderRateLimitException,
@@ -81,7 +82,9 @@ class ErrorClassifier:
         "context_length_exceeded",  # 上下文长度超限
         "content_length_limit",  # 请求内容长度超限 (Claude API)
         "content_length_exceeds",  # 内容长度超限变体 (AWS CodeWhisperer)
-        "max_tokens",  # token 数超限
+        # 注意：移除了 "max_tokens"，因为 max_tokens 相关错误可能是 Provider 兼容性问题
+        # 如 "Unsupported parameter: 'max_tokens' is not supported with this model"
+        # 这类错误应由 COMPATIBILITY_ERROR_PATTERNS 处理
         "invalid_prompt",  # 无效的提示词
         "content too long",  # 内容过长
         "input is too long",  # 输入过长 (AWS)
@@ -136,6 +139,19 @@ class ErrorClassifier:
         "CONTENT_POLICY_VIOLATION",
     )
 
+    # Provider 兼容性错误模式 - 这类错误应该触发故障转移
+    # 因为换一个 Provider 可能就能成功
+    COMPATIBILITY_ERROR_PATTERNS: Tuple[str, ...] = (
+        "unsupported parameter",  # 不支持的参数
+        "unsupported model",  # 不支持的模型
+        "unsupported feature",  # 不支持的功能
+        "not supported with this model",  # 此模型不支持
+        "model does not support",  # 模型不支持
+        "parameter is not supported",  # 参数不支持
+        "feature is not supported",  # 功能不支持
+        "not available for this model",  # 此模型不可用
+    )
+
     def _parse_error_response(self, error_text: Optional[str]) -> Dict[str, Any]:
         """
         解析错误响应为结构化数据
@@ -261,6 +277,25 @@ class ErrorClassifier:
         search_text = f"{parsed['message']} {parsed['raw']}".lower()
         return any(pattern.lower() in search_text for pattern in self.CLIENT_ERROR_PATTERNS)
 
+    def _is_compatibility_error(self, error_text: Optional[str]) -> bool:
+        """
+        检测错误响应是否为 Provider 兼容性错误（应触发故障转移）
+
+        这类错误是因为 Provider 不支持某些参数或功能导致的，
+        换一个 Provider 可能就能成功。
+
+        Args:
+            error_text: 错误响应文本
+
+        Returns:
+            是否为兼容性错误
+        """
+        if not error_text:
+            return False
+
+        search_text = error_text.lower()
+        return any(pattern.lower() in search_text for pattern in self.COMPATIBILITY_ERROR_PATTERNS)
+
     def _extract_error_message(self, error_text: Optional[str]) -> Optional[str]:
         """
         从错误响应中提取错误消息
@@ -425,6 +460,16 @@ class ErrorClassifier:
                 ),
             )
 
+        # 400 错误：先检查是否为 Provider 兼容性错误（应触发故障转移）
+        if status == 400 and self._is_compatibility_error(error_response_text):
+            logger.info(f"检测到 Provider 兼容性错误，将触发故障转移: {extracted_message}")
+            return ProviderCompatibilityException(
+                message=extracted_message or "Provider 不支持此请求",
+                provider_name=provider_name,
+                status_code=400,
+                upstream_error=error_response_text,
+            )
+
         # 400 错误：检查是否为客户端请求错误（不应重试）
         if status == 400 and self._is_client_error(error_response_text):
             logger.info(f"检测到客户端请求错误，不进行重试: {extracted_message}")

src/services/orchestration/fallback_orchestrator.py

@@ -427,6 +427,9 @@ class FallbackOrchestrator:
             )
             # str(cause) 可能为空（如 httpx 超时异常），使用 repr() 作为备用
             error_msg = str(cause) or repr(cause)
+            # 如果是 ProviderNotAvailableException，附加上游响应
+            if hasattr(cause, "upstream_response") and cause.upstream_response:
+                error_msg = f"{error_msg} | 上游响应: {cause.upstream_response[:500]}"
             RequestCandidateService.mark_candidate_failed(
                 db=self.db,
                 candidate_id=candidate_record_id,
@@ -439,6 +442,9 @@ class FallbackOrchestrator:
 
         # 未知错误：记录失败并抛出
         error_msg = str(cause) or repr(cause)
+        # 如果是 ProviderNotAvailableException，附加上游响应
+        if hasattr(cause, "upstream_response") and cause.upstream_response:
+            error_msg = f"{error_msg} | 上游响应: {cause.upstream_response[:500]}"
         RequestCandidateService.mark_candidate_failed(
             db=self.db,
             candidate_id=candidate_record_id,

src/services/request/result.py

@@ -289,11 +289,17 @@ class RequestResult:
             status_code = 500
             error_type = "internal_error"
 
+        # 构建错误消息，包含上游响应信息
+        error_message = str(exception)
+        if isinstance(exception, ProviderNotAvailableException):
+            if exception.upstream_response:
+                error_message = f"{error_message} | 上游响应: {exception.upstream_response[:500]}"
+
         return cls(
             status=RequestStatus.FAILED,
             metadata=metadata,
             status_code=status_code,
-            error_message=str(exception),
+            error_message=error_message,
             error_type=error_type,
             response_time_ms=response_time_ms,
             is_stream=is_stream,