fix: fix retry

README.md

@@ -75,6 +75,12 @@ The server currently supports the following endpoints:
 > **Note**: `/v1/completions` is not supported because Gemini models do not support completion functionality, and Google's PaLM model (which does) is likely to be deprecated.  
 
 
+## Roadmap
+
+- [x] v1
+- [ ] 支持配置多个 api key ，并且可以轮询、分配权限、重试等调度操作
+- [ ] 支持配置和切换代理
+
 ## Building the Project  
 
 ```

package.json

@@ -12,6 +12,7 @@
     "@fastify/swagger": "^8.15.0",
     "@fastify/swagger-ui": "^2.1.0",
     "@google/generative-ai": "^0.21.0",
+    "@types/async-retry": "^1.4.9",
     "async-retry": "^1.3.3",
     "commander": "^12.1.0",
     "dotenv": "^16.4.5",

src/v1/chat/completions.ts

@@ -192,6 +192,7 @@ class ChatCompletionsHandler {
         const is_hit_err = retry_error.some((x) => err instanceof x);
         if (is_hit_err === false) {
           bail(err);
+          if (err) err.bail = true;
         }
         throw err; // 触发重试
       }
@@ -519,7 +520,12 @@ const chatCompletionsRoute: FastifyPluginAsync = async (app) => {
     url: "/v1/chat/completions",
     schema: { body: CHAT_COMPLETION_SCHEMA },
     handler: async (req, reply) => {
-      const handler = new ChatCompletionsHandler(app, req.body, req, reply);
+      const handler = new ChatCompletionsHandler(
+        app,
+        req.body as any,
+        req,
+        reply
+      );
       await handler.processRequest();
     },
   });