Harden async task flows and enhance analysis tooling

server/_core/llm.ts

@@ -1,4 +1,5 @@
 import { ENV } from "./env";
+import { fetchWithTimeout } from "./fetch";
 
 export type Role = "system" | "user" | "assistant" | "tool" | "function";
 
@@ -323,13 +324,17 @@ export async function invokeLLM(params: InvokeParams): Promise<InvokeResult> {
     payload.response_format = normalizedResponseFormat;
   }
 
-  const response = await fetch(resolveApiUrl(apiUrl), {
+  const response = await fetchWithTimeout(resolveApiUrl(apiUrl), {
     method: "POST",
     headers: {
       "content-type": "application/json",
       authorization: `Bearer ${apiKey || ENV.llmApiKey}`,
     },
     body: JSON.stringify(payload),
+  }, {
+    timeoutMs: ENV.llmTimeoutMs,
+    retries: ENV.llmRetryCount,
+    retryMethods: ["POST"],
   });
 
   if (!response.ok) {