Use billable AI model for session memory

backend/src/ai/LocalAIApi.js

@@ -154,7 +154,7 @@ async function awaitResponse(aiRequestId, options = {}) {
   const interval = Math.max(Number(options.interval ?? 5), 1);
   const deadline = Date.now() + Math.max(timeout, interval) * 1000;
 
-  while (true) {
+  while (Date.now() < deadline) {
     const statusResp = await fetchStatus(aiRequestId, {
       headers: options.headers,
       timeout: options.timeout_per_call,
@@ -184,16 +184,14 @@ async function awaitResponse(aiRequestId, options = {}) {
       return statusResp;
     }
 
-    if (Date.now() >= deadline) {
-      return {
-        success: false,
-        error: "timeout",
-        message: "Timed out waiting for AI response.",
-      };
-    }
-
     await sleep(interval * 1000);
   }
+
+  return {
+    success: false,
+    error: "timeout",
+    message: "Timed out waiting for AI response.",
+  };
 }
 
 function extractText(response) {
@@ -283,7 +281,7 @@ function config() {
     projectId,
     projectUuid: process.env.PROJECT_UUID || null,
     projectHeader: process.env.AI_PROJECT_HEADER || "project-uuid",
-    defaultModel: process.env.AI_DEFAULT_MODEL || "gpt-5-mini",
+    defaultModel: process.env.AI_DEFAULT_MODEL || "gpt-5.5",
     timeout,
     verifyTls,
   };