ai-codegen-cli-vrk 2.1.0 → 2.1.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-codegen-cli-vrk",
-  "version": "2.1.0",
+  "version": "2.1.1",
   "description": "Minimalist Terminal-based AI code generator",
   "type": "module",
   "bin": {

package/src/aiClient.js

@@ -1,51 +1,47 @@
 import { GoogleGenerativeAI } from "@google/generative-ai";
 
 let genAI = null;
-let activeModel = null;
+let modelQueue = [];
+let currentModelIndex = 0;
 
 export function setApiKey(apiKey) {
   genAI = new GoogleGenerativeAI(apiKey.trim());
 }
 
 /**
- * Ultimate discovery: Asks Google for a list of working models for your key.
+ * Lists all usable models and sorts them by stability.
  */
-async function getWorkingModel() {
-  if (activeModel) return activeModel;
+async function refreshModelQueue() {
+  if (modelQueue.length > 0) return;
 
   try {
-    // Attempt to list models to see what this specific key is allowed to use
-    // We use v1beta as it has the most complete model list
     const response = await fetch(`https://generativelanguage.googleapis.com/v1beta/models?key=${genAI.apiKey}`);
     const data = await response.json();
     
     if (data.models) {
-      // Look for 1.5-flash or 2.0-flash in the permitted list
-      const best = data.models.find(m => 
-        m.supportedGenerationMethods.includes("generateContent") && 
-        (m.name.includes("1.5-flash") || m.name.includes("2.0-flash"))
-      );
+      // Prioritize 1.5-flash (most stable) then 2.0-flash (experimental)
+      const usable = data.models
+        .filter(m => m.supportedGenerationMethods.includes("generateContent"))
+        .map(m => m.name.split('/').pop());
+
+      const preferred = ["gemini-1.5-flash", "gemini-1.5-flash-latest", "gemini-2.0-flash-exp", "gemini-pro"];
       
-      if (best) {
-        // Strip "models/" prefix if present for the SDK
-        const modelName = best.name.split('/').pop();
-        activeModel = genAI.getGenerativeModel({ model: modelName }, { apiVersion: "v1beta" });
-        return activeModel;
-      }
+      modelQueue = preferred.filter(p => usable.includes(p));
+      // Add any other usable models found as backup
+      usable.forEach(u => { if(!modelQueue.includes(u)) modelQueue.push(u); });
     }
   } catch (e) {
-    // If listing fails, fall back to a safe guess
+    modelQueue = ["gemini-1.5-flash", "gemini-pro"];
   }
-
-  // Final fallback to the most common working model
-  activeModel = genAI.getGenerativeModel({ model: "gemini-1.5-flash" }, { apiVersion: "v1beta" });
-  return activeModel;
 }
 
 export async function generateFullProject(task, tests, retryCount = 0) {
+  await refreshModelQueue();
+  
+  const modelName = modelQueue[currentModelIndex] || "gemini-1.5-flash";
+  const model = genAI.getGenerativeModel({ model: modelName }, { apiVersion: "v1beta" });
+
   try {
-    const model = await getWorkingModel();
-    
     const prompt = `
 Generate the ENTIRE project in a SINGLE response. 
 Strictly pass all test cases.
@@ -74,11 +70,22 @@ ${tests}
     return text.replace(/```[a-z]*\n([\s\S]*?)\n```/gi, "$1").trim();
 
   } catch (error) {
-    // Handle Overloaded (503) or Rate Limit (429) - Critical for Free Tier
     const msg = error.message || "";
-    if ((msg.includes("503") || msg.includes("429") || msg.includes("overloaded")) && retryCount < 5) {
-      console.log(`..... (Server busy, waiting 15s to retry ${retryCount + 1}/5)`);
-      await new Promise(r => setTimeout(r, 15000));
+    
+    // Handle Quota Exceeded (429) or Server Busy (503)
+    if ((msg.includes("429") || msg.includes("quota") || msg.includes("503") || msg.includes("overloaded")) && retryCount < 5) {
+      
+      // If a specific model is exhausted, try to switch to the next one in the queue
+      if (currentModelIndex < modelQueue.length - 1) {
+        console.log(`..... (Model ${modelName} exhausted, switching to ${modelQueue[currentModelIndex+1]}...)`);
+        currentModelIndex++;
+        return generateFullProject(task, tests, retryCount + 1);
+      }
+
+      // If all models are exhausted, wait a full 60 seconds
+      console.log(`..... (All models busy/exhausted, waiting 60s to reset quota...)`);
+      await new Promise(r => setTimeout(r, 60000));
+      currentModelIndex = 0; // Reset queue and try again
       return generateFullProject(task, tests, retryCount + 1);
     }
     throw error;