Update ai_engine.js

ai_engine.js

@@ -1,44 +1,61 @@
 import OpenAI from "openai";
 
-// Initialize Azure OpenAI strictly from environment variables
-const azureOpenAI = new OpenAI({
-    apiKey: process.env.AZURE_OPENAI_API_KEY,
-    baseURL: `${process.env.AZURE_OPENAI_ENDPOINT}/openai/deployments/${process.env.AZURE_OPENAI_DEPLOYMENT_NAME}`,
-    defaultQuery: { "api-version": process.env.AZURE_OPENAI_API_VERSION || "2024-05-01-preview" },
-    defaultHeaders: { "api-key": process.env.AZURE_OPENAI_API_KEY }
+// Initialize OpenRouter using the standard OpenAI SDK
+const openRouterAI = new OpenAI({
+    apiKey: process.env.OPENROUTER_API_KEY,
+    baseURL: "https://openrouter.ai/api/v1",
+    defaultHeaders: {
+        "HTTP-Referer": process.env.YOUR_SITE_URL || "https://everydaycats.web.apl", // Optional: Helps with OpenRouter rankings
+        "X-Title": process.env.YOUR_SITE_NAME || "Aura Scanner", // Optional: Shows in OpenRouter dashboard
+    }
 });
 
+// --- DYNAMIC MODEL ROUTER ---
+const getModelId = (modelName) => {
+    switch(modelName?.toLowerCase()) {
+        case "llama":
+        case "maverick":
+            return "meta-llama/llama-4-maverick";
+        case "gpt":
+        case "gpt-5.4-nano":
+            return "openai/gpt-5.4-nano";
+        case "qwen":
+        default:
+            // The Default Winner: Cheapest input, massive throughput, perfect for JSON & Lore
+            return "qwen/qwen3-vl-30b-a3b-thinking";
+    }
+};
+
 // --- DEDICATED ERROR LOGGER ---
-const logAzureError = (error, context) => {
+const logOpenRouterError = (error, context) => {
     const status = error.status || error.response?.status;
     
-    if (status === 429) {
-        console.error(`🚨 [AZURE THROTTLING - 429] ${context}: Rate limit exceeded or quota exhausted. Slow down requests.`);
+    if (status === 402) {
+        console.error(`💸 [OPENROUTER FUNDS - 402] ${context}: Insufficient credits! Top up your OpenRouter account.`);
+    } else if (status === 429) {
+        console.error(`🚨[OPENROUTER THROTTLING - 429] ${context}: Rate limit exceeded. Provider is congested.`);
     } else if (status === 401) {
-        console.error(`🛑 [AZURE AUTH - 401] ${context}: Invalid API Key or Endpoint URL.`);
+        console.error(`🛑[OPENROUTER AUTH - 401] ${context}: Invalid API Key.`);
     } else if (status === 400) {
-        console.error(`⚠️ [AZURE BAD REQUEST - 400] ${context}: Invalid payload, image too large, or Azure Content Filter triggered.`);
+        console.error(`⚠️ [OPENROUTER BAD REQUEST - 400] ${context}: Invalid payload, image format, or context length exceeded.`);
     } else if (status >= 500) {
-        console.error(`🔥[AZURE SERVER ERROR - ${status}] ${context}: Azure backend is struggling or down.`);
+        console.error(`🔥[OPENROUTER SERVER ERROR - ${status}] ${context}: The specific AI provider backend is down.`);
     } else if (error.code === 'ECONNRESET' || error.code === 'ETIMEDOUT') {
-        console.error(`⏳ [NETWORK TIMEOUT] ${context}: Connection to Azure dropped.`);
+        console.error(`⏳ [NETWORK TIMEOUT] ${context}: Connection to OpenRouter dropped.`);
     } else {
-        console.error(`❌[AZURE ERROR] ${context}:`, error.message || error);
-    }
-    
-    // Log the deep Azure inner error if it exists (crucial for content filter details)
-    if (error.error?.innererror) {
-        console.error(`   ↳ Deep Error:`, JSON.stringify(error.error.innererror));
+        console.error(`❌[OPENROUTER ERROR] ${context}:`, error.message || error);
     }
 };
 
-export const generateCompletion = async ({ prompt, system_prompt, images }) => {
+export const generateCompletion = async ({ model, prompt, system_prompt, images }) => {
     try {
+        const targetModel = getModelId(model);
+        
         let messagesPayload =[
             { role: "system", content: system_prompt || "You are an elite AI assistant." }
         ];
 
-        // 🚨 VISION SUPPORT: OpenAI requires the data URI prefix for base64
+        // 🚨 VISION SUPPORT: OpenAI/OpenRouter requires the data URI prefix for base64
         if (images && Array.isArray(images) && images.length > 0) {
             let userContent = [{ type: "text", text: prompt }];
             
@@ -54,20 +71,22 @@ export const generateCompletion = async ({ prompt, system_prompt, images }) => {
             messagesPayload.push({ role: "user", content: prompt });
         }
 
-        const response = await azureOpenAI.chat.completions.create({
-            model: process.env.AZURE_OPENAI_DEPLOYMENT_NAME,
+        const response = await openRouterAI.chat.completions.create({
+            model: targetModel,
             messages: messagesPayload,
-            // reasoning_effort: "high" // Uncomment if you want to force deeper reasoning
+            // Force JSON output (Supported by Qwen, Llama, and GPT on OpenRouter)
+            response_format: { type: "json_object" } 
         });
         
         return { 
             success: true, 
             data: response.choices[0].message.content,
-            usage: response.usage 
+            usage: response.usage,
+            model_used: targetModel
         };
         
     } catch (error) {
-        logAzureError(error, "generateCompletion");
+        logOpenRouterError(error, "generateCompletion");
         return { 
             success: false, 
             error: error.message, 
@@ -76,8 +95,10 @@ export const generateCompletion = async ({ prompt, system_prompt, images }) => {
     }
 };
 
-export const streamCompletion = async ({ prompt, system_prompt, images, res }) => {
+export const streamCompletion = async ({ model, prompt, system_prompt, images, res }) => {
     try {
+        const targetModel = getModelId(model);
+        
         let messagesPayload =[
             { role: "system", content: system_prompt || "You are an elite AI assistant." }
         ];
@@ -97,8 +118,8 @@ export const streamCompletion = async ({ prompt, system_prompt, images, res }) =
             messagesPayload.push({ role: "user", content: prompt });
         }
 
-        const stream = await azureOpenAI.chat.completions.create({
-            model: process.env.AZURE_OPENAI_DEPLOYMENT_NAME,
+        const stream = await openRouterAI.chat.completions.create({
+            model: targetModel,
             messages: messagesPayload,
             stream: true,
             stream_options: { include_usage: true } 
@@ -109,6 +130,7 @@ export const streamCompletion = async ({ prompt, system_prompt, images, res }) =
         for await (const chunk of stream) {
             const delta = chunk.choices[0]?.delta;
             
+            // Handle Qwen's specific "Thinking" blocks if they stream them
             if (delta?.reasoning_content) {
                 res.write(`__THINK__${delta.reasoning_content}`);
             } else if (delta?.content) {
@@ -124,9 +146,9 @@ export const streamCompletion = async ({ prompt, system_prompt, images, res }) =
         res.end();
 
     } catch (error) {
-        logAzureError(error, "streamCompletion");
+        logOpenRouterError(error, "streamCompletion");
         
-        // Safely write the error to the stream so the frontend doesn't hang forever waiting for chunks
+        // Safely write the error to the stream so the frontend doesn't hang forever
         res.write(`\n\n[ERROR]: ${error.message}`);
         res.end();
     }