Upload server.js

server.js

@@ -368,6 +368,13 @@ app.post('/v1/chat/completions', gssAuth, async (req, res) => {
     ? (process.env.OLLAMA_BASE_URL || 'https://ollama.com/api')
     : (process.env.GROQ_BASE_URL   || 'https://api.groq.com/openai/v1');
 
+  // For Ollama cloud, use native /chat endpoint with different model format
+  const isOllamaCloud = useOllama && baseUrl.includes('ollama.com');
+  const chatEndpoint  = isOllamaCloud ? `${baseUrl}/chat` : `${baseUrl}/chat/completions`;
+  if (isOllamaCloud && !req.body.model?.includes(':')) {
+    req.body.model = 'gpt-oss:120b';
+  }
+
   const poolKeys = (subscriberPools[subscriberId]?.[provider]) || (subscriberPools['env']?.groq) || [];
   const maxRetries = Math.min(poolKeys.length, 10);
 
@@ -387,7 +394,7 @@ app.post('/v1/chat/completions', gssAuth, async (req, res) => {
           try {
             const ollamaUrl = process.env.OLLAMA_BASE_URL || 'https://ollama.com/api';
             const ollamaBody = {
-              model: 'llama3.2',
+              model: 'gpt-oss:120b',
               messages: req.body.messages || [],
               stream: false,
             };
@@ -411,12 +418,21 @@ app.post('/v1/chat/completions', gssAuth, async (req, res) => {
     }
 
     try {
-      const upstream = await fetch(`${baseUrl}/chat/completions`, {
+      const upstream = await fetch(chatEndpoint, {
         method:  'POST',
         headers: { 'Authorization': `Bearer ${entry.key}`, 'Content-Type': 'application/json' },
         body: JSON.stringify(req.body),
       });
 
+      // For Ollama cloud responses, convert to OpenAI format
+      const processResponse = async (r) => {
+        const data = await r.json();
+        if (isOllamaCloud && data.message) {
+          return { choices: [{ message: { role: 'assistant', content: data.message.content || '' }, finish_reason: 'stop' }], model: req.body.model };
+        }
+        return data;
+      };
+
       // 401/403 = invalid/expired key — cooldown and try next
       if (upstream.status === 401 || upstream.status === 403) {
         cooldownKey(entry);
@@ -429,9 +445,9 @@ app.post('/v1/chat/completions', gssAuth, async (req, res) => {
             ollamaFallbackAttempted = true;
             console.warn(`[GSS] All Groq keys invalid for sub ${subscriberId}, switching to Ollama cloud`);
             const ollamaUrl = process.env.OLLAMA_BASE_URL || 'https://ollama.com/api';
-            // Convert OpenAI format to Ollama native format
+            // Ollama cloud native format
             const ollamaBody = {
-              model: 'llama3.2',
+              model: 'gpt-oss:120b',
               messages: req.body.messages || [],
               stream: false,
             };
@@ -442,7 +458,6 @@ app.post('/v1/chat/completions', gssAuth, async (req, res) => {
             });
             if (ollamaUpstream.ok) {
               const ollamaData = await ollamaUpstream.json();
-              // Convert Ollama response to OpenAI format
               const content = ollamaData.message?.content || ollamaData.response || '';
               return res.status(200).json({
                 choices: [{ message: { role: 'assistant', content }, finish_reason: 'stop' }],
@@ -461,7 +476,7 @@ app.post('/v1/chat/completions', gssAuth, async (req, res) => {
         continue;
       }
 
-      const data = await upstream.json();
+      const data = await processResponse(upstream);
       return res.status(upstream.status).json(data);
 
     } catch (err) {