Update app.js

app.js

@@ -50,7 +50,7 @@ const supabase = createClient(
 
 let memoryChats = {}; 
 let dirtyChats = new Set(); 
-const activeGenerations = new Map(); // Tracks AbortControllers for stopping generation
+const activeGenerations = new Map(); 
 
 async function initDB() {
     try {
@@ -150,7 +150,6 @@ app.delete('/api/chats/:id', async (req, res) => {
     res.json({ success: true });
 });
 
-// --- STOP GENERATION ENDPOINT ---
 app.post('/api/chats/:id/stop', (req, res) => {
     const { id } = req.params;
     log.info("User requested to stop generation.", id);
@@ -193,7 +192,6 @@ app.post('/api/chats/:id/stream', async (req, res) => {
     memoryChats[id].isGenerating = true;
     dirtyChats.add(id);
 
-    // Create Abort Controller for this generation
     const abortController = new AbortController();
     activeGenerations.set(id, abortController);
 
@@ -239,19 +237,21 @@ app.post('/api/chats/:id/stream', async (req, res) => {
         
         historicalMessages.push({ role: "user", content: contentBlock });
 
+        // THE FIX: Uncapped Token limit for Claude 3.7 to allow massive reasoning + coding
+        const commandMaxTokens = model.includes("claude") ? 64000 : 8192;
+
         const command = new ConverseStreamCommand({
             modelId: bedrockModelId,
             system:[{ text: system_prompt || CLAUDE_SYSTEM_PROMPT }],
             messages: historicalMessages,
-            inferenceConfig: { maxTokens: model.includes("haiku") ? 32000 : 4000, temperature: 1 },
+            inferenceConfig: { maxTokens: commandMaxTokens, temperature: 1 },
             additionalModelRequestFields: model.includes("claude") ? { 
-                thinking: { type: "adaptive" }, output_config: { effort: "high" }
+                thinking: { type: "adaptive" } 
             } : undefined
         });
 
-        // Pass abortSignal to allow stopping mid-generation
         const response = await bedrockClient.send(command, { abortSignal: abortController.signal });
-        log.info("Bedrock connected successfully, streaming chunks...", id);
+        log.info(`Bedrock connected successfully (Max Tokens: ${commandMaxTokens}), streaming chunks...`, id);
         
         for await (const chunk of response.stream) {
             if (chunk.contentBlockDelta) {
@@ -264,6 +264,10 @@ app.post('/api/chats/:id/stream', async (req, res) => {
                     safeWrite(delta.text);
                 }
             }
+            // Log exactly why it stopped (e.g. max_tokens vs end_turn)
+            if (chunk.messageStop) {
+                log.info(`Stream stopped. Reason: ${chunk.messageStop.stopReason}`, id);
+            }
             if (chunk.metadata && chunk.metadata.usage) {
                 streamInputTokens = chunk.metadata.usage.inputTokens || 0;
                 streamOutputTokens = chunk.metadata.usage.outputTokens || 0;
@@ -286,11 +290,9 @@ app.post('/api/chats/:id/stream', async (req, res) => {
     } finally {
         activeGenerations.delete(id);
         
-        // Save tokens generated up to point of failure/abort
         memoryChats[id].inputTokens += streamInputTokens;
         memoryChats[id].outputTokens += streamOutputTokens;
         memoryChats[id].totalTokens += streamTotalTokens;
-        
         memoryChats[id].isGenerating = false;
         dirtyChats.add(id);