@mmmbuto/zai-codex-bridge 0.1.5 → 0.1.7

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mmmbuto/zai-codex-bridge",
-  "version": "0.1.5",
+  "version": "0.1.7",
   "description": "Local proxy that translates OpenAI Responses API format to Z.AI Chat Completions format for Codex",
   "main": "src/server.js",
   "bin": {

package/src/server.js

@@ -103,8 +103,10 @@ function translateResponsesToChat(request) {
   }
 
   // Build chat request
+  // Normalize model name to lowercase (Z.AI requirement)
+  const model = (request.model || 'glm-4.7').toLowerCase();
   const chatRequest = {
-    model: request.model,
+    model: model,
     messages: messages,
     stream: request.stream !== false // default true
   };
@@ -206,7 +208,9 @@ async function makeUpstreamRequest(path, body, headers) {
     path: path,
     cleanPath: cleanPath,
     base: ZAI_BASE_URL,
-    hasAuth: !!upstreamHeaders.Authorization
+    hasAuth: !!upstreamHeaders.Authorization,
+    bodyKeys: Object.keys(body),
+    bodyPreview: JSON.stringify(body).substring(0, 200)
   });
 
   const response = await fetch(url, {
@@ -221,22 +225,34 @@ async function makeUpstreamRequest(path, body, headers) {
 /**
  * Handle streaming response from Z.AI
  */
-async function* streamChatToResponses(stream, res) {
+async function streamChatToResponses(stream, res) {
   const decoder = new TextDecoder();
   let buffer = '';
+  let chunkCount = 0;
+
+  log('debug', 'Starting to process stream');
 
   for await (const chunk of stream) {
     buffer += decoder.decode(chunk, { stream: true });
     const lines = buffer.split('\n');
     buffer = lines.pop() || '';
 
+    chunkCount++;
+
     for (const line of lines) {
-      if (!line.trim() || !line.startsWith('data: ')) continue;
+      if (!line.trim() || !line.startsWith('data: ')) {
+        if (line.trim() && !line.startsWith(':')) {
+          log('debug', 'Non-data line:', line.substring(0, 50));
+        }
+        continue;
+      }
 
       const data = line.slice(6).trim();
+      log('debug', 'SSE data:', data.substring(0, 100));
 
       // Check for stream end
       if (data === '[DONE]') {
+        log('debug', 'Stream end received');
         res.write(`event: completed\n`);
         res.write(`data: ${JSON.stringify({ status: 'completed' })}\n\n`);
         return;
@@ -244,17 +260,30 @@ async function* streamChatToResponses(stream, res) {
 
       try {
         const parsed = JSON.parse(data);
+        log('debug', 'Parsed SSE:', JSON.stringify(parsed).substring(0, 150));
+
         const delta = parsed.choices?.[0]?.delta;
 
-        if (delta?.content) {
+        // Z.AI uses reasoning_content instead of content
+        const content = delta?.content || delta?.reasoning_content || '';
+
+        if (content) {
+          log('debug', 'Writing delta:', content.substring(0, 30));
           res.write(`event: output.text.delta\n`);
-          res.write(`data: ${JSON.stringify({ value: delta.content })}\n\n`);
+          res.write(`data: ${JSON.stringify({ value: content })}\n\n`);
         }
       } catch (e) {
-        log('warn', 'Failed to parse SSE chunk:', e.message);
+        log('warn', 'Failed to parse SSE chunk:', e.message, 'data:', data.substring(0, 100));
       }
     }
+
+    if (chunkCount > 100) {
+      log('warn', 'Too many chunks, possible loop');
+      return;
+    }
   }
+
+  log('debug', 'Stream ended naturally');
 }
 
 /**
@@ -287,7 +316,8 @@ async function handlePostRequest(req, res) {
   log('info', 'Incoming request:', {
     path,
     format: detectFormat(request),
-    model: request.model
+    model: request.model,
+    authHeader: req.headers['authorization'] || req.headers['Authorization'] || 'none'
   });
 
   let upstreamBody;
@@ -330,13 +360,19 @@ async function handlePostRequest(req, res) {
 
     // Handle streaming response
     if (upstreamBody.stream) {
+      log('info', 'Starting streaming response');
       res.writeHead(200, {
         'Content-Type': 'text/event-stream; charset=utf-8',
         'Cache-Control': 'no-cache',
         'Connection': 'keep-alive'
       });
 
-      await streamChatToResponses(upstreamResponse.body, res);
+      try {
+        await streamChatToResponses(upstreamResponse.body, res);
+        log('info', 'Streaming completed');
+      } catch (e) {
+        log('error', 'Streaming error:', e);
+      }
       res.end();
     } else {
       // Non-streaming response