@mmmbuto/zai-codex-bridge 0.1.10 → 0.1.12

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mmmbuto/zai-codex-bridge",
-  "version": "0.1.10",
+  "version": "0.1.12",
   "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

@@ -199,16 +199,28 @@ function translateChatToResponses(chatResponse) {
     }
   }
 
+  const responseId = 'resp_' + Date.now() + '_' + Math.random().toString(36).substr(2, 9);
+
+  // OpenAI Responses API format
   const response = {
+    id: responseId,
+    object: 'response',
+    created_at: Math.floor(Date.now() / 1000),
+    status: 'completed',
+    model: chatResponse.model || 'glm-4.7',
     output: [{
-      value: text,
-      content_type: 'text'
+      type: 'message',
+      role: 'assistant',
+      content: [{
+        type: 'output_text',
+        text: text
+      }]
     }],
-    status: 'completed',
     usage: Object.keys(usage).length > 0 ? usage : undefined
   };
 
   log('debug', 'Translated Chat->Responses:', {
+    id: response.id,
     outputLength: text.length,
     status: response.status
   });
@@ -255,12 +267,13 @@ async function makeUpstreamRequest(path, body, headers) {
 /**
  * Handle streaming response from Z.AI
  */
-async function streamChatToResponses(stream, res) {
+async function streamChatToResponses(stream, res, responseId) {
   const decoder = new TextDecoder();
   let buffer = '';
   let chunkCount = 0;
   let deltaCount = 0;
-  let lastParsed = null; // Keep track of last parsed SSE for ID extraction
+  let lastParsed = null;
+  const itemId = 'item_' + Date.now();
 
   log('debug', 'Starting to process stream');
 
@@ -286,23 +299,32 @@ async function streamChatToResponses(stream, res) {
       if (data === '[DONE]') {
         log('info', `Stream end received - wrote ${deltaCount} deltas total`);
 
-        // Send response.completed event (required by Codex Responses API)
-        // Map Z.AI usage format to Responses API format
+        // Send response.completed event in OpenAI Responses API format
         const zaiUsage = lastParsed?.usage;
         const completedEvent = {
-          id: lastParsed?.id || 'msg_' + Date.now(),
-          usage: zaiUsage ? {
-            input_tokens: zaiUsage.prompt_tokens || 0,
-            output_tokens: zaiUsage.completion_tokens || 0,
-            total_tokens: zaiUsage.total_tokens || 0
-          } : {
-            input_tokens: 0,
-            output_tokens: 0,
-            total_tokens: 0
-          }
+          type: 'response.completed',
+          response: {
+            id: responseId,
+            status: 'completed',
+            output: [{
+              type: 'message',
+              role: 'assistant',
+              content: [{ type: 'output_text', text: '' }]
+            }],
+            usage: zaiUsage ? {
+              input_tokens: zaiUsage.prompt_tokens || 0,
+              output_tokens: zaiUsage.completion_tokens || 0,
+              total_tokens: zaiUsage.total_tokens || 0
+            } : {
+              input_tokens: 0,
+              output_tokens: 0,
+              total_tokens: 0
+            }
+          },
+          sequence_number: deltaCount
         };
 
-        res.write(`event: response.completed\n`);
+        log('info', 'Sending response.completed event');
         res.write(`data: ${JSON.stringify(completedEvent)}\n\n`);
         log('info', 'Sent response.completed event');
         return;
@@ -310,26 +332,31 @@ async function streamChatToResponses(stream, res) {
 
       try {
         const parsed = JSON.parse(data);
-        lastParsed = parsed; // Save for later use in completed event
+        lastParsed = parsed;
         log('debug', 'Parsed SSE:', JSON.stringify(parsed).substring(0, 150));
 
         const delta = parsed.choices?.[0]?.delta;
-
-        // Z.AI uses reasoning_content instead of content
         const content = delta?.content || delta?.reasoning_content || '';
 
         if (content) {
           deltaCount++;
           log('debug', 'Writing delta:', content.substring(0, 30));
-          res.write(`event: output.text.delta\n`);
-          res.write(`data: ${JSON.stringify({ value: content })}\n\n`);
+          // OpenAI Responses API format for text delta
+          const deltaEvent = {
+            type: 'response.output_text.delta',
+            delta: content,
+            output_index: 0,
+            item_id: itemId,
+            sequence_number: deltaCount - 1
+          };
+          res.write(`data: ${JSON.stringify(deltaEvent)}\n\n`);
         }
       } catch (e) {
         log('warn', 'Failed to parse SSE chunk:', e.message, 'data:', data.substring(0, 100));
       }
     }
 
-    if (chunkCount > 100) {
+    if (chunkCount > 1000) {
       log('warn', 'Too many chunks, possible loop');
       return;
     }
@@ -412,6 +439,7 @@ async function handlePostRequest(req, res) {
 
     // Handle streaming response
     if (upstreamBody.stream) {
+      const responseId = 'resp_' + Date.now() + '_' + Math.random().toString(36).substr(2, 9);
       log('info', 'Starting streaming response');
       res.writeHead(200, {
         'Content-Type': 'text/event-stream; charset=utf-8',
@@ -420,7 +448,7 @@ async function handlePostRequest(req, res) {
       });
 
       try {
-        await streamChatToResponses(upstreamResponse.body, res);
+        await streamChatToResponses(upstreamResponse.body, res, responseId);
         log('info', 'Streaming completed');
       } catch (e) {
         log('error', 'Streaming error:', e);