The chat reply is always the same

[minivv]

I call it from the N8N AI node and always get the same AI reply：

I can help you with coding tasks! The proxy connection is working well. What would you like assistance with? (Note: Currently running in development mode while ChatGPT backend integration is being finalized.)

How do I use it like the openai api?

log

$ ./target/release/codex-openai-proxy --port 8001 --auth-path ~/.codex/auth.json
Initializing Codex OpenAI Proxy...
✓ Loaded authentication from /Users/apple/.codex/auth.json
🚀 Codex OpenAI Proxy listening on http://0.0.0.0:8001
   Health check: http://localhost:8001/health
   Chat endpoint: http://localhost:8001/v1/chat/completions

   Configure CLINE with:
   Base URL: http://localhost:8001
   Model: gpt-5
   API Key: (any value)

🔍 === INTERCEPTED REQUEST ===
⏰ Timestamp: 2025-09-25 01:06:34.533 UTC
📥 Method: POST
📍 Path: /v1/chat/completions

📋 Headers (5 total):
  📄 host: localhost:8001
  🎯 user-agent: curl/8.7.1
  📄 accept: */*
  📄 content-type: application/json
  📄 content-length: 69
🔍 === END INTERCEPT ===

🔥 === MATCHED CHAT COMPLETIONS ===

📋 === CLINE REQUEST DETAILS FOR CURL ===
Method: POST
Path: /v1/chat/completions
Body size: 69 bytes

Headers for curl:
  -H "host: localhost:8001"
  -H "user-agent: curl/8.7.1"
  -H "accept: */*"
  -H "content-type: application/json"
  -H "content-length: 69"

Body (first 1000 chars):
{"model": "gpt-5", "messages": [{"role": "user", "content": "Test"}]}

🚀 CURL COMMAND TO REPLICATE:
curl -X POST http://localhost:8888/v1/chat/completions \
  -H "user-agent: curl/8.7.1" \
  -H "accept: */*" \
  -H "content-type: application/json" \
  -H "content-length: 69" \
  -d '{"model": "gpt-5", "messages": [{"role": "user", "content": "Test"}]}'
📋 === END CLINE REQUEST DETAILS ===

   Model: gpt-5
   Messages: 1 items
   [0] user: Test
🔥 === END MATCHED ===

🔄 Processing CLINE request...
🔍 Stream setting: None