codex-proxy

An OpenAI Responses API proxy for Gemini and Z.AI (GLM) providers.

Translates OpenAI's Responses API to Gemini and Z.AI APIs. Handles wire format differences, role mapping, and SSE stream formatting so Codex can use these providers instead of GPT.

Features

• Responses API - Full lifecycle with SSE events
• Multi-Provider - Gemini (OAuth2) and Z.AI (GLM) support
• Context Compaction - Both Gemini and Z.AI models support
• Tool Support - Function calling and web search
• Docker Ready - Production container with hot-reload

Quick Start

# Clone and start
git clone https://github.com/cornellsh/codex-proxy.git
cd codex-proxy

# Start proxy (Docker)
./scripts/control.sh start

# Or run directly (Python 3.14+ required)
python -m codex_proxy

Configuration

Configuration lives at ~/.config/codex-proxy/config.json. Environment variables override all settings.

Server Settings

Env Var	Config Key	Description	Default
CODEX_PROXY_PORT	port	Port to listen on	8765
CODEX_PROXY_LOG_LEVEL	log_level	Logging level (DEBUG, INFO, WARNING, ERROR)	DEBUG
CODEX_PROXY_DEBUG	debug_mode	Enable debug mode (logs raw requests)	true

Authentication Settings

Env Var	Config Key	Description	Default
CODEX_PROXY_ZAI_API_KEY	z_ai_api_key	Z.AI API key	-
CODEX_PROXY_GEMINI_API_KEY	gemini_api_key	Google AI Studio API key	-
CODEX_PROXY_GEMINI_CLIENT_ID	client_id	Override OAuth client ID	Built-in
CODEX_PROXY_GEMINI_CLIENT_SECRET	client_secret	Override OAuth client secret	Built-in

API Endpoints

Env Var	Config Key	Description	Default
CODEX_PROXY_ZAI_URL	z_ai_url	Z.AI endpoint	https://api.z.ai/api/coding/paas/v4/chat/completions
CODEX_PROXY_GEMINI_API_INTERNAL	gemini_api_internal	Gemini internal API	https://cloudcode-pa.googleapis.com
CODEX_PROXY_GEMINI_API_PUBLIC	gemini_api_public	Gemini public API	https://generativelanguage.googleapis.com

Model Settings

Config Key	Description	Default
models	Available models (any model accepted if empty)	[]
compaction_model	Model for context compaction	First configured model
fallback_models	Map model names to fallbacks	{}
model_prefixes	Map model prefixes to providers	{"gemini": "gemini", "glm": "zai", "zai": "zai"}

Reasoning Settings

Config Key	Description	Default
reasoning_effort	Default reasoning effort level	medium
reasoning.effort_levels	Custom effort levels with budgets	Pre-configured levels

Timeout Settings

Config Key	Description	Default
request_timeout_connect	Connection timeout (seconds)	10
request_timeout_read	Read timeout (seconds)	600
compaction_temperature	Temperature for compaction requests	0.1

Example Config

{
  "port": 8765,
  "log_level": "INFO",
  "z_ai_api_key": "your-z-ai-key",
  "gemini_api_key": "your-google-ai-studio-key",
  "compaction_model": "glm-4.6",
  "fallback_models": {
    "gemini-3-pro-preview": "gemini-2.5-flash",
    "glm-4.7": "glm-4.6"
  },
  "reasoning": {
    "default_effort": "medium"
  },
  "model_prefixes": {
    "custom-model": "gemini"
  }
}

Available Reasoning Effort Levels

none, minimal, low, medium, high, xhigh. You can customize budgets and levels in reasoning.effort_levels config if needed.

Documentation

• CONTRIBUTING.md - Development guide and contribution process
• AGENTS.md - Deep technical context for AI assistants

License

MIT License - see LICENSE