Summary

codex-lb addresses managing multiple ChatGPT accounts via a load balancer and proxy. It targets developers and power users needing centralized API access, usage tracking, and cost control, offering a unified, OpenAI-compatible interface.

How It Works

This project acts as a proxy pooling multiple ChatGPT accounts, distributing requests to balance load and optimize utilization. It features per-account usage tracking (tokens, cost), granular API key management with rate limits, and a web dashboard for monitoring and configuration. Its advantage lies in abstracting LLM API endpoint complexity behind a single, consistent interface.

Quick Start & Requirements

The recommended installation is via Docker:

docker volume create codex-lb-data
docker run -d --name codex-lb \
  -p 2455:2455 -p 1455:1455 \
  -v codex-lb-data:/var/lib/codex-lb \
  ghcr.io/soju06/codex-lb:latest

Access the dashboard at http://localhost:2455 to add accounts. Clients point to http://127.0.0.1:2455/v1 or http://127.0.0.1:2455/backend-api/codex. API key authentication can be enabled via the dashboard.

Highlighted Details

• Account Pooling: Distributes requests across multiple ChatGPT accounts.
• Usage Tracking: Detailed per-account token consumption, cost analysis, and 28-day trends.
• API Key Management: Keys with configurable rate limits (tokens, cost, time, model).
• OpenAI Compatibility: Endpoints compatible with Codex CLI, OpenCode, OpenClaw, and OpenAI Python SDK.
• Dashboard: Web interface for account management, API key generation, and monitoring.
• WebSocket Support: Optional native upstream WebSocket transport for streaming.

Maintenance & Community

The project lists several contributors, including Soju06 and Jonas Kamsker. It follows the all-contributors specification, welcoming contributions. No specific community channels (e.g., Discord, Slack) are linked in the README.

Licensing & Compatibility

The project's license is not explicitly stated in the README, requiring clarification for adoption, especially concerning commercial use. It is designed for compatibility with OpenAI-compatible clients.

Limitations & Caveats

A significant caveat is the absence of a stated license. Some WebSocket features are experimental, and upstream Codex support for stable WebSocket APIs is noted as lacking. Reverse proxies must be configured to handle WebSocket upgrades.