Soju06/codex-lb

What it solves

It solves the problem of managing multiple ChatGPT accounts to avoid rate limits and track usage. Instead of manually switching between accounts, users can pool them into a single proxy that distributes requests across the available accounts.

How it works

The project acts as a load balancer and API proxy. It provides OpenAI-compatible endpoints that allow any standard OpenAI client to connect to it. It manages a pool of accounts, tracks token usage and costs, and provides a web-based dashboard for administration. It can be deployed via Docker or run locally using uvx.

Who it’s for

It is designed for developers and power users who utilize multiple ChatGPT accounts and want a centralized way to manage them, track their spending/quotas, and integrate them into various AI clients like Codex CLI, OpenCode, or the OpenAI Python SDK.

Highlights

• Account Pooling: Distributes API requests across multiple accounts to maximize availability.
• Usage Tracking: Monitors per-account tokens, costs, and 28-day usage trends.
• API Key Management: Allows setting rate limits based on tokens, cost, window, and model for specific keys.
• OpenAI Compatibility: Works with any OpenAI-compatible client and supports specific endpoints for Codex and OpenCode.
• Management Dashboard: Includes a secure dashboard with password and optional TOTP authentication.