hjanuschka/pi-multi-pass

What it solves

This project addresses the common frustration of hitting rate limits when using AI services. It allows users to manage multiple subscriptions (e.g., multiple ChatGPT or Claude accounts) and automatically rotates between them when one account reaches its usage quota, ensuring uninterrupted access to AI models.

How it works

It functions as an extension for the pi CLI tool. Users add multiple OAuth accounts, which are then grouped into "pools" or "chains." When a rate limit error occurs, the tool detects it, marks the current account as exhausted, and seamlessly switches to the next available account in the pool or chain to retry the request. It supports various selection strategies, including round-robin, quota-aware prioritization, and custom JavaScript-based logic.

Who it’s for

It is designed for power users, developers, and researchers who rely heavily on LLMs and frequently encounter usage caps, as well as those who need to manage different AI accounts across work and personal projects.

Highlights

• Automatic Failover: Transparently switches accounts during a session when rate limits are hit.
• Smart Routing Strategies: Supports round-robin, quota-first (chooses the account with the most remaining usage), and scheduled (time-based) routing.
• Project-Level Affinity: Allows restricting specific accounts or pools to individual project directories.
• Model Presets: Enables creating named routing shortcuts (e.g., coding-premium) that can span across different providers.
• TUI Management: Provides a full suite of terminal commands to manage subscriptions, pools, and chains without manually editing configuration files.