A new AI large model API gateway, MIXAPI, addresses the challenge of unified management and distribution for diverse LLM APIs. It transforms various model interfaces into standard OpenAI, Claude, and Gemini compatible formats, serving individuals and enterprises. Offering a single executable, Docker image, and one-click deployment, it provides an open-source, self-controlled solution for API key management and secondary distribution.

How It Works

Built upon and integrating features from New-API and One-API, MIXAPI offers a comprehensive feature set. It emphasizes significant performance optimizations, claiming over 50% improvement in high-concurrency scenarios through AI-assisted code refinement in critical forwarding paths, aiming for efficient, high-throughput API aggregation and translation.

Quick Start & Requirements

• Installation: Deploy via single binary (Windows .exe recommended for ease), Docker image, or by building from Go source (go run main.go). Access the interface at http://localhost:3000.
• Prerequisites: Requires a local database (SQLite default, Docker uses SQLite) or remote MySQL/PgSQL. Go environment needed for source builds.
• Links: PRO version: https://github.com/aiprodcoder/MixAPI-PRO.

Highlighted Details

• Unified gateway for OpenAI, Claude, Gemini formats, supporting numerous LLMs (Gemini, Claude, Qwen, Kimi, etc.).
• Features a new UI, multi-language support, flexible billing (online recharge, token packages, per-use charging), channel weighting, model restrictions, Rerank models, and diverse authorization methods (LinuxDO, Telegram, OIDC).
• Robust token management includes usage limits (minute, day, total), package binding, and channel group controls.
• Supports channel retry and caching mechanisms (Redis, Memory).

Maintenance & Community

Community engagement is fostered via a WeChat group for discussions, customization, and consulting. Users are encouraged to star the repository to show support.

Licensing & Compatibility

The project is described as "completely open source" but lacks a specific license declaration. This ambiguity requires clarification for commercial use or integration into closed-source projects.

Limitations & Caveats

A separate "PRO" version targets enterprise needs (auditing, user registration), implying the open-source version may lack these specific enterprise-grade features. Session multi-round request recording is optimized to store only the last interaction, potentially losing conversational context. Input filtering specifically targets Chinese text and code characters.