lfzk550/fanzha-ai-proxy

What it solves

It allows users to access the "National Anti-Fraud AI" assistant through a standard OpenAI-compatible API. This enables the AI to be integrated into popular third-party AI clients and development frameworks that expect the /v1/chat/completions format.

How it works

The project acts as a high-performance reverse proxy. It translates standard OpenAI API requests into the specific format required by the National Anti-Fraud AI backend. It handles authentication via JWT tokens (Access and Refresh tokens) and supports streaming (SSE) outputs to ensure a smooth chat experience.

Who it’s for

Developers or users who want to use the National Anti-Fraud AI within their own apps or existing AI interfaces like NextChat, LobeChat, or OneAPI.

Highlights

• OpenAI Compatibility: Fully supports /v1/chat/completions and /v1/models specifications.
• Streaming Support: Implements Server-Sent Events (SSE) for real-time incremental text delivery.
• Session Maintenance: Includes a built-in mechanism to use Refresh Tokens for automatic session renewal (up to 90 days).
• Flexible Auth: Supports authentication via environment variables or dynamic request headers.