aurora-develop/aurora

What it solves

Aurora converts the internal backend capabilities of the ChatGPT Web interface into a standard OpenAI-compatible API. This allows developers to use ChatGPT Web's features—such as chat, file analysis, and image generation—within applications that expect the OpenAI API format, without needing official API keys for every feature.

How it works

The project acts as a proxy and translation layer. It takes requests sent to an OpenAI-style endpoint and translates them into requests that the ChatGPT Web backend can understand. To handle authentication and avoid detection, it manages a pool of accounts using various token types (access, refresh, session, and free UUIDs). Each account is assigned its own TLS client, proxy IP, browser fingerprint, and WebSocket connection to mimic real user behavior.

Who it’s for

Developers who want to integrate ChatGPT Web's full suite of capabilities (including multimodal features) into their own software using a standardized API interface.

Highlights

• Broad Feature Support: Supports chat completions, file-based Q&A, image generation/editing, speech-to-text (transcription), and text-to-speech (TTS).
• Tool Calling Simulation: Mimics function calling by using a custom text protocol (<tool_call>) and automatically injecting system prompts to convert model output into standard OpenAI tool call formats.
• Advanced Account Management: Features a robust system for managing account pools, including automatic token renewal and deduplication based on account IDs.
• Flexible Authentication: Supports multiple token sources, including OAuth refresh tokens, session tokens, and external access tokens provided at runtime.
• Infrastructure Ready: Provides built-in support for proxy pools and TLS to ensure stability and connectivity.