Go bot that wires GPT-4, DALL·E, and Whisper into Feishu

What it does This Go project is a chatbot bridge that drops OpenAI’s models—GPT-4, GPT-4V, DALL·E-3, and Whisper—directly into Feishu and its international counterpart, Lark. Users can hold multi-topic conversations in private or group chats, generate images from text, reason over uploaded pictures, and send voice messages, all inside the collaboration platform they already use.

The interesting bit The bot isolates conversation topics into threaded contexts with automatic timeouts, which keeps group chats from collapsing into one garbled prompt history. It also load-balances across multiple OpenAI API keys, a practical concession to teams hitting rate limits in production.

Key highlights

• Supports GPT-4, GPT-4V, DALL·E-3, and Whisper through a single Feishu bot interface
• Context-aware multi-topic conversations in both private and group chats
• Built-in scene presets and role-playing modes for templated interactions
• Multi-token load balancing and reverse-proxy support for production use
• Deployable via Docker, serverless functions, Railway, Repl.it, or local binaries

Caveats

• Several advertised features—table analysis, document-to-PPT conversion, Feishu document interaction, private data training, and admin mode—are marked as work-in-progress (🚧) in the README, so they may not be usable yet.
• The project is fundamentally integration code; most of the intelligence lives upstream in OpenAI’s APIs.

Verdict Worth a look if your team lives in Feishu/Lark and wants a quick, self-hosted way to share GPT-4 access without managing separate accounts. Skip it if you’re not on Feishu or if you need deep custom logic beyond what OpenAI’s API already provides.