[Question] Public API endpoint for client applications?

[alreshi]

Summary

I'm building a web application that needs to send AI inference requests to the Cocoon network. I'd like to understand the recommended way for client applications to connect.

My Setup

• Application: Next.js web app with chat interface
• Integration: Using OpenAI-compatible API format (/v1/chat/completions)
• Environment: macOS (development), will deploy to cloud

What I've Tried

1. Downloaded the worker release from the repo
2. Attempted to build locally with ./scripts/cocoon-launch --local-all
3. Build fails on macOS due to assembly incompatibilities (BLST library)

Questions

1. Is there a public proxy endpoint that client applications can connect to for inference requests?

   • If yes, what is the endpoint URL?
   • What authentication/payment setup is required?

2. For developers without TDX/H100 hardware, what is the recommended way to:

   • Test real inference (not mock)?
   • Build production applications?

3. Is there a client-only package that doesn't require building the full worker infrastructure?

Environment Details

• macOS 14.x (Sonoma)
• Node.js 18+
• No access to Intel TDX or NVIDIA H100 hardware

Thank you for building this! Looking forward to integrating Cocoon into my application.

[kenorb]

Related:

• How to use the cocoon now as a developer? #18
• https://toncenter.com/api/v3/index.html
• https://cocoon.doge.tg/tools/transaction