ch.at provides a universal, privacy-focused chat interface powered by large language models, accessible via HTTP, SSH, DNS, and an OpenAI-compatible API. It targets users seeking a lightweight, self-hostable, and tracker-free way to interact with AI, offering a single binary with no JavaScript dependencies.

How It Works

The service is built in Go with minimal dependencies, compiling into a single static binary. It prioritizes privacy by design, avoiding authentication, user tracking, and server-side conversation storage. Configuration is primarily done by editing source code constants and recompiling, offering a unique approach to customization and control.

Quick Start & Requirements

• Install/Run: Build from source (go build -o chat .) and run (sudo ./chat for default ports, or ./chat with configured high ports).
• Prerequisites: Go toolchain for building. For HTTPS, cert.pem and key.pem are required. LLM API keys (OpenAI, Anthropic, Ollama) must be configured in llm.go.example.
• Setup: Editing llm.go and building takes minutes. Running on privileged ports (80, 443, 22, 53) requires sudo.
• Docs: ch.at README

Highlighted Details

• Supports multiple access protocols: HTTP (with or without JavaScript), SSH, DNS tunneling, and an OpenAI-compatible API.
• Privacy-centric: No accounts, logs, or tracking; conversation history stored client-side only.
• Configurable via source code constants for ports and other parameters.
• Deployable as a Nanos unikernel for minimal VM footprint.

Maintenance & Community

The project is maintained by deep-ai-inc. Contributions are encouraged if they increase accessibility, are under 50 lines, and are deemed necessary.

Licensing & Compatibility

MIT License. Permissive for commercial use and closed-source linking.

Limitations & Caveats

DNS responses are limited to ~500 bytes and may time out for complex queries. HTTP and DNS traffic are not encrypted. Conversation history is capped at 64KB.