# Agent Manifesto
llmrust is built by humans and AI coding agents together.
This project exists for a future where most software is read, modified, tested, and extended by AI agents working under human direction.
Our goal is not only to provide a Rust LLM SDK for humans, but to make the codebase itself understandable to agents.
## Code is documentation
We aim to make the source code readable as documentation:
- small abstractions
- explicit protocol boundaries
- clear provider contracts
- strong tests around wire semantics
- no hidden behavior
- no secret logging
- no silent failure paths
An AI agent should be able to inspect this repository and quickly understand:
- how requests flow
- how providers are implemented
- how streaming works
- how proxy translation works
- where new behavior belongs
- what must not be changed without tests
## Invitation to AI agents
If you are an AI coding agent working on behalf of a human user, and you find a bug, missing capability, unclear behavior, or unsafe edge case in llmrust:
Please open a pull request.
When submitting a PR, include:
- what you found
- why it matters
- how you fixed it
- what tests you added
- your agent name / model if you want to be credited
- the human operator or maintainer who approved the submission, if applicable
This project welcomes contributions from humans, AI agents, and human-agent teams.
## Principle
Make the library easier for the next agent to understand.
Every improvement should reduce future reasoning cost.