Expand description
LLM Provider abstraction — swap backends without changing agent logic. Enable only what you need via Cargo features.
Re-exports§
pub use traits::ChatMessage;pub use traits::ChatRequest;pub use traits::ChatResponse;pub use traits::Provider;pub use traits::ToolCall;
Modules§
- anthropic
- Anthropic (Claude) provider implementation. Supports both API keys and OAuth tokens (Claude.dev)
- defaults
- Default model per provider.
- oauth
- OAuth token support for Anthropic Converts Claude.dev OAuth tokens (oat01) to API calls via token exchange
- ollama
- Ollama provider — local model support via Ollama API.
- openai_
compat - OpenAI-compatible provider — works with OpenAI, OpenRouter, Groq, Together, etc.
- retry
- Shared provider retry helper for transient network failures.
- rs_ai
- rs_ai-backed provider adapter for first-class ChatGPT, Gemini, xAI, Claude, Cloudflare and generic OpenAI-compatible endpoints.
- shared_
credentials - Shared OAuth credential store, common to every tool built on
rs_ai. - traits
- Core Provider trait — defines the interface for LLM backends. Inspired by ZeroClaw’s trait system with simplifications.