Installation
Add TypedLM from GitHub, pick the features you need, and check the requirements.
TypedLM is not on crates.io yet. Add it from the repository:
cargo add typedlm --git https://github.com/casoon/typedlm
cargo add serde --features derive
cargo add schemars
cargo add tokio --features macros,rt-multi-threadYour own types derive Serialize, Deserialize and JsonSchema, so serde and schemars are
direct dependencies of your crate. Any async runtime works; the examples use tokio.
Features
| Feature | Default | Adds |
|---|---|---|
derive |
yes | #[derive(TypedLm)] |
http |
yes | OpenAiCompatible, using reqwest with rustls and the tokio timer |
eval |
yes | Dataset, metrics and evaluate |
optimize |
no | optimizing worked examples and instructions |
Without default features the crate depends on serde, serde_json, schemars and tracing only,
and on no async runtime. Bring your own Provider implementation in that case.
Requirements
| Minimum | |
|---|---|
| Rust | 1.85 (edition 2024); 1.88 with the optimize feature |
| Endpoint | any OpenAI-compatible chat completions API, or your own Provider |