imp-llm

imp-llm is the provider and streaming client layer used by imp.

It contains the model/provider abstraction, streaming response handling, OAuth/API-key auth support, model metadata, and provider-specific request/response code for the imp coding agent.

What this crate provides

• streaming LLM client primitives
• provider implementations for Anthropic, OpenAI, Google, and compatible APIs
• model registry and model alias handling
• OAuth and API-key authentication helpers
• response/event normalization for higher-level agent runtimes
• provider-specific error/context handling

Intended use

Most users should install and run the imp binary rather than depend on this crate directly:

brew tap kfcafe/tap && brew install imp
imp

Use imp-llm directly if you want the lower-level Rust client pieces without the full agent loop, TUI, sessions, and tools from imp-core / imp-cli.

Status

This crate is an active surface of imp, but APIs may still evolve as provider behavior, model capabilities, and the imp runtime change.

Repository

• Main README: https://github.com/kfcafe/imp
• Crate: https://crates.io/crates/imp-llm