crates/asist-llm/src/lib.rs


1
2
3
4
5
6
7
8
9
10
11
12

//! llama-server client.
//!
//! It speaks the OpenAI chat API in streaming mode, because in a voice
//! assistant the answer is not waited for: it is split into sentences and
//! synthesized while the model keeps writing. Waiting for the whole answer
//! would add the model time to the synthesizer time instead of overlapping them.

pub mod chat;
pub mod client;

pub use chat::{Message, Role};
pub use client::{Delta, LlmClient, StreamOutcome};