Crate rig_cat 

rig-cat

An LLM agent framework built on comp-cat-rs.

No async, no tokio. All effects are Io<Error, A>, all concurrency is Fiber, all streaming is Stream. The entire runtime is synchronous composition with thread-based parallelism when needed.

Quick start

use rig_cat::provider::openai::{OpenAiCompletion, ApiKey, ModelName};
use rig_cat::agent::AgentBuilder;

let model = OpenAiCompletion::new(
    ApiKey::new(std::env::var("OPENAI_API_KEY")?),
    ModelName::new("gpt-4o".into()),
);

let agent = AgentBuilder::new(model)
    .preamble("You are a helpful assistant.")
    .temperature(0.7)
    .build();

let response = agent.prompt("What is a Kan extension?").run()?;
println!("{response}");

Modules

agent

Agent: the high-level LLM abstraction.

embedding

Embedding model trait and vector types.

error

Project-wide error type.

model

Completion model trait: the core LLM abstraction.

pipeline

Pipeline: composable RAG and multi-step workflows.

provider

LLM provider implementations.

tool

Tool trait: functions that agents can invoke.

vector_store

Vector store trait and in-memory implementation.