LLM provider abstraction for clawft.

This crate provides a unified interface for calling LLM APIs via OpenAI-compatible endpoints. It is a standalone library with no dependencies on other clawft crates.

Architecture

• [Provider] trait defines the chat completion interface
• [OpenAiCompatProvider] implements it for any OpenAI-compatible API
• [ProviderRouter] routes model names (e.g. "openai/gpt-4o") to providers
• [LlmProviderConfig] describes how to connect to a provider

Quick Start

use clawft_llm::{ProviderRouter, ChatRequest, ChatMessage};

let router = ProviderRouter::with_builtins();
let (provider, model_name) = router.route("openai/gpt-4o").unwrap();

let request = ChatRequest::new(model_name, vec![
    ChatMessage::system("You are a helpful assistant."),
    ChatMessage::user("What is Rust?"),
]);

let response = provider.complete(&request).await?;
println!("{}", response.choices[0].message.content);