Crate apprentice_lib

Apprentice-lib is a library that allows to create agent applications. It allows to create a chat with an LLM and use tools/functions.

Features

• several providers
• light-weight
• configurable
• extensible

Providers

• Anthropic (Claude models)
• OpeanAI (GPT models)
• Google Cloud Platform (Gemini)

Examples

use apprentice_lib::llm::{get_llm_chat, Message, Role};
use apprentice_lib::tools::ToolChoice;
use apprentice_lib::request::get_reqwest_client;
use apprentice_lib::ModelProvider;
use apprentice_lib::Config;

let config = Config::new(ModelProvider::OpenAI, "gpt-4".into(), "<api-key>".into(), "https://api.openai.com/v1/chat/completions".into());
 
let reqwest_client = get_reqwest_client().expect("transport created");
 
let mut chat = get_llm_chat(config, reqwest_client, vec![]).expect("chat created");
 
chat.set_system_prompt("You are a helpful assistant.".into());
 
let user_message = Message::text(Role::User, "Hi assistant!".into());
 
let response = chat.get_inference(&[user_message], ToolChoice::None).expect("LLM response");

for message in response.iter() {
    match message {
        Message::Text(text) => { /* process text message */ }
        Message::ToolCall(tool_call) => { /* process tool use request */ }
        Message::ToolResult(_) => { panic!("LLM must not respond with tool result!") }
    };
}

Modules

llm

LLM interface

request

API transport.

tools

Agent tools.

Macros

val_as_str

Interpret value as str

Structs

Config

Model parameters.

Enums

Error

App errors.

ModelProvider

Model providers.