#generative-ai #ai #llm #generative #agent #language-model #nlp

ai-agents

a Rust library designed for building and managing generative AI agents, leveraging the capabilities of large language models (LLMs)

5 releases

0.2.3 Apr 18, 2024
0.2.2 Apr 16, 2024
0.2.1 Apr 1, 2024
0.1.4 Mar 15, 2024
0.1.3 Mar 15, 2024

#187 in Machine learning

Download history 189/week @ 2024-03-13 17/week @ 2024-03-20 136/week @ 2024-03-27 45/week @ 2024-04-03 96/week @ 2024-04-10 169/week @ 2024-04-17

446 downloads per month

MIT license

42KB
994 lines

ai-agents

This repository is a Rust library designed for building and managing generative AI agents, leveraging the capabilities of large language models (LLMs), such as ChatGPT. The aim of this project is to provide a robust and scalable framework that is adaptable to a wide range of scenarios.

ai-agents is at a very early stage of development.

Features

  • Structured Data Flow: Leverage PipelineNet for organized and efficient data flow between processing units, enabling complex data transformation and decision-making capabilities.
  • Flexible Architectures: Utilize dynamic flow control within PipelineNet to adapt AI agent behaviors.
  • Extendibility: Easily extend core functionalities with custom unit implementations.
  • Contextual Grouping: Organize units into coherent groups for focused execution, simplifying task management and enhancing processing clarity.
  • Asynchronous Support

Crates

  • ai-agent-macro
  • sllm-rs: A crate dedicated to interfacing with Large Language Models (LLMs), including utilities for sending requests and processing responses.

Examples

The following examples are simulations of limited situations, demonstrating the application of ai-agents to specific scenarios:

Run Examples

To run the examples, you need to set an environment variable OPEN_API_KEY with your API key. This can be done by creating a .env file in the root of the project.

OPEN_API_KEY=your_api_key_here
  • Find Treasure: A game simulation where the player's goal is to find treasure in a dynamically generated scenario by interacting with NPCs.

  • Ecommerce Chat Assistant: A limited simulation agent that, based on customer inputs (such as name and order ID), explains the current state of an order.

Dependencies

~11–23MB
~348K SLoC