LLM Functions

This project allows you to enhance large language models (LLMs) with custom tools and agents developed in bash/javascript/python. Imagine your LLM being able to execute system commands, access web APIs, or perform other complex tasks – all triggered by simple, natural language prompts.

Prerequisites

Make sure you have the following tools installed:

argc: A bash command-line framewrok and command runner
jq: A JSON processor

Getting Started with AIChat

1. Clone the repository:

git clone https://github.com/sigoden/llm-functions

2. Build tools and agents:

Create a ./tools.txt file with each tool filename on a new line.

get_current_weather.sh
execute_py_code.py
search_tavily.sh

Create a ./agents.txt file with each agent name on a new line.

coder
todo

Run argc build to build tools and agents

3. Configure your AIChat:

Symlink this repo directory to AIChat functions_dir:

ln -s "$(pwd)" "$(aichat --info | grep -w functions_dir | awk '{print $2}')"
# OR
argc install

4. Start using your functions:

Now you can interact with your LLM using natural language prompts that trigger your defined functions.

Writing Your Own Tools

Writing tools is super easy, you only need to write functions with comments.

llm-functions will automatically generate binaries, function declarations, and so on

Refer to ./tools/demo_tool.{sh,js,py} for examples of how to use comments for autogeneration of declarations.

Bash

Create a new bashscript in the ./tools/ directory (.e.g. may_execute_command.sh).

#!/usr/bin/env bash
set -e

# @describe Runs a shell command.
# @option --command! The command to execute.

main() {
    eval "$argc_command" >> "$LLM_OUTPUT"
}

eval "$(argc --argc-eval "$0" "$@")"

Javascript

Create a new javascript in the ./tools/ directory (.e.g. may_execute_js_code.js).

/**
 * Runs the javascript code in node.js.
 * @typedef {Object} Args
 * @property {string} code - Javascript code to execute, such as `console.log("hello world")`
 * @param {Args} args
 */
exports.main = function main({ code }) {
  return eval(code);
}

Python

Create a new python script in the ./tools/ directory (e.g., may_execute_py_code.py).

def main(code: str):
    """Runs the python code.
    Args:
        code: Python code to execute, such as `print("hello world")`
    """
    return exec(code)

Writing Agents

Agent = Prompt + Tools (Function Callings) + Knowndge (RAG). It's also known as OpenAI's GPTs.

The agent has the following folder structure:

└── agents
    └── myagent
        ├── functions.json                  # Function declarations file (Auto-generated)
        ├── index.yaml                      # Agent definition file
        ├── tools.txt                       # Reuse tools
        └── tools.{sh,js,py}                # Agent tools script

The agent definition file (index.yaml) defines crucial aspects of your agent:

name: TestAgent                             
description: This is test agent
version: 0.1.0
instructions: You are a test ai agent to ... 
conversation_starters:
  - What can you do?
documents:
  - files/doc.pdf

Refer to ./agents/todo-{sh,js,py} for examples of how to implement a agent.

License

The project is under the MIT License, Refer to the LICENSE file for detailed information.

3.8 KiB Raw Blame History Unescape Escape