LLM Functions

This project enables the creation of LLM tools and agents using Bash, JavaScript, and Python, leveraging function calling to connect LLMs to custom code. It allows executing system commands, processing data, and interacting with APIs.

Tools Showcase

Agents showcase

Prerequisites

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

Getting Started with AIChat

The project is designed to work with AIChat. To get started, clone the repository, build tools and agents by creating tools.txt and agents.txt files, and then run argc build and argc check. Link the repository to AIChat using symlinks or environment variables.

Writing Your Own Tools

Tools can be created in Bash, JavaScript, or Python, with JSON declarations automatically generated from comments.

Writing Your Own Agents

Agents consist of prompts, tools, and documents, similar to OpenAI's GPTs. They have a specific folder structure including functions.json, index.yaml, and tools.txt.

MCP (Model Context Protocol)

• mcp/server: Allows LLM-Functions tools/agents to be used through the Model Context Protocol.
• mcp/bridge: Enables the use of external MCP tools by LLM-Functions.

Documents

• Tool Guide
• Agent Guide
• Argc Commands

License

The project is under the MIT License.