ToolHandler

A Python library for interacting with LLMs (Large Language Models) and function calling capabilities.

Features

• Integration with OpenAI-compatible LLM APIs, especially llama-3.3
• Function registration and tool calling system
• Optimization of the tool calling process with the selection of the relevant top N functions to be fed into the LLM prompt
• Support for conversation history and recursive function calls

Prerequisites

• Python 3.10+
• OpenAI API key
• Moralis API key

Create a .env file in the root directory and add your API keys. Look at the .env.example file for reference.

Directory Structure

• chat.py: Main file for the chat interface based on gradio
• llm.py: File for the Generic Tool Calling LLM class that enables tool registration and tool calling
• tools.py: File for the Tool Manager class that stores tool embeddings and select the top N tools to be fed into the LLM prompt related to the user query
• functions.py: File for the definition of the functions available for the LLM to call. It contains the actual functions that interact with the Moralis Solana API
• app.py: File for the initialization of the tool calling LLM and its integration with the Moralis Solana API
• requirements.txt: File for the dependencies
• .env.example: Reference file to create the .env file for the environment variables
• README.md: This file.

Installation

pip install -r requirements.txt

Usage

python chat.py

This will start the chat interface. Copy the URL of the interface and paste it in your browser (e.g., http://127.0.0.1:7860). Look at functions.py to formulate your queries based on the list of available functions.

Sample queries:

• "What is the balance of the account [account_address] on mainnet?"
• "What is the balance of the account [account_address]?" (Default network is mainnet)
• "What is the token price of [token_address]?"
• "I want to know the global token metadata for the contract [contract_address]"
• "What are the latest news on [topic]?" (Should not call any function)

Note: You can take a look at the terminal output to see the tool selection process (similarity scores, top N functions), tool calls, message history, raw outputs, etc.

Limitations

The focus is now on tool calling, the integration with the Moralis Solana API has just started.