bots: Making LLM Tool Use Convenient and Powerful

Overview

bots (bɒts), n.pl. : Language Models which are instruct-tuned, have the ability to use tools, and are encapsulated with model parameters, metadata, and conversation history. The bots library provides a structured interface for working with such agents, aiming to make LLM tools more convenient, accessible, and sharable for developers and researchers. Foundation (bots.foundation) ————————— The core of the Bots library is built on a robust foundation: - Tool handling capabilities - any well-structured Python function can be used by a bot - Simple primary interface: bot.respond(), with supporting operations add_tool(s), save(), load(), and chat() - Tree-based conversation management:

  • Implements a linked tree structure for conversation histories

  • Allows branching conversations and exploring multiple dialogue paths

  • Efficiently manages context by only sending path to root

  • Enables saving and loading specific conversation states

  • Abstract base classes for wrapping LLM API interfaces into a unified “bot” interface

  • Pre-built implementations for ChatGPT and Anthropic bots

  • Complete bot portability - save and share bots with their full context and tools

Contents

Key Features

Auto Terminal (bots.dev.auto_terminal)

  • Advanced terminal interface for autonomous coding

  • Full conversation tree navigation (/up, /down, /left, /right)

  • Autonomous mode (/auto) - bot works until task completion

  • Tool usage visibility controls (/verbose, /quiet)

  • Save/load bot states for different tasks

  • Integrated Python and PowerShell execution

Tool System (bots.tools)

  • Standardized tool requirements:

    • Clear docstrings with usage instructions

    • Consistent error handling

    • Predictable return formats

    • Self-contained with explicit dependencies

  • Built-in tools for:

    • File operations (read, write, modify)

    • Code manipulation

    • GitHub integration

    • Terminal operations

  • Tool portability and preservation

Functional Prompts (bots.flows.functional_prompts)

  • Core operations: chain(), branch(), tree_of_thought()

  • Composable patterns for complex tasks

  • Iteration control (prompt_while, chain_while)

  • Support for parallel exploration

  • Parallel execution functions

API Reference

Indices and Tables