specialagent 0.10.0

No project description provided

specialagent

A minimalist LLM-driven agent

An agent is just something that acts (agent comes from the Latin agere, to do). Of course, all computer programs do something, but computer agents are expected to do more: operate autonomously, perceive their environment, persist over a prolonged time period, adapt to change, and create and pursue goals.

Russel & Norvig, 1995

Designed with educational use in mind, this agent harness is intentionally only a few hundred lines of readable, dependency-free Python. It acts on its environment by running shell commands and receives output of those commands as percepts to drive subsequent actions.

Features

• Tool use. The agent has 4 tools:
  • exec - Run bash commands
  • write - Overwrites the contents of a file
  • replace - Find and replace in file
  • read_skill - Load a skill file
  • exit - Terminate session

Security

This package allows LLMs to directly execute shell commands. It should never be used without appropriate sandboxing.

Usage

uvx specialagent

or

pipx specialagent

Configuration

In its default configuration, the agent will attempt to use the model hosted at http://127.0.0.1:8080/v1/chat/completions (the llama.cpp default). Alternatives may be provided by env variables. For example, to use Gemma 4 31B hosted on AI Studio:

export LLM_BASE_URL=https://generativelanguage.googleapis.com/v1beta/openai/chat/completions
export LLM_API_KEY=your-api-key-here
export LLM_MODEL=gemma-4-31b-it