Qwerty v3 — Tool Agent with Optional LLM Brain

Pure Python stdlib. Zero dependencies. Optional AI via Ollama.

pip install qwerty-agent → qwerty

How It Works

You → Qwerty TUI
       │
       ├── LLM Brain (Ollama) ← if installed
       │     └── understands you, calls tools, generates response
       │
       └── Direct tools ← fallback
             ├── /web, /wiki, /run, /read, /write
             ├── /search, /find, /list
             └── text normalization + fuzzy matching

Without Ollama: Qwerty works as a direct tool agent (commands like /web, /run, /read). With Ollama: Qwerty becomes an AI agent that understands natural language and decides when to use tools.

Quick Start

pip install qwerty-agent
qwerty "explain what a kernel is"

For AI mode:

# Install Ollama: https://ollama.com
ollama pull qwen2.5:7b
qwerty

Features

• 12 tools: read, write, edit, replace, append, list, search, find, run, web search, wikipedia, learn
• LLM Brain: optional Ollama integration — understands natural language, calls tools autonomously
• Text normalization: shorthand expansion, contraction handling, character dedup, fuzzy matching
• Constitution safety: blocks dangerous commands, secrets leaks, unauthorized file overwrites
• Persistent memory: learned facts survive across sessions
• Beautiful TUI: colors, dividers, timestamps, tab completion, command history
• No dependencies: pure Python stdlib — runs on any system with Python 3.10+

Commands

Architecture

qwerty-agent/
├── qwerty_agent/
│   ├── agent.py        # Tool implementations + process()
│   ├── brain.py        # Ollama LLM connector
│   ├── interface.py    # TUI (colors, history, tab completion)
│   ├── protocol.py     # JSON-RPC / MCP server
│   └── constitution.json  # Safety rules
├── test_all.py         # 46 tests
├── pyproject.toml
└── README.md

Requirements

• Python 3.10+
• No pip packages needed
• Ollama (optional) for AI mode

Tests

python test_all.py