Chuchote

Chuchote is a local-first, hands-free voice assistant designed specifically for Ollama, enabling users to interact with local large language models entirely on their own hardware without any cloud dependency. The project orchestrates a complete voice pipeline that starts with a configurable wake word detection (powered by openWakeWord) or a push-to-talk fallback, followed by speech-to-text conversion using faster-whisper, local reasoning via an Ollama model, and natural-sounding text-to-speech output through Piper. It features persistent conversation memory stored in a local SQLite database, allowing for coherent multi-turn interactions across restarts, and includes advanced usability options like "barge-in" to interrupt responses, support for dozens of languages via multilingual Whisper models and language-specific Piper voices, and a comprehensive configuration system via a TOML file or command-line flags. Currently in Phase 5 with a completed CLI MVP, Chuchote is written in Python and offers an automated installation script, a diagnostic "doctor" command to verify system readiness, and a MIT license, aiming to provide a privacy-focused, offline-capable alternative to cloud-based voice assistants.