Dragonfire

It uses offline speech recognition to process voice commands locally, preserving user privacy for command invocation while connecting to internet services only when the task requires external information.

The project was one of the early open-source desktop AI assistants targeting privacy-conscious users who wanted Siri or Cortana functionality without the cloud dependency.

The assistant handles a range of commands: web searches, weather lookups, Wikipedia queries, system application control, file management, calculation, and conversational interaction backed by a locally-running language model or rule-based response system.

Dragonfire's modular command system allows developers to extend its capabilities by defining new intent handlers in Python, making it a practical starting point for custom desktop automation tools tailored to specific workflows.

Its open-source codebase provides transparency into exactly what is and isn't sent to external services.

Privacy-focused users who want voice assistant capabilities without commercial cloud surveillance, developers building custom voice-controlled tools for specific applications, and tinkerers interested in DIY home assistant projects use Dragonfire as a foundation.

The project represents an earlier generation of local AI assistants that has been largely superseded in capability by local LLM runners like Ollama combined with faster offline speech recognition, but remains relevant as a complete, self-contained example of how to build a fully offline voice assistant pipeline in Python.