Local AI Assistant

Local-first personal AI assistant foundation for Windows + Ollama.

Run

python app.py

Default model: qwen3:14b via Ollama at http://localhost:11434.

Voice mode is push-to-talk and optional:

python app.py --voice

Set ENABLE_VOICE=true and configure PIPER_VOICE in .env for full STT/TTS. Install optional extras as needed:

python -m pip install -e ".[voice,browser]"
python -m playwright install chromium

Commands

• /mode study
• /mode productivity
• /mode research
• /mode interview
• /modes
• /memory stats
• /memory list
• /memory search <query>
• /memory show <id>
• /memory delete <id>
• /memory export
• /reset
• /exit

Memory commands can also be used naturally:

python app.py --once "remember My test phrase is banana rocket 947"
python app.py --once "What is my test phrase?"

Coding assistant commands:

python app.py --once "analyze repo ."
python app.py --once "explain error Traceback TypeError: bad operand"
python app.py --once "coding task refactor the router"

Architecture

• core: settings, logging, constants, exceptions
• brain: Ollama client and session orchestration
• router: intent classification and dispatch
• tools: pluggable tool registry with placeholder capabilities
• voice: push-to-talk microphone, faster-whisper STT, Piper TTS
• security: permission levels and confirmation policy
• memory: SQLite-backed durable memories with bounded retrieval
• coding: repository indexing, error analysis, and coding context helpers
• modes: configuration-driven assistant behavior modes
• shared: DTOs and utilities