Sansa-MP

A talking AI "friend" bot with timecode-driven spontaneous chatter. Part of the MProcs suite.

   ____                                    __  __  ____
  / ___|                                  |  \/  ||  _ \
  \___ \   __ _  _ __    ___   __ _       | |\/| || |_) |
   ___) | / _` || '_ \  / __| / _` | ____ | |  | ||  __/
  |____/ | (_| || | | || (__ | (_| ||____||_|  |_||_|
   \__,_||_| |_| \___| \__,_|

What it does

• Talks like a close human friend (persona seeded deterministically from a meta-value), powered by the DeepSeek chat API.
• Decides on its own when to jump in and chat: an algorithm reads the live clock's decimals through a seeded gate, so the rhythm of when it talks is decided purely by the timecode decimals.
• It pauses while you type: any current speech is cut off mid-word and no new remarks start; once your input is empty again (idle a couple of seconds), it resumes talking on its own.
• Voice output via the piper neural voice (amy), falling back to espeak.

Install

pip install sansa-mp

First run checks/installs what it needs interactively: python packages (piper-tts, requests), the piper CLI + alsa-utils (apt) and the amy voice model (~63 MB download).

Usage

sansa-mp

• On first run you paste your DeepSeek API key once; it is saved to ~/.config/friendbot/deepseek_api_key (or set the DEEPSEEK_API_KEY environment variable). Type k anytime to change it.

Credits

Sansa-MP is part of the MProcs suite.

License

MIT