Skip to content

scarter84/0411

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

36 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Whim

Local-first desktop + mobile ecosystem for voice, AI, connectivity, and automation.

Whim is a Python desktop terminal and Android mobile companion that runs entirely on your own hardware. No cloud APIs required. No telemetry. ~2 MB of code.

Full Manual with Screenshots


What's in it

Component Description
Whim Terminal Tkinter desktop app with 16 tabs: AI chat, voice cloning, wake word, SmartThings IoT, Signal/Discord, screen share, archive editor, audio capture, and more
CURSOR tab Built-in code editor with file browser, syntax highlighting, and AI assist. Pick any local Ollama model from a dropdown, hit Explain/Fix/Refactor/Complete/Tests, stream the response, apply diffs
Whim.AI Streaming Ollama chat with presets, observability (tokens/s, VRAM, context meter), tool trace, export
AVR Lab XTTS v2 voice cloning with speaker references and spectrogram visualization
Voice Engine Wake word ("Hey Whim") with live spectrogram, HPF/AGC/parametric EQ, VAD, confidence ghost bar
Persona Coined response playlists per voice clone — pre-rendered WAV for <100ms playback
Whim.m Android mobile app: voice recorder, file library, device chat, AI chat via Ollama proxy
Networking Reverse SSH tunnel to VPS (primary) + Tailscale mesh VPN (fallback)

Stack

  • Python 3.12 + Tkinter (desktop)
  • Ollama (local LLM inference)
  • Coqui XTTS v2 (voice cloning, conda env)
  • FFmpeg (audio processing)
  • autossh + systemd (tunnel)
  • ADB (Android device management)

Quick Start

# 1. Clone
git clone https://github.com/scarter84/0411.git
cd 0411

# 2. Install AI models (optional — app works without them)
bash scripts/setup_models.sh --status   # check what you have
bash scripts/setup_models.sh --ollama   # install Ollama + pull llama3.1:8b-16k (~5 GB)
bash scripts/setup_models.sh --xtts     # install XTTS v2 voice synthesis (~10 GB)

# 3. Copy config templates
cp config/openclaw.example.json ~/.openclaw/openclaw.json
cp config/device_locations.example.json config/device_locations.json

# 4. Run the desktop terminal
python3 app/openclaw_tkui.py

# 5. Run the mobile server (optional)
python3 whim_m_v2.1.py --port 8089

Code vs. Weights

The repo is ~2 MB of code. All AI models install separately:

Component Size Install
Whim code ~2 MB git clone
llama3.1:8b-16k 4.9 GB ollama pull llama3.1:8b-16k
deepseek-r1:32b 19 GB ollama pull deepseek-r1:32b
XTTS v2 ~10 GB bash scripts/setup_models.sh --xtts

The app works without any models — AI tabs show connection errors but everything else functions normally.

It's not perfect

This is a one-person project built for daily use. There are rough edges. But the codebase is small enough that you can point Cursor, Factory droids, Aider, or any AI coding tool at it and they can reason about the whole thing.

Fork it. Assign a droid to it. Steal the parts you like.

License

MIT

About

ai backplane

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors