Radar check
What we found when we opened it. Every bar is one fact, with its date.
Last code pushed 7 Jul 2026, 3 months before this check.
No releases published on GitHub; the code date above is the one to trust.
Public since May 2026, 5 months.
4,145 stars and 3 forks.
0 commits and 1 new issues in the last 90 days.
1 contributor. No pull requests merged in 90 days; changes go in directly. A small team carries it.
A security policy says how to report a problem. No published security advisories.
How the bars are scored
- Alive. Days since the last code push: a week or less 100, a month 90, three months 70, six months 50, a year 30, eighteen months 15, older 0.
- Releases. Age of the latest GitHub release, on a similar scale; no bar when the project publishes no releases.
- Maturity. Age of the project: eight years 100, five 90, three 70, two 55, one 40, six months 25, younger 10.
- Adoption. Stars on a log scale: 100,000 stars is 100, 10,000 is 80, 1,000 is 60, 100 is 40.
- Activity. Commits in the last 90 days: 300 or more 100, 100 85, 30 65, 10 45, any 25.
- Community. People who have contributed: 100 or more 100, 30 80, 10 60, 5 40, 2 25, one 10.
- Security. A published security policy 80, none 40; published advisories are listed, not scored.
Checked by Repo Radar on 28 Oct 2026 from the repository on GitHub.
In simple English
This program puts a small animated character on your computer screen. It watches what you do, listens through your microphone, and talks back to you with a computer voice. It can read your screen to understand your work and offer help when you need it. You need a computer with a fast processor and enough memory to run local models or an internet connection for cloud AI services. It is free to use and runs entirely on your own desktop.
What it is
An open-source desktop assistant built in Python that combines screen capture, speech recognition, voice synthesis, and a graphical interface. It runs as a floating overlay window that reacts to your workflow and can speak or display messages.
What it is good for
Solo builders who want a personalized, interactive desktop companion that watches active windows and provides proactive suggestions during deep work.
Summary (Key Points)from the README
Key features
- 60fps event loop: A central tick runs every 16ms, and the face renders at 60fps through PyQt6 QPainter.
- Event-driven modules: All modules talk only through the `EventBus` pub/sub broker, so no module depends directly on another.
- Semantic screen reading: `SemanticSensor` walks the UIA tree, scores extraction quality, and applies an adaptive cooldown.
- OCR fallback: When UIA fails, `VisionSensor` captures the screen with MSS and reads it with Tesseract.
- Flexible brain: `CognitiveEngine` runs on Groq's cloud API or a local llama-cpp model.
- Voice output: Piper ONNX TTS passes through a DSP chain (pitch shift, bitcrush, chorus) for a robotic voice.
- Speech input: `EarsSensor` uses Faster-Whisper with neural VAD for speech-to-text.
- Sentence-level streaming: `StateMachine` parses streamed LLM output into sentences as they arrive, and the README gives a 1.2s thinking delay before responses.
- Work Mode: `Ctrl+Shift+Alt+W` makes B monitor your task and offer help, staying silent when content isn't relevant.
- Global hotkeys: Four `Ctrl+Shift+Alt` shortcuts toggle the input box, speak mode, and Work Mode, and trigger the kill switch.
Ways to install
| Platform or method | What it needs | Effort |
|---|---|---|
| Windows 10/11 | Python 3.10+, `venv\Scripts\activate`, `pip install -r requirements.txt`, a Groq API key in `.env` (or a local GGUF model) | Primary platform, follows the README steps directly |
| Linux/macOS | Same steps, with `source venv/bin/activate` | Experimental, per the README |
Important
- Platform support: The README lists Windows 10/11 as primary and Linux/macOS as experimental.
- Exit: B has no standard close button, so the `Ctrl+Shift+Alt+Q` kill switch is the only way to exit.
- Hardware target: The README targets an Intel i5 quad-core CPU and 16 GB of RAM.
- API key: Cloud inference needs a Groq API key stored in `.env`, and requests go to Groq's cloud service.
- Local model size: Local llama-cpp inference needs a 4GB+ GGUF model, which must be downloaded separately because the `models/` folder is gitignored.
- OCR dependency: The OCR fallback needs Tesseract installed, and the README does not give setup steps for it.
- Global hooks: B registers system-wide hooks and an always-on-top, click-through overlay, so it runs outside normal app windows.
- Development status: The README marks the project as "Active Development," so expect changes.
Written by an AI model from the project’s README; not part of the review, and it can be wrong.
- Features a smooth 60 frames-per-second animated facial interface built with PyQt6.
- Includes privacy-first local processing options for speech and vision tasks.
- Provides a global kill shortcut to instantly terminate all system hooks if needed.
- Requires 16 GB of memory and a quad-core processor to run smoothly.
- Complex setup involving Python environments, audio packages, and optional API keys.
- Primarily designed for Windows with experimental support for Linux and macOS.
Traditional desktop notification widgets and static AI chat windows
Groq API for cloud-based language model inference
- abi/screenshot-to-code81k stars
Turns screenshots into code using vision models, while B observes desktop activity dynamically.
- openai/openai-python32k stars
Provides programmatic access to language models that power AI assistant backends.
- vllm-project/vllm96k stars
A serving engine that makes one graphics card answer many requests at once.
- infiniflow/ragflow93k stars
A question-answering system built on a document collection, with the sources shown beside every answer.
- basicmachines-co/basic-memory4.3k stars
Gives an AI assistant a memory that is stored as plain text files you can read.
git clone https://github.com/AhmadHassan-BTed/B.git cd B python -m venv venv source venv/bin/activate pip install -r requirements.txt python main.py