Pearl is your assistant. She is 100% private and runs locally on your computer.
- Private AI chats
- Dynamic context switching based on keywords
- Interactive install for dependencies, making setup simple
- Saving chats to disk is optional, supporting privacy
- Offline use is fully supported, assuming dependencies are installed
- Sends no data outside the local machine by default
- Comes with 5 themes
- Supports file uploads
- Ollama (with at least one LLM - llama3.2:latest is currently the most compatible)
- Kokoro Fast-API for optional TTS integration
- Docker (a dependency for Kokoro, optional)
- 4-Core CPU
- 8GB RAM
- Linux Ubuntu 18.04 or later, or Window 10/11
- 50GB+ Disk Space recommended
Pearl can be built from source files via Nuitka & bundled into an AppImage on Linux:
pip install -r requirements.txt
nuitka \
--standalone \
--onefile \
--remove-output \
--output-dir=dist \
--include-package=pyttsx3.drivers \
--include-data-dir=defaults=defaults \
--include-data-files=CHANGELOG.md=CHANGELOG.md \
--include-data-files=README.md=README.md \
pearl.py
Set AppRun inside Pearl.AppDir as executable prior to building the AppImage
Copy pearl.bin into the deepest subdirectory of Pearl.AppDir
Run this command from the app's root directory:
appimagetool Pearl.AppDir Pearl-Linux.AppImage -v
On Windows, Pearl can also be built via Nuitka
IMPORTANT: Must have Visual Studio Build Tools installed with C++ / MVSC bindings
pip install -r requirements.txt
nuitka --standalone --msvc=latest --onefile --remove-output --enable-plugin=tk-inter --windows-disable-console --output-dir=dist --include-package=pyttsx3.drivers --windows-icon-from-ico=defaults/assets/pearl.ico --include-data-dir=defaults=defaults --include-data-files=CHANGELOG.md=CHANGELOG.md --include-data-files=README.md=README.md pearl.py
Report bugs via GitHub or to bugs@phillipplays.com.
Pearl integrates third-party artificial intelligence models selected by the user. The developer of this software is not responsible for the accuracy, appropriateness, legality, or safety of any output generated by these AI models. All responses from the AI are provided "AS-IS" with no warrenties of any kind, express or implied. Selecting an AI model, configurings its safeguards, and ensuring safe interaction are entirely the user's responsibility. By using Pearl, you acknowledge that any consequences arising from AI output are solely your own.