About This Project • Key Features • Applications • Install From Scratch • Operations • How To Use • Developer Guide • Community • Credits • License
This project is for a fully customizable AI persona usable for streaming or private companionship. Feel free to download and use how you wish.
This software uses libraries from the FFmpeg project under the LGPLv2.1
- Configurable AI persona that integrates with applications for general interaction and streaming
- Text and talk with AI persona in real-time
- Support for various services and local models
- Support for custom contexts
Feel free to build and share your own! See the Developer Guide for more info.
Note To simplify setup across platforms, setup now uses conda. Conda is not necessary to run this project.
Create and enter a virtual environment with specific Python and pip version.
conda create -n jaison-core python=3.12 pip=24.0 -y
conda activate jaison-coreInstall PyTorch 2.5.1 with the right integration. Example below for computers with RTX graphics card.
conda install pytorch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 pytorch-cuda=12.4 -c pytorch -c nvidiaFor NVidia cards, ensure you have the latest drivers and CUDA toolkit
Install remaining dependencies.
If on Windows, please enable Developer Mode
pip install .
python -m spacy download en_core_web_smInstall FFmpeg
sudo apt install ffmpegbrew install ffmpegDownload executables and place them in the root folder:
- Download latest
ffmpeg-git-essentials.7z - Extract and copy all contents from
bin/to root of this project.
Configuration
FOR A FREE, 3RD PARTY T2T INTEGRATION: Use openai type but configure for use with Groq.
Add keys and other sensitive information for services you intend to use in .env (make a new file and copy the contents of .env-template).
For immediate setup using the example configuration, just provide the OpenAI API key.
Overall configuration can be done in configs/ and an example with all configurable options is located in configs/example.yaml. See Development guide for details on configuration.
Dealing with duplicate libiomp5md.dll.
It might not be necessary, but in case you encounter this error when running:
- Go to environment directory (where conda stores installed packages)
- Search for
libiomp5md.dll - Delete the version under package
torch
python ./src/main.py --helpExample usage: python ./src/main.py --config=example
See the specification for building applciations for Project J.A.I.son, creating custom integrations, and configuration tips below:
Join the community!