CustardPie is an application that customizes Large Language Models (LLMs) based on user preferences. Users can select from a range of pre-trained LLMs, then personalize them further by adding their own documents in formats such as PDFs, .txt, and .docx. The application processes these inputs, fine-tuning the model as per the specified customizations, and provides the modified model file for download.
- LLM Customization: (Work in progress) Users can select their preferred LLM model type.
- Document & Website Integration: Upload documents in
.pdf,.txt, or.docxformats or specify websites for model tuning. - Personalized Model Output: (Work in progress) Receive a tailored LLM file based on the provided customizations.
Ensure you have the following installed:
- Node.js (v20 or above)
- Python (v3.10 recommended)
pnpm/yarn/npm(use any based on your preference)
Clone the repository and navigate to the ui directory to install dependencies.
cd ui
pnpm install-
Backend:
pnpm run server
-
Frontend:
pnpm run dev
To install the required Python packages, run:
pip install -r requirements.txtchromadb==0.4.24
langchain-huggingface
llama-index-llms-ollama
llama-index-vector-stores-chroma
llama-index-embeddings-langchain
llama-index
langchain
langchain-community
Flask
- Launch the backend and frontend servers.
- Access the application from your browser.
- Choose your preferred LLM model type (Work in progress).
- Upload your documents in
.pdf,.txt, or.docxformats. - Download your customized LLM model (Work in progress).