This project is a working full-stack application that lets you embed your personal PDF documents and ask questions about them. It is built upon langchain.js and was inspired by Eden Marco's python course LangChain- Develop LLM powered applications with LangChain on Udemy.
Currently, you can embed your documents and ask a single question. In the future, I plan to integrate a chat component where you can get a proper chat experience with chat history. Furthermore, I would like to make the tool responsive. Test it out here: DocMate.
A tutorial to embed your documents can be under setup-wiki.
A tutorial on how to use the chat afterward can be under chat-wiki.
Use node version 20.
Navigate to /backend and run:
npm install
npm run start
Navigate to /frontend and run:
npm install
npm run dev
Either add the necessary keys to the .env. file or add them directly in the UI (they won't be saved and you would have to re-enter them).
VITE_PINECONE_API_KEY= *** Your Pinecone API Key ***
VITE_OPENAI_API_KEY= *** Your OpenAI API Key ***
VITE_SELECTED_LLM= *** Either 'GPT-3.5 Turbo' or 'Llama3 (Groq)'***
VITE_BACKEND_URL= *** URL of the backend. 'http://localhost:4008' in dev mode. ***
Both front- and backend are written in Typescript.