mishra-tapas / llm-pdf-chat

Chat with LLMs using PDFs as context! Experimental exploration: FastAPI + Streamlit + Langchain

llm-pdf-chat

FastAPI + Streamlit + Langchain!

Running Locally

Make sure to have poetry package manger installed locally.

Configure the .env file as per the .env.template

• Install all the dependancies locally:

  poetry install
  poetry shell # activate the virtual environment

• To start the FastAPI backend server:

  # first make sure you have a mongo db server up and running
  # and that you have set the connection string in your .env file
  cd server
  uvicorn main:app --reload
  # --reload if you are developing and changing the source files

  This would start running the API on your localhost:8000 and you can access the Swagger Docs at /docs endpoint.

• To start the streamlit web-interface:

  cd client
  streamlit run app.py

  This would start the streamlit app at localhost:8501.

Files Directory

Note: Files are shared between the streamlit and FastAPI code.

• It is expected that both the processes are running on the same machine.
• The environment variable, FILES_STORAGE_DIR is used by both the streamlit and FastAPI code.
• The files uploaded from the streamlit interface are stored in this directory, and are accessed by langchain running in the server code of FastAPI.
• Either write an absolute path, or the path must be wrt the parent of the client or server dirs. like ../uploaded_files/