LangChain Documentation Helper

A repository for learning LangChain by building a generative ai application. Original work from Eden Marco.

This is a web application demo that use Pinecone as a vectorstore and answers questions about LangChain. Pinecone act a context database, first you have to download the docs from sources from LangChain official documentation. Obviously you can do the same with every dataset you want.

Basically you have three scripts in this project:

langchain_retrieval_doc_helper.py --> my work. Original work from Eden was not working for me. main.py --> it invokes the logic from core_agent.py core_agent.py --> basically is a refactor from first script.

Environment Variables

To run this project, you will need to add the following environment variables to your .env file

PINECONE_API_KEY PINECONE_ENVIRONMENT_REGION OPENAI_API_KEY

Run Locally

Clone the project

  git clone https://github.com/alonsoir/documentation-helper.git

Go to the project directory

  cd documentation-helper

Download LangChain Documentation

  mkdir langchain-docs
  wget -r -A.html -P langchain-docs  https://api.python.langchain.com/en/latest

Run ingestion script to create the index in Pinecone. https://app.pinecone.io My index is langchain-doc-index, be sure to match this name with the one you use in the code.

    python ingestion.py

Install dependencies

  pipenv install

Start the flask server

  streamlit run main.py

Running Tests

To run tests, run the following command

  pipenv run pytest .

🔗 Links