Skip to content

Latest commit

 

History

History
49 lines (38 loc) · 1.56 KB

README.md

File metadata and controls

49 lines (38 loc) · 1.56 KB
Raw

This is a repo for a human-interface LLM-powered QA Slack chatbot for answering questions related to DataTalks.Club Zoomcamps

Current version

Please follow this report that explains in detail how the bot works

First version

Behind the scenes

Course FAQ Google Document and the Course repo get indexed to the Pinecone vector store. Then semantic search retrieves the most similar (and hopefully most relevant) pieces to the question asked. Then this information is passed as a context to a conversational LLM to form the final answer.

Diagram

Before you start

Use any python virtual environment manager of your preference and install the dependencies from requirements.txt

Fill Pinecone index

  1. Set PINECONE_API_KEY and PINECONE_ENV environmental variables accordingly

  2. run ingest/ingest.py

    python ingest_mlops_old.py

Test QA pipeline locally

  1. Set OPENAI_API_KEY, PINECONE_API_KEY, and PINECONE_ENV environmental variables accordingly

  2. Run test.py

    python test.py

Launch the bot

  1. Set OPENAI_API_KEY, SLACK_APP_TOKEN, SLACK_BOT_TOKEN, PINECONE_API_KEY, and PINECONE_ENV environmental variables accordingly
  2. To launch the bot on the CLI run slack_bot/main.py 
    python main.py

Alternatively it can be launched with Docker, please follow this README