Skip to content

Latest commit

 

History

History
70 lines (46 loc) · 2.17 KB

README.md

File metadata and controls

70 lines (46 loc) · 2.17 KB

HuberGPT

This app enables AI-powered search for Andrew Huberman.

This is also a testbed for exploring Langchain functionality and building a production-ready app.

This app currently has the following features:

  • GPT-3.5 based QA search over Huberman's content
  • Giving source transcripts for the answers
  • Login authentication with NextAuth
  • Database support with Pinecone for embedding and MySql for queries / user login

Future features that I would like to include:

  • User profiles for better prompts and customized results
  • Allow multiple modes of search (highlevel, detailed, proscons, etc.)
  • Show most popular questions and answers per week
  • Allow users to add questions to Huberman
  • Cleaner settings:
    • Authentication
    • Profile
    • Search parameters
    • Action Buttons
  • Cron job to give updates on app activity
  • Cache for Db queries

Services Used

Dataset

Get all videos from

Trascribe all episodes with Whisper.

Transcribed data is split / embedded (Pinecone) with Langchain.

All steps outlined in: scripts/get_data_huberman.ipynb TODO: add script to automate this process.

Search

Use Langchain VectorDBQAChain to:

  • Embed the user query
  • Perform similarity search on Pinecone embeddings
  • Synthesize the answer from relevant chunks with GPT 3.5

Search

Relevant chunks with metadata (links) are displayed as source documents.

This builds on the excellent UI from https://github.com/mckaywrigley/wait-but-why-gpt.

Deploy

Note: the app that supports streaming is deployed to vercel.app: https://huber-gpt.vercel.app/

Credits

Thanks to Mckay Wrigley for open-sourcing his UI.

Thanks to Andrew Huberman for the excellent podcast.