Question answering over an index

[tardigrde]

Realize this example: https://python.langchain.com/en/latest/modules/chains/index_examples/vector_db_qa.html

Goal: I can provide (almost) any type of file or any URL, and have an educated discussion about the docs with the GPT model.

Usecase: I want to learn more about LangChain. If the above goal is satisfied, I can do the following:

• load LangChain documentation into a VectorStore (index),
• spin up the Q&A bot with the VectorStore as a retrieval source
• I can:
  • [ ] use GPT4 model
  • [ ] ask N number of questions from the bot
  • [ ] the answers reflect the current state of the docs (as seen in the index) and not the state from GPT snapshot in 2021

See repo for native implementation: https://github.com/tardigrde/docs-gpt-qa

[tardigrde]

@dil0gcc could you enable Github Discussions on this project? Or is that a paid feature?

I found a good starting point /alternative for what I outlined above: https://github.com/akshata29/chatpdf

Other resources I found after a quick lookaround: GPT4 awesome page https://gpt4.tools/

also similar to what I want to do https://www.youtube.com/watch?v=ih9PBGVVOO4

Lot of tutorials with examples https://github.com/gkamradt/langchain-tutorials/tree/main

[tardigrde]

this should be it : https://github.com/arc53/DocsGPT