AI Q/A from Openai (RAG)

Working on this project was very fun and taught me the basics of embeds. This inadvertently opened a pandora box in sentence encoders, decoders, and other large language models.

The idea for this app was simple.

Input a text document and when people come in with questions for the community, it would use DaVinci to read off the document to provide answers. This would engage customers and community users to answer simple questions *as long as your docs are up to date :) *

  • Set your training document train_faq

  • Ask the questions answer_faq

Its pretty awesome what you can make with a docker image slapping on FastApi and a talented developer working along side you @Zane

Improvements?

  1. Text upload instead of copy and paste
  2. If it gets any bigger, chunk the texts into postgres and utilize pinecone as a vector database
  3. Actually find a customer who would pay for this feature. 😔