FAISS vector database has been used to create the embeddings for the data collected
Then using the langchain we are retreiving the best possible solutions for the user queries, combining it with gpt-3 providing user a proper output in a friendly manner
The response is then show using the Textbase chat bot UI available.
Improvement
we can use opensource Large Language Model's such as llamma- 7B 13b etc..
train for our specific business purpose, this would avoid the expense of using open-ai api for the each user, and reduce overall cost and be helpful in long term.