Implement Conversational Search

[rajveer43]

Description

I've been thinking about enhancing our search experience to make it more user-friendly and conversational. I propose implementing a chatbot-like feature that allows users to ask questions in natural language about our products, instead of the traditional keyword-based searches.

Technical Details

We can integrate the Open Source Model to handle natural language queries and extract the key information needed for effective searches. The backend logic will need to be designed to interpret the user's intent and dynamically fetch relevant product data from our API

Imagine users being able to ask things like "What are the latest tech gadgets?" or "Can you recommend a good laptop for gaming?" and getting accurate results.

[rajveer43]

You can assign it to someone else as well.

[rajveer43]

let me know on how you want this feature to be built. Lets discuss approach first.
Approach can be.

1. build a small prototype on this in Notebook
2. Train the system to classify user queries into specific intents, such as buying, comparing, or exploring.
3. Enable the system to remember previous interactions and use that context to provide more accurate responses.
4. Use a RLHF method to make it human interactive,

[rajveer43]

Ok so, I will start working on it, as you have approved.

[rajveer43]

@khushi2706 there is a module of a langchain where a function can access directly API of the products that we are using in our project. what I think is it could be a good solution where conversational search will direclty interact with APIs, and then fetch relevant information based on the user's NLP query.

Consider it as ReAct + RAG = Reasoning + Action + Retrieval augmented generation(QnA) from API.

the two files I added before were a partial solution, but the above solution seem to be appropriate.