RAG chatbot for nutrition science website

[pankaj2609]

Description

Create a chatbot (for a nutrition science website) that can help health conscious people to know the nutrient value for each food item that they desire to eat.

Features

• Try to answer the user query from the website data.
• If website data insufficient, explore the web and get the best possible answer.

Tech Stack

• Embedding Models: HuggingFace Embeddings
• Vector Database: ChromaDB, MongoDB
• Backend: Python(Flask)
• Dummy data from Kaggle
• Feel free to use services (other than above mentioned)

Resources

• MongoDB vector database
• Langchain for RAG chatbot
• HuggingFace embeddings

[Naman-Vasudev]

@pankaj2609 sir can you please what is meant by website here. For which website we have to make rag chatbot

[pankaj2609]

@Naman-Vasudev create a simple website, that shows some content related to nutrition science. For the same website, create a chatbot which can answer user query.

For instance, assume a blogging website (let say it has 10 articles)
The website should have a chatbot, where the user can ask their question.
Try searching for the answer to that question from those 10 articles first.
If answer found from these 10 articles, show that answer.
If no valid answer found from those 10 articles, the bot should be able to get the answer from the internet.

Try to explore langchain for appropriate tools to build this.

[Naman-Vasudev]

@pankaj2609 Sir, I can create a simple HTML website and can use AI for styling all; if that will do, assign the issue to me, and my approach will be the following:.
Create a website with 10 articles mentioning some basic foods and nutritional value.
For backend use Flask and my existing Claude API code (#96) that will run on local host this time rather than terminal.
and if data not on the HTML website, use Kaggle's data in the form of local excel file and then returning

[pankaj2609]

@Naman-Vasudev just one modification, we first try to look up in the local data (be it website data or the local excel file or both).
If that is insufficient, the bot should search on some search engine. (this is what we want to achieve)

Website data and excel data, both fall under the local data itself. We want that bot should be able to access the search engine results too.

[maheshaggarwal21]

@pankaj2609 please assign this issue to me

[pankaj2609]

Hey @maheshaggarwal21 . Good work.
I wanted to recommend some changes in the python script (line 74 onwards)

1. All the import commands, keep in the beginning of the code.
2. Try setting up the LLM outside the if condition (although the current code will also work fine).

Try to improve the html part, have a chatbot support on the site itself. U might need to write some Javascript code for it. Do give it a try.

[maheshaggarwal21]

@pankaj2609 please check the pr now