codex

To Replicate:

You will be needing Exa Ai API which searches over the internet for the required ursl to gain information.
You will also need Gemini API and OpenAI API for embedding and pushing the contents into PineCone, Upstash for which you will be needing Pinecone API and Upstash API for pushing the embeddings into the vector databases for utilizing the context in generating reponses .
streamlit run doc.py and streamlit run app.py for making the two streamlit apps and then you can use the rest as you would use a web app.

Inspiration

As a geek, we always try to explore some websites for research purposes or documentation pages for understanding tools but sometimes it takes so much time just to understand the content in it so we thought if there is a companion with us who can help us with understanding stuff it would be great so we came up with this tool.

What it does

It has many features on the way that you can interact with a website or page

AI-powered Browser
- With this feature whenever you search for something we return some web pages. not only that it has some context once you search so whenever you do some search related to the old topic it uses cache memory and helps you with that. The more user uses the tool the better it gets.
Chat with webpage
- So with this feature, you can use the thing that's shown on a page and chat with the webpage. It uses the content from the site and using AI you can interact and chat with that website
CodeGenerator
- Using our tools you can also generate code of any type that you want which is very fast

How we built it

We used Open source LLM's like Gemini
Used Pinecone & up stash vector databases
- We used them to store information which makes the tool better. The more people use this service the better it gets
Auth0 for authentication stuff

Challenges we ran into

Integrating Vector DB felt hard but the journey was awesome
using gemini , llama-index was a pain
and godaddy's coupon code wasn't even working for us

Accomplishments that we're proud of

Finally, We made an application by integrating vector DB into it
We made a tool that helps every geek or techies with research and study purpose

What we learned

Learned how to use VectorDB
Integrated Auth in our application
And a lot of small things which changed our POV

What's next for CodeX

We are thinking of making a graph of nodes that links all our searches so we can visualize and navigate easily in the future
Make it into a real search engine without using external API's

Ashuradhipathi/CodeX