SRT Clarity Search

Clarity is simple perplexity.ai clone. Use the code for whatever you like! :)

If you have any questions, feel free to reach out to me on Twitter.

How It Works

Given a query, Clarity fetches relevant, up-to-date information from the web and uses OpenAI's API to generate an answer.

The app works as follows:

1. Get query from user
2. Scrape Google for relevant webpages
3. Parse webpages for text
4. Build prompt using query + webpage text
5. Call OpenAI API to generate answer
6. Stream answer back to user

Requirements

Get OpenAI API key here.

Running Locally

1. Clone repo

git clone https://github.com/SolidRusT/srt-clarity-search.git

2. Install dependencies

npm i

3. Run app

npm run dev

Using Docker

docker build -t solidrust/srt-clarity-search -f Dockerfile .

docker run -it -p 3000:3000  solidrust/srt-clarity-search

Improvement Ideas

Here are some ideas for how to improve Clarity:

• Speed up answers by replacing link scraping with the Google Search API (scraping was used to circumvent cost + rate limits)
• Add "follow up" searches
• Improve the prompt
• Get sources working in non text-davinci-003 models
• Train your own model to use for answer synthesis

Credits

Shoutout to Perplexity AI for the inspiration. I highly recommend checking their product out.

This repo is meant to show people that you can build powerful apps like Perplexity even if you don't have a large, experienced team.

LLMs are amazing, and I hope Clarity inspires you to build something cool!