Search Refine

Inspiration

I always have questions that I just cant quite put into words so that I can put it into google, and if so I waste time thinking of a question instead of finding an answer. Now the answer to that is to use AI, but its been shown time and time again that AI is not reliable in facts.

What it does

To fix this I built Search Refine which uses AI to optimize your prompt to get the best outcome from a search engine. This puts it in both worlds.

How we built it

I use a nextjs framework with an expressjs server. I use ChatGPT with a fine tuned response based on hundreds upon hundreds of optimization of the prompt to get the best one. I then use google search engine api to serach the web.

Challenges we ran into

Since I mostly work on linux software I am not that comfortable in web development, and because of this I had to spend the full time in the night and mornings too debuging and trying new things. This nextjs was tricky as it had diffferent features that changes since a few weeks ago and it change things that were constant in js development like .env variables.

Accomplishments that we're proud of

I was able to create a webapp for the first time of using the nextjs frameworks and beginning web development in general.

What we learned

I learned that everything is not as it seems, like somteimes http/s requests change in some modules in js and so it is important to think about that when coding it.

What's next for Search Refine

The next thing to do is to extend the range of the search to images as well as more responses. I would also begin writing a custom AI model to learn from the user on Search Refine. https://github.com/axelKeizoStahl/search-refine

Built With

• axios
• cors
• dotenv
• express.js
• google
• javascript
• nextjs
• openai
• react