Building a Smarter Web Crawler for Nonprofits (and the World)

We have a vision for PixlieAI - an engine that takes whatever data you put into it, builds sold knowledge graphs out of that data, that you can then use with any AI agent you desire to analysis to your heart’s desire, without the risk of hallucinations. All run privately on your hardware.

It open ups many possibility for uses. And as a team, we weren’t sure where to start. Even building out a generic tool that can be used for anything, you need to start somewhere.

Creating Tech to Making the World a Better Place

Then Tech to the Rescue found us. It aligned perfectly with our vision of how we are building Pixlie. We want tech to make the world a better place, and what a great place to start but at a nonprofit that matches nonprofits with tech companies willing to do pro bono work to help them find solutions to their problem.

When we got access to their platform and immediately saw a pattern. A need to gather public data (that is available on websites or social media) and analyze this data to send relevant notifications.

This would be our first focus.

An Easier Way to Monitor Government Updates

Our first partnership is with Fundatia ADEPT. Their aim is to give the high nature value farmed landscapes of Transylvania and their communities an economic future and relevance in the 21st century without sacrificing their sustainability and productivity. These landscapes are an important model for future productive and sustainable farming in Europe.

To create an automated, AI-driven solution that can proactively monitor relevant government websites and sources, extracting and analyzing policy updates, legislative changes, and new consultations.

This project is a perfect first focus for PixlieAI. We are building an interface that will be able to:

• Crawl a website inputted by the user
• Search for keywords inputted by the user
• Create search results based on these keywords
• Utilizing AI, give users notifications when there is a new, relevant search result based on their input

All with the ability to run locally and keeping cost down, while utilizing a knowledge graph to avoid hallucinations.

See our early versions crawling the Ollama website

Choose a domain to crawl:

Set your keywords:

See the search results:

Pixlie is open source. And while we are still in the early phases, we will be finalizing this piece of the puzzle in the coming weeks. Try it out for your self and see what PixlieAI can do.

Or contribute to PixlieAI and be part of something different.