Inspiration
Here, we're working on "RAD!" - a rag assistant for disability resources.
Finding resources can be really tough for people with disabilities or their caretakers. They're often left bouncing between different counselors and departments. Can we make resources finding easier?
What it does
Here, we scrape webpages and pdfs across the california department of rehab website and connect it up to a rag-bot.
We eventually hope to build increased capabilities relative to e.g. perplexity by building bespoke agents that annotate the resources from the website(s) from the simulated perspective of e.g. a social worker or a special disabilities teacher, or really anyone who helps people with disabilities. that way the search can be more precise and tailored to this sector.
How we built it
- Apify + scrapy to scrape
- Openai + instructor to annotate pdfs (classify if they're forms or informational)
- Llama-index + openai for RAG over the websites and their linked pdfs
- Chainlit for front end
Challenges we ran into
A lot of stuff... :)
Accomplishments that we're proud of
Getting an mvp off the ground!!!
What we learned
- Llama index is dope (but we knew that already)
- vecatara and render and diffbot n all are quite cool too! though we didn't end up getting them off the ground past demos :(
What's next for Rad!
- more complex RAG patterns (with databases differentiated appropriately)
- "simulated professionals" accessing those databases
- a webviewer that references webpages + pdfs as you talk to it
Log in or sign up for Devpost to join the conversation.