fetch — The (API) Key to Entertainment.

Not all things can define us. We are all audiences. We are all creators. But no matter who we are, we all need entertainment. That’s why we built fetch — a universal API key that turns any natural language query into structured JSON data and, in doing so, unlocks entertainment for everyone.

Who It Serves

Developers
Fetch removes one of the biggest barriers to creating: finding the right data. Maybe you have an idea for a movie recommendation app or a sneaker deal tracker but can’t find the right API. With fetch, you don’t need one — we give you a single key that scrapes, cleans, and structures data even from sources that don’t provide APIs, making it possible to build what you imagine.

Audiences
Fetch makes discovery richer and more nuanced than a basic Google search. If you’re bored on a Saturday night and want something underground, fetch goes beyond the surface results, digs deeper, and delivers exactly the type of entertainment you’re craving.

Creators
Fetch makes it easier to be discovered. If someone is searching for your genre, your work can rise to the top because fetch doesn’t just privilege the mainstream — it surfaces hidden gems, too.

How It Works

User enters a query and the structure of the data they want.
Query runs through the Google Search API to pull the top links.
Beautiful Soup scrapes and cleans each page into usable text.
Each page is processed separately, embedded using NumPy and scikit-learn, and ranked with vector similarity.
Distilled content is sent into Google Gemini, which fills in the structured JSON the user asked for.
Results are returned to the frontend — clean, usable, and ready.

Stack: Next.js (frontend), FastAPI + Python (backend).

Inspiration

Our inspiration came from frustration. Entertainment data is scattered, messy, and inconsistent. Some platforms offer APIs, many don’t, and even when they do, they often limit what you can actually access. We wanted something universal — a tool that would help developers build, audiences discover, and creators be found.

Along the way, we learned about web scraping, embeddings, and RAG pipelines. We saw the trade-offs between depth and speed, and we refined our system so that Gemini only receives the most relevant context. We discovered the importance of smarter ranking systems, caching common queries, and balancing universality with reliability.

Challenges

Our first RAG pipeline was too slow, feeding Gemini too much raw text.
Search results often returned generic links instead of specific ones.
Scaling required rethinking caching and databases.

Each challenge forced us to improve, and in building fetch we learned not just about technology but about building off each other’s work to create something greater.