Fetch

Fetch

Inspiration

Some people don't have time to order food or groceries, and even if they do, it's hard to tell if you are paying a good price.

What it does

Our project, Fetch, lets anyone find the best prices on food with just one text.

Fetch utilizes Linq to allow anyone to send a food request. As a first-time user, when you send a request, it will ask for your info. Once a request goes through, it goes to a ChatGPT API to standardize the request. That request then goes to Browserbase so it can look for the products across different sites and return the cheapest option.

Fetch also has memory—every request you send gets sent to a MongoDB database and then gets sent to the ChatGPT API, so you can just text things like "get me the same as last" or ask "what were my previous orders?"

How we built it

Linq to receive and parse incoming SMS/text requests from users.
ChatGPT API to process natural language, standardize order requests, and maintain conversational context.
Browserbase to run headless web browsing across multiple sites to scrape and compare product prices in real-time.
MongoDB to store user order history and session state for long-term memory.

Challenges we ran into

Standardizing unstructured user text messages into clean, actionable search queries using the ChatGPT API.
Automating real-time web navigation across multiple different e-commerce and food delivery sites using Browserbase to reliably extract and compare prices.
Managing state and context history across SMS interactions by syncing MongoDB database records with prompt contexts for OpenAI.

Accomplishments that we're proud of

Built a fully functional end-to-end SMS bot that allows users to compare food prices seamlessly using a single text message.
Integrated automated browser execution to pull real-time pricing data directly from web storefronts.
Implemented persistent memory so users can easily query or repeat past orders conversationally.

What we learned

How to integrate headless browser automation tools like Browserbase into a backend pipeline for price comparison.
How to leverage LLM APIs to parse unstructured human text into structured data queries.
Strategies for storing and injecting user session history into chat contexts via MongoDB.

What's next for Fetch

Direct Ordering: Allowing users to complete the checkout/payment process directly through text message without leaving the SMS interface.
Broader Merchant Support: Expanding Browserbase automation scripts to cover more local grocery chains, delivery apps, and specialty markets.
Smart Recommendations: Using order history to suggest deals, price drops on frequently purchased items, or automated reordering schedules.