Inspiration
We often found ourselves missing out on deals because prices would drop and go back up before we noticed. Existing tools were either limited to certain websites or required too much manual setup. We wanted to create something that feels like having an AI assistant constantly scouting the web for us.
What it does
AutoScout lets users describe in plain English what they want to track. For example: “Monitor this Nike hoodie every 60 seconds and notify me when it’s less than $100.” The system automatically extracts the interval, condition, and URL, stores it, and continuously monitors the product page. When the condition is met, the user is notified instantly.
How we built it
We built the backend with FastAPI and APScheduler to manage recurring checks. User descriptions are processed with Gemini to extract structured fields like description, condition, interval, and URL. Monitors are saved in DynamoDB, and product pages are fetched using Requests or Playwright screenshots. Gemini also helps extract price/stock values even when buried in dynamic or visual content. Notifications are delivered via AWS SNS to email or SMS.
Challenges we ran into
We faced multiple challenges with circular imports and package structure when wiring up Lambda-style functions for local development. Parsing user instructions reliably with AI required handling inconsistent outputs, so we had to sanitize JSON responses from Gemini. Getting price data consistently from product pages was also tricky due to dynamic content, forcing us to combine both HTML scraping and screenshot-based extraction.
Accomplishments that we’re proud of
We got the entire end-to-end flow working: from natural language monitor creation to recurring checks and real-time notifications. We are proud of building a system flexible enough to monitor any product on the web, not just pre-integrated stores.
What we learned
We learned how to combine traditional scheduling and scraping tools with AI to make systems smarter and more user-friendly. We also got hands-on experience with AWS DynamoDB, SNS, and integrating Gemini into a production-style backend.
What’s next for AutoScout
We want to expand notifications to include WhatsApp and push alerts, add a simple front-end dashboard for users to manage their monitors, and improve the AI parsing to handle even more complex conditions. Long term, we envision AutoScout becoming a personal shopping scout that not only tracks prices but also suggests better alternatives and predicts future sales.
Log in or sign up for Devpost to join the conversation.