Inspiration
We’ve all been stuck in that moment with friends or family where nobody knows what to do next. The options feel endless, yet nothing feels right. Whether you just moved, are traveling, or simply want something new, the stress of planning can kill the excitement of discovery. That frustration inspired us to build OdysseyAI: an intelligent agent that makes exploration seamless and fun.
What it does
OdysseyAI is a personalized discovery agent that curates activities and events based on your preferences, group type, budget, and even mood. It scrapes real-world, trending data from TikTok, Instagram, event sites, and Google Maps, then uses AI to recommend experiences that are personalized, context-aware, and fun. Instead of endless scrolling, you get a tailored guide to exploration.
How we built it
We started with a Google Form survey to capture user preferences (location, mood, budget, etc.). The data was processed with Python into structured user profiles. Using MiniMax, we generated dynamic hashtags and queries relevant to each profile. Apify scrapers pulled real-world activity data from TikTok, Instagram, Eventbrite, and Google Maps. We merged these results and fed them into LlamaIndex for searchability and personalization. Finally, OdysseyAI ties it all together into a conversational chatbot that recommends activities in real time.
Challenges we ran into
Normalizing survey responses across multiple categories and formats. Handling inconsistent outputs from the LLM and ensuring JSON was parseable. Working with multiple Apify scrapers, each expecting slightly different payloads (hashtags, queries, etc.). Balancing between scraping speed, data freshness, and API rate limits. Integrating multiple tools into one coherent pipeline under time pressure.
Accomplishments that we're proud of
Successfully integrated four different platforms (TikTok, Instagram, Maps, Eventbrite). Created a unique specialization approach: OdysseyAI as a “discovery agent” vs. general-purpose AI. Turned a common frustration (decision fatigue) into a fun, engaging solution. Collaborated as a team, each member owning a critical piece of the system (survey parsing, AI prompts, scraping, indexing).
What we learned
How to connect multiple AI and scraping tools into one workflow. The importance of prompt engineering for reliable, structured outputs. That users value context-aware recommendations far more than static lists. The challenges of cleaning and merging data from diverse sources. How collaboration and modular design (separating survey parsing, scraping, and indexing) accelerates progress.
What's next for OdysseyAI
Improve ranking logic so recommendations consider popularity, recency, and personal preferences. Build a lightweight web app with a chatbot interface for real-time exploration. Expand integrations to more data sources (Yelp, Ticketmaster, Meetup). Add continuous scraping and a backend database for live, always-updated results. Explore personalization at scale: collaborative filtering, session memory, and adaptive recommendations.
Log in or sign up for Devpost to join the conversation.