FOMO
A cron-driven pipeline that scrapes events from 9+ sources, scores them with a local LLM, and writes the best ones to Google Calendar.
Details
FOMO scrapes events from 9+ sources, scores each one with a local LLM, and writes the ones worth attending directly to Google Calendar. It runs on cron with no server.
How it works
- Scrape 9+ event sources in parallel (DoTheBay, Fandango, FunCheap, Lectures on Tap, etc.)
- Filter out past events
- Auto-pass movies on your watchlist and artists from your Spotify top 200
- Score everything else 0-10 with a local LLM based on your preferences
- Drop anything below the FOMO threshold (e.g., 9.2)
- Write survivors to Google Calendar with score, reason, emoji, and source links
Events are deduplicated by fingerprint hash so re-runs don’t create duplicates. Failed scrapers fall back to cached results so one broken source doesn’t kill the whole run.
Why a local 3B model
Because I’m cheap. Also because it works- the task is classification, not generation. Given an event and a user profile, output a number and a short reason. A 3B model on Apple Silicon handles that fine at ~19 tok/s with no API costs or rate limits.
The real work is in the prompt, not the model. It includes my full preference profile- specific cuisines, artists, genres, hard dislikes, and a blacklist. Score anchors tell the model that 9+ means “would drop everything” and fewer than 0.5% of events should qualify. Without that calibration the model scored everything 6-8 and the filter was useless. With it, the pipeline surfaced events I’d never have found on my own- like Repair Cafe SF and Lectures on Tap- while nailing obvious fits like Lord of the Rings screenings at Alamo Drafthouse. No fine-tuning required.
The source is on GitHub.
Gallery
