Event Portal
Tracking 14 Toronto music venues. AI reads event posters, enriches every artist across six platforms.
Visit project →
Toronto's music scene is scattered across fourteen venue websites that all look different, update on different schedules, and present event information in completely different formats. Some list events in clean HTML tables. Others post flyer images with artist names buried in graphic design. Finding out who is playing where this weekend requires visiting fourteen separate sites, some of which use JavaScript rendering that breaks simple scrapers. And knowing the artist name is just the start. You want to know their Spotify following, their YouTube presence, their Beatport releases. That information lives across six different platforms with no connection between them.
AI-First Scraping
Claude Vision reads each venue page the way a human would. It decides which listings are real upcoming events versus archived shows, placeholder content, or promotional banners. When a venue redesigns their website, the scraper keeps working because it understands page content, not CSS selectors. For events posted as poster images, Claude Vision processes the flyer artwork and extracts artist names directly from the graphic design. No OCR library. The model sees the image and reads it.
Cross-Platform Enrichment
Every discovered artist gets enriched across six platforms. Spotify for follower counts and genre tags. YouTube for subscriber counts and video stats. Beatport for track and release history. SoundCloud for play counts. Resident Advisor for event history. Instagram and TikTok scrapers collect social profiles, posts, reels, and video URLs using Playwright with stealth fingerprinting. The result is a complete profile for every artist playing in Toronto, assembled automatically from scattered sources that have no native connection to each other.
Currently tracking 119 events from 14 Toronto venues with 121 enriched artist profiles. Supabase database across 10+ tables covering events, venues, artists, promoters, event managers, and junction relationships. Eight rotating browser fingerprints for stealth automation. Multi-provider AI settings: DeepSeek for cost-effective processing, Anthropic Claude for vision and reasoning, OpenRouter for flexibility across models.
Have something like this?
Let's talk