How arjundivecha/curate-my-world works
Project Type and Purpose
• This is a full-stack, AI-powered personalized event curation system (“Squirtle”).
• Frontend: React + TypeScript (Vite).
• Backend: Node.js + Express (curate-events-api/).
• Persistence: Supabase (Postgres + edge functions).
• Orchestration of 5 external data sources + Claude Sonnet 4 for conversation analysis and dynamic prompt generation.
File Structure
• Top-level directories:
• src/ – React app with pages, components, hooks, UI library (shadcn-ui).
• curate-events-api/ – Express server integrating multi-source clients, pipeline, managers, parsers.
• supabase/functions/ – Serverless functions for collecting, deduplicating, orchestrating event ingestion.
• supabase/migrations/ – Database schema migrations.
• public/ – Static assets for the frontend.
• Config & scripts:
• .env.example + curate-events-api/.env.example – env var templates.
• scripts/ – start/stop scripts, health monitors, cleanup, port management.
• supabase/config.toml – Supabase project config.
• External integrations indicated by src/clients/ under the backend: Perplexity, Apyflux, PredictHQ, Exa, SerpApi, Ticketmaster, etc.
README for Additional Insights
• Architectural pattern:
• Multi-source parallel ingestion → aggregation pipeline → AI categorization → deduplication → location filtering → response to frontend.
• “Venue learning” and “category management” components.
• Real-time UI: dashboard, calendar, filtering.
• Key technologies:
• Claude Sonnet 4 (Anthropic) for conversation analysis.
• Supabase edge functions for serverless event collectors.
• Express for REST API on port 8765.
• React/Vite dev server on port 8766.
Detail
• Supabase functions folder as an orchestration layer that may be invoked by cron or directly by the backend.
• Caching, logging, environment/config modules as side utilities.