How timelessco/recollect works
Project type and purpose
• Recollect is a full-stack web application built with Next.js on the frontend and Supabase on the backend.
• Purpose: allow users to add and manage bookmarks, images, and documents; assign them to collections (public or private); tag and search items; generate automatic image descriptions via AI; and optionally self-host.
File structure
• Top-level:
• public/ (static assets: SVGS, fonts)
• src/pages/ (Next.js pages and API routes)
• src/components/, src/pageComponents/ (React UI components & page-specific layouts)
• src/async/ (custom React Query hooks and Supabase CRUD helpers)
• src/utils/ (Supabase clients, API helpers, constants)
• cypress/ (end-to-end tests)
• env/ (environment/schema management)
• Next.js serverless functions under src/pages/api/ implement REST endpoints for bookmarks, categories, files, sharing, tags, settings, and AI (v1/ai/*).
• Two Supabase clients: supabaseClient (browser) and supabaseServerClient (server/API).
• CI/CD and self-hosting scripts (.github/actions, vercel.json, release-it).
Key architectural insights from README
• Frontend: React + Next.js with TypeScript, Tailwind CSS, hydration, SSR/ISR where needed.
• Backend: Supabase Auth, Postgres, Storage.
• AI: asynchronous OCR & embeddings via Supabase Edge functions and third-party APIs.
• Testing: Cypress end-to-end covering UI and API.
• Deployment: Vercel (serverless) or self-hosted.