paulh
/
recipe-manager
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
recipe-manager/PROJECT.md

4.5 KiB
Raw Blame History

Recipe Manager — Project Vision

Project Name: Recipe Manager
Owner: Paul Huliganga
Created: 2026-03-23
Status: Planning → MVP Development

Vision

Build a self-hosted, privacy-first recipe management app that Anne and Elizabeth can use to organize, import, and cook from their recipes—replacing CopyMeThat with full data ownership and extensibility.

Why Build This?

  1. Data sovereignty — We control our data, not a SaaS provider
  2. Agentic engineering practice — Perfect scope for autonomous agent-driven development
  3. Extensibility — Future AI features (meal planning, cost tracking via Fintrove integration)
  4. Learning — Hands-on experience with modern web stack + agent collaboration

Success Criteria

MVP (v0.1) — "Anne and Elizabeth can use it"

  • ✅ Manual recipe entry (title, ingredients, instructions, source URL, notes)
  • ✅ Recipe viewing/editing/deletion
  • ✅ Basic organization (tags/collections)
  • ✅ Search (text-based)
  • ✅ Cook mode (ingredient checkboxes, step tracking, wake lock)
  • ✅ Self-hosted deployment (Docker Compose on paje.ca)

v1.0 — "Feature parity with CopyMeThat"

  • Recipe scraping (browser extension or bookmarklet)
  • Multi-device sync
  • Print styles
  • Scaling (adjust serving sizes)
  • Random recipe suggestion

v2.0 — "AI-enhanced"

  • Ingredient substitution suggestions
  • Meal planning
  • Shopping list generation
  • Integration with Fintrove (recipe cost tracking)

Constraints

Technical

  • Stack: Node.js + TypeScript + React + SQLite
  • Deployment: Docker Compose, self-hosted on paje.ca
  • Storage: Local-first (SQLite), cloud sync optional in v1
  • Browser support: Modern evergreen browsers (Chrome, Firefox, Safari, Edge)

Development

  • Agent-driven: Sub-agents build, test, document autonomously
  • Version control: Git with conventional commits
  • Testing: Unit tests for core logic, E2E tests for critical flows
  • Documentation: Auto-generated architecture decision records (ADRs)

Timeline

  • MVP: 48-72 hours of agent work (overnight runs)
  • v1.0: 1-2 weeks (incremental agent iterations)
  • v2.0: TBD (depends on AI integration complexity)

Non-Goals (for MVP)

  • ❌ Mobile apps (web-first, PWA later)
  • ❌ User authentication (single household deployment)
  • ❌ Recipe sharing/community features
  • ❌ Video support
  • ❌ Nutrition tracking (maybe v2)

Key Decisions

Why SQLite?

  • Local-first, zero-config, portable
  • Perfect for single-household deployment
  • Can sync via file replication (Syncthing, Dropbox) if needed

Why React?

  • Agent familiarity (Codex knows React cold)
  • Rich ecosystem for forms, UI components
  • Can evolve to Next.js if SSR needed

Why Docker Compose?

  • Consistent deployment across machines
  • Easy backup/restore (just copy SQLite file)
  • Can add services incrementally (reverse proxy, backup cron, etc.)

Agent Instructions Reference

See AGENT_INSTRUCTIONS.md for how agents should approach development, testing, and documentation.

Decisions Made (2026-03-23)

  1. Domain/URL: recipes.paje.ca ✓
  2. Initial users: Whole family (Anne, Elizabeth, Paul, Joseph) ✓
  3. Data migration: Export from CopyMeThat if possible (free tier may be limited now)
  4. First priority after MVP: Recipe scraping extension (v1.0 milestone)

AI Features Vision (v2.0+)

These are the "smart" features planned for future releases:

Cooking Intelligence

  • Ingredient substitutions — "Out of buttermilk? Use milk + vinegar"
  • Meal planning — Generate weekly menus from your recipe collection
  • Shopping lists — Aggregate ingredients, de-duplicate, organize by grocery aisle
  • Recipe recommendations — "You haven't made this in 3 months, try it tonight!"

Data & Integration

  • Nutrition estimates — Calorie/macro breakdown via USDA database
  • Fintrove integration — Track recipe costs, generate budget-aware meal plans
  • Recipe costing — "This dinner costs $4.50/serving"

Advanced Input

  • Voice input — Add ingredients/steps via speech-to-text
  • OCR import — Scan physical recipe cards or cookbooks
  • Smart parsing — AI-enhanced recipe extraction from non-standard sites

Why later? Need MVP data (recipes + usage patterns) to tune AI features effectively. Scraping comes first so Anne & Elizabeth can import their existing collections.

This is a living document. Agents should update it as the project evolves.