Vegangenius Chef
AI-powered vegan recipe generator & personal cookbook
Live: free AI vegan recipe generator · Browse the published recipes
Generate recipes from natural-language prompts • AI food photography • Organize cookbooks • SSR public recipe pages
Overview
Vegangenius Chef is a full-stack vegan recipe app with three tiers:
Browser → Angular 22 SPA → Express reverse-proxy (:8080) → Flask API (:5000) → Cloud SQL
- Generate vegan recipes via Google Gemini (
gemini-3.8-flash in production)
- AI food photos via Google Gemini (
gemini-nano-banana-2.1 in production)
- Save & organize recipes into named cookbooks
- Publish recipes as SSR pages at
/r/<slug> for SEO & sharing
- Auth via Google OAuth or guest mode (localStorage)
Quick Start
Prerequisites: Node.js 26+, Python 3.11+, uv
# 1. Clone with submodules
git clone --recurse-submodules https://github.com/adamtasteslikegood/tasteslikegoodtheangularsvegancookbook.git
cd tasteslikegoodtheangularsvegancookbook
# 2. Frontend + Express proxy
npm install
cp .env.example .env.local # Set GEMINI_API_KEY, FLASK_BACKEND_URL
# 3. Flask backend
cd Backend
uv sync
cp .env.example .env # Set GOOGLE_API_KEY, GOOGLE_CLIENT_ID, etc.
./init_database.sh
cd ..
# 4. Run (two terminals)
cd Backend && uv run python app.py # Flask on :5000
npm run dev # Angular :3000 + Express :8080; dev routes are proxied
Commands
| Command | Description |
|---|
npm run dev | Angular :3000 + Express :8080 |
npm run build | Production build (Angular + Express) |
npm start | Production server (port 8080) |
npm run lint | ESLint check |
npm run format | Prettier format |
npm run type-check | TypeScript type check |
npm test | Vitest (server tests) |
Architecture
See docs/architecture/ for ADRs and diagrams.
├── src/ # Angular 22 SPA (signals, standalone components)
├── server/ # Express reverse proxy + static hosting
├── Backend/ # Flask API (submodule: tasteslikegood.com)
├── scripts/ # Utility scripts (PM tooling, git, gcloud)
├── specs/ # PM planning docs (synced to Confluence)
├── docs/ # All documentation
│ ├── architecture/ # ADRs, diagrams, rate limiting
│ ├── guides/ # Developer guides, quickstart
│ ├── deployment/ # Cloud Build, CI/CD checklists
│ ├── security/ # Security docs & guidelines
│ └── phases/ # Historical phase documentation
└── public/ # Static assets (privacy policy, etc.)
Deployment
Two Cloud Run services in us-central1:
express-frontend — Node.js (Angular + proxy)flask-backend — Python (API + AI + auth + DB)
See docs/deployment/DEPLOYMENT_CHECKLIST.md for the full checklist.
gcloud builds submit --config=cloudbuild.yaml
Atlassian PM workflow
This repo has an official cross-agent PM workflow outside git (all on tasteslikegood.atlassian.net — the only Atlassian site for work items):
- Jira KAN = active execution state, who is working on what now
- Jira RCP = delivery state, epics, sprints, acceptance scope
- Confluence TLG = durable planning/session context and documentation
specs/*.md = local working copies that sync non-destructively into Confluence
The PM tooling is hard-locked to that site by scripts/pm/_atlassian_guard.py:
Jira writes are limited to KAN/RCP, read-only rollups/briefings may also
include PLZG/TO, and any other site or project key is refused with an error.
Quick commands:
npm run pm:start # connectivity check + local PM briefing
npm run pm:brief # refresh local PM briefing
npm run pm:reflect # compare local git state to Jira/KAN linkage
npm run pm:sync # publish/update Confluence briefing
npm run pm:status # live Jira + PR + Confluence + prod snapshot
npm run pm:daemon # start daemon in background on this VM
npm run pm:daemon:status # show pid + log path
npm run pm:daemon:logs # tail daemon log
npm run pm:daemon:stop # stop background daemon
npm run pm:daemon:foreground # run watcher/MCP helper in foreground for debugging
Put real Atlassian credentials in .env using the variables shown in .env.example.
Contributing
See CONTRIBUTING.md for workflow, branching strategy, and code style.
License
MIT
