Architecture Specs
Feature Blueprints
Don't rely on stale boilerplate. Use these Markdown specifications to guide your AI agent in generating fresh, context-aware implementations.
.md
Stripe Billing
Hybrid mock/real subscription system. Toggle keys to go from dev to prod.
blueprints/billing-stripe.md
- Subscription table schema
- Checkout & Portal actions
- Webhook synchronization
.md
Google Authentication
Activation pattern for production-ready OAuth with Google and NextAuth.js.
blueprints/auth-google.md
- Environment configuration
- GCP Console setup
- Redirect URI mapping
.md
AI SDK Integration
Standardized pattern for AI chat/completion integration using Vercel AI SDK and OpenRouter.
blueprints/ai-sdk.md
- Streaming chat interfaces
- Model agnostic (Claude/GPT)
- Runtime key validation
.md
Async Jobs Queue
Reliable, zero-dependency asynchronous job queue using the existing Postgres database.
blueprints/async-jobs.md
- Offload long-running tasks
- Persist jobs across restarts
- Simple retry logic
How to use Blueprints
Blueprints are "Prompt Engineering as Code". Instead of pasting code snippets, tell your AI agent (Cursor, Copilot, etc) to read the spec file.
# Example Prompt
"Read blueprints/async-jobs.md and implement the job queue system. Use the existing db/schema.sql patterns."
"Read blueprints/async-jobs.md and implement the job queue system. Use the existing db/schema.sql patterns."