backend-api-developmentBuilds lean FastAPI backends: routing and dependency injection, Pydantic models as the single validation/serialization layer, thin routes over feature-module services, auth (session cookies vs JWT, chosen by client type), one error shape with correct status codes, pagination on every list endpoint,
database-and-ormDesigns and operates the application data layer the lean way: schema design with real constraints, SQLAlchemy/SQLModel models, Alembic migrations as the only schema-change path, query patterns that avoid N+1 and load only what's needed, transactions committed inside the request, connection pooling s
deploy-and-operateShips and runs full-stack apps the lean way: small multi-stage Docker images, a CI pipeline shaped lint → test → build → migrate → deploy, twelve-factor environment and secrets discipline, health endpoints, structured logging with request IDs, and the minimal observability that answers "is it up and
frontend-modern-uiBuilds lean, dynamic frontends: React + Vite when the UI is a real application (components by feature, server state via TanStack Query vs local UI state, forms, accessibility basics), htmx + server templates when it's mostly forms and tables, and the judgment call between them. Use when building or
full-stack-app-architectureChooses and structures a full-stack application the lean way: picking the stack (default: FastAPI + React/Vite or htmx, SQLite-first), monolith-first project layout, module boundaries that follow features not layers, twelve-factor config/env handling, and the criteria for when (rarely, late) to spli
lean-code-principlesApplies the lean-code discipline that anchors this plugin: minimizing lines of code by leaning on frameworks and the standard library, YAGNI, small public surface area, deleting code as a feature, and judging when an abstraction pays for itself versus when it's speculative cost. Use when writing or
ml-in-productionPuts machine-learning models into applications the lean way: packaging a trained model as a versioned artifact, serving it behind a FastAPI endpoint with Pydantic-validated inputs, choosing batch vs realtime inference by the product's actual latency need, keeping the feature pipeline identical betwe
realtime-and-dynamic-featuresAdds the highly dynamic layer to full-stack apps the lean way: choosing polling vs Server-Sent Events vs WebSockets by actual need, streaming responses (including LLM token streams), live-updating dashboards, optimistic UI, and background jobs with progress reporting, using FastAPI primitives and mi
testing-strategyDesigns minimal effective test suites for full-stack apps — testing behavior at the API boundary over mocking internals, a pytest fixture spine whose per-test isolation is proved rather than assumed, the production engine for anything constraint- or dialect-dependent, a handful of Playwright end-to-
ui-and-ux-inspectionInspects a bespoke web application for usability, cognitive-load, accessibility, interaction, workflow, performance, and privacy defects — tracing critical user processes backward from successful end states, separating observed evidence from inference, and producing reproducible severity- and confid