ai-recipes

name	ai-recipes
description	Use when implementing a complete AI feature in a BRICKS app end-to-end, after generator and model choice. Proven wiring patterns with events, actions, and Property Bank state - voice kiosk (VAD → STT → LLM → TTS, or the assistant generator with MCP tools and file search), document Q&A / product finder (file ingestion → vectorStore → reranker → LLM), realtime transcription display, and AI image/video signage with the GenerativeMedia brick. Each recipe includes the generators and bricks to add, state shape, event wiring, offline behavior, and how to verify in the simulator and on device. Triggers on "build a voice assistant kiosk", "make a FAQ bot from these files", "live captions", "AI-generated visuals", "wire it up". Do NOT use for picking generators or models (use ai-generators / ai-model-selection first).
license	MIT
metadata	{"version":"0.1.0","author":"BRICKS"}

AI Feature Recipes

End-to-end wiring patterns. Each recipe lives in references/ — read the one that matches, then adapt:

Recipe	File	Use for
Voice kiosk	references/voice-kiosk.md	Talk-to-it kiosks, voice ordering, information desks
Document Q&A / RAG	references/doc-qa.md	"Answer questions about our menu/manual/products"
Live captions	references/live-captions.md	Realtime transcription on screen
Generative signage	references/generative-signage.md	AI-generated ambient visuals

Announce at start: "I'm using the ai-recipes skill — following the recipe."

Rules that apply to every recipe

1. Prerequisites first. Generator and model must already be chosen (ai-generators, ai-model-selection). If not, do that first.
2. State lives in the Property Bank. Every recipe defines a small state machine (e.g. idle → listening → thinking → speaking). Drive ALL UI from those properties — never from ad-hoc event side effects — so the UI stays consistent and testable.
3. Offline behavior is part of the feature. Local pipelines: state what happens while models download/heat up (first launch needs network unless models are preloaded). Cloud pieces: define the degrade path (cached content, default media, hidden entry point) — a public screen must never show an error card.
4. Lifecycle on low-RAM devices. Use heat-up options when latency matters and RAM allows; use only-in-use live policies when several AI contexts share a constrained device.
5. Verification loop. After wiring: compile the project → walk the state machine in the simulator (drive properties manually to preview each UI state) → deploy to a real device for latency and audio checks — mic, speaker, and true model speed exist only there. The built-in bricks-cli skill covers deployment and on-device inspection.
6. Announce failure states. Wire the error events of every generator to a visible-but-graceful state (retry, fallback content) and log them — silent AI failures on a kiosk look like a frozen app.

When Not To Use

• Choosing generators or models — ai-generators / ai-model-selection.
• Importing the documents that feed RAG — the document-essentials plugin's pdf/docx/xlsx/pptx skills handle file → content; this skill consumes their output.
• The CTOR editor's own chat — app settings.