Vendune
Product Lab — complete external app example
This package adds a Studio navigation module, a product-detail question panel, a storefront page, managed translated/JSONB guide records, namespaced HTTP APIs, MCP tools, selected AI context and an independent SQLite/event service.
Run PRODUCT_LAB=1 ./scripts/dev.sh from the repository root, then install
manifest.json in an owner/admin's synthetic shop through POST /api/apps.
The launcher generates private ignored credentials; none enter the browser.
Open Product Lab in the Studio sidebar or Product advice in the shop.
Manage guide records under Apps → Product Lab. The native host supplies context
and allowed actions; app.js owns the UI and handles product-context updates.
| File | Responsibility |
|---|---|
manifest.json |
Versioned data/action/surface/API/event/AI declarations |
server.py |
Separate app service, sample facts and durable tenant-keyed event inbox |
index.html, app.js |
Versioned standalone guest UI in four languages |
Dockerfile, compose.yaml |
Non-root, read-only and resource-limited container example |
The recommendation action retrieves fixed sample care facts, not merchant
knowledge or an LLM answer. Managed guides demonstrate separate shop records;
a real provider can implement its own retrieval/model behind the same action.
The sample fact table is deliberately shared fixture content; tenant-dependent
records must include tenant ownership. Event delivery requires an independent
PROCESS_ROLE=app-worker process and receiver deduplication.
Complete contract, setup, SDK, tests and limits. Container limits do not constitute microVM isolation or a hostile-code runner.