Shutter Expert — Test Prototype

1 Window photo

Upload a photo for preview. Photo analysis runs through the xAI Grok proxy (server-side key).

Drop a window photo here or click to choose

JPG/PNG · installer measure photos preferred

Load sample:

Existing shutters answered? (HARD GATE)

Existing shutters in the home?

If yes: Match decision

Product line

Wet location (actual water)?

Window surround

Mount

Frame type

Frame subtype (optional)

Width (in)

Height (in)

Available depth to glass/hardware (in)

Extension on back of frame (in)

Louver size (in)

Tilt

Window pane / divided-light count

Ordered panel count

Diagonal difference (in)

Divider rail?

Split tilt?

Obstruction within 3 in of frame?

Furniture in panel swing path?

Installer / order notes

Quick scenarios:

Clearly separated from the measurement rules engine. Vision model call is designed but not wired.

Provider: xAI Grok vision via OpenAI-compatible https://api.x.ai/v1/chat/completions with image_url input (JSON mode)
Default model: grok-4.3 (reasoning none; image downscaled) (override with env XAI_VISION_MODEL)
API key: env XAI_API_KEY on the box / Worker only — never in browser JS
Local proxy: run python3 server.py → POST /api/vision-analyze; if the key is unset, this step shows “photo AI not connected”
Later: same handler as a Cloudflare Worker / Netlify function — see VISION.md