SeeFix · Green Robin Labs
It sees the part.
It checks your work.
SeeFix is a camera-first repair copilot. Point your phone at a running toilet and it finds the worn flapper, shows you the replacement step on your own photo, and verifies the fix before you call it done.
Avery, first-time renter
The toilet keeps running at 2 AM. Avery doesn't know what a flapper is, and doesn't want to wait three days for the landlord.
Morgan, weekend fixer
Morgan can follow a video, but wants proof the part matches the model and that the job was done right, not just "looks fine".
How it works
1. /api/identify
Identify
Snap the model plate. Muse Spark Open (via Baseten) reads brand, model and serial into a typed record.
2. /api/segment
See the part
SAM 3.1 finds the exact part you name, like “toilet flapper”, and masks it on your camera frame.
3. /api/card
Show, don't tell
Muse Image turns your own photo into an annotated instruction card for the next step.
4. /api/safety
Safety gate
Muse Glimmer, Meta's open-weights model self-hosted on Modal, stops the job on gas, live mains or scald risk.
Built by Green Robin Labs for the Meta Global AI Developer Hackathon (Indie · Oct 26–Nov 4, 2026). Lab routes call real models (Spark Open via Baseten, SAM 3.1 + Muse Image via Meta, Glimmer on Modal). Missing keys or inactive Meta billing return a labeled 503 — nothing is mocked. Photo-based fix verification is live in /lab; the continuous camera verify loop is next.