Vision model
Multimodal image + context
Output
Strict JSON schema
Fallback
Demo analysis engine
Safety
Non-diagnostic by design
STRUCTURED OUTPUT SCHEMA
{
image_quality: { focus, lighting, resolution, usable, notes },
visible_findings:[{ label, descriptor, confidence }],
differential_compatibility:[
{ condition, compatibility: Higher|Moderate|Lower, rationale, icd_hint }
],
triage_level: routine|soon|urgent,
triage_action: string,
red_flags: [string],
supportive_care: [string],
patient_summary / professional_summary,
questions_for_doctor: [string],
contextual_genomic_pathways:[
{ gene, pathway, function, role }
]
}ENGINE STATUS
Scans run against the live vision model when it is reachable. If the model is unavailable, unconfigured or returns an unusable payload, the demo analysis engine immediately supplies a complete clinical case so the product never stalls during a presentation.
- Cases stored on this device: 1
- Produced by live RUBI: 0
- Produced by demo engine: 1
The model never returns a diagnosis. Differentials are expressed as visual compatibility only, and genomic pathways are contextual education, never inferred from an image.