Ask Video
Ask questions over evidence, not raw pixel guesses.
Ask Video is the user-facing workflow that turns World-State and evidence into grounded answers. It is designed to cite artifacts, expose uncertainty, and block high-risk automation.
Ask Video contract
POST /api/product/videos/{video_id}/ask
{
"question": "When did the delivery vehicle enter the loading zone?",
"evidence_required": true,
"include_cost_context": true
}
{
"answer": "A delivery vehicle entered the loading zone at 00:01:14.",
"evidence_refs": ["ev_loading_zone_00114"],
"review_required": false,
"confidence": "medium",
"cost_context_ref": "cost_20260624_001"
}Product behavior
| Situation | Expected behavior | Why it matters |
|---|---|---|
| Evidence is strong | Answer with evidence_refs | User can inspect the answer |
| Evidence is weak | Return review_required | Avoids false certainty |
| Question is high-risk | Block or require review | Prevents unsafe autonomy |
| Question asks identity or intent | Refuse or route to policy review | Avoids sensitive unsupported inference |
| Repeated question | Reuse artifacts and cost context | Improves economics and debugging |
Ask Video evaluation checklist
A buyer should not evaluate Ask Video by asking only easy questions. They should ask a visible event question, a timestamp question, a negative question, an ambiguous question, and a high-risk question. The system should answer what is supported, cite evidence, and refuse or review what is not supported.
Product boundaries are part of the product, not footnotes. Ayneye is not presented as a replacement for every GPU video foundation model, not a free-infinite-query engine, and not an autonomous surveillance decision system. The current beta path is controlled signup, dashboard, REST API, bounded processing, evidence artifacts, visible limits, and read-only agent/MCP-style evaluation. Hard scenes, identity-sensitive workflows, emergency response, physical access, discipline, and destructive actions require human review or remain blocked.