Pick one workflow
Do not test every video problem at once. Select one representative question and one safe source.
Use cases
Choose a workflow, not a generic computer-vision promise.
Ayneye use cases are organized around operational workflows: video agents, live streams, security review, media operations, and developer platforms. Each path explains the artifact bundle, safety boundary, evaluation checklist, and signup flow.
Use-case operating model
Every use case follows the same beta sequence: define one safe workflow, process one representative video or bounded stream window, inspect artifacts, ask evidence-grounded questions, review limits and cost, then decide whether to expand capacity.
Use case
Video agents
Read-only agent memory over scene state and evidence, without camera authority.
Open →Use caseLive streams
Bounded windows from continuous video with timeline and cost reports.
Open →Use caseSecurity review
Evidence-first review workflows with human gates for high-risk outcomes.
Open →Use caseMedia operations
Long-footage indexing, segment search, timestamp extraction, and reuse.
Open →Use caseDeveloper platforms
Embed video intelligence into SaaS products with REST and artifacts.
Open →Evaluation matrix
| Use case | Primary artifact | Buyer question | Safety boundary |
|---|---|---|---|
| Video agents | scene_graph.json + evidence_bundle.json | Can agents answer from reusable video memory? | Read-only tools before action |
| Live streams | timeline.json + cost.json | Can we process bounded windows instead of endless streams? | No unbounded capture jobs |
| Security review | evidence_bundle.json + detector_report.json | Can reviewers find events with evidence? | No identity or access-control automation |
| Media operations | timeline.json + scene_graph.json | Can editors locate segments faster? | Human editorial confirmation |
| Developer platforms | API + artifacts + billing state | Can this embed into a SaaS product? | Tenant, limits, and review gates |
How to start
Generate artifacts
Run bounded processing and inspect scene_graph.json, timeline.json, evidence_bundle.json, and cost.json.
Review evidence
Compare answers with cited time spans and decide whether the workload needs human escalation.