LoginStart beta
◉Ayneye
QuickstartStart beta

Quickstart

From beta signup to a verified artifact bundle.

Create a workspace, register a source, execute bounded processing, inspect artifacts, ask from evidence, and record pass/fail results.

Prerequisites

You need a beta account, dashboard session, one representative video, and a clear question that can be checked against the footage. Do not start with a polished demo clip. Start with ordinary footage from the target environment so the team can see confidence boundaries, missed events, and cost behavior.

Seven-step evaluation path

  1. 1

    Create beta account

    Use /beta. The signup creates a tenant-scoped sandbox and dashboard session.

  2. 2

    Open dashboard

    Confirm sandbox limits, onboarding checklist, and demo workspace CTA are visible.

  3. 3

    Register source

    Use /api/videos/add or dashboard source creation. Save video_id in your app.

  4. 4

    Execute processing

    Run a short CPU-first pass and wait for artifacts to materialize.

  5. 5

    Inspect artifacts

    Open scene_graph.json, timeline.json, evidence_bundle.json, detector_report.json, and cost.json.

  6. 6

    Ask from evidence

    Ask one narrow operational question and verify evidence_refs.

  7. 7

    Record outcome

    Log evidence coverage, errors, cost, latency, and review_required states.

Copyable starter commands

export AYNEYE_API_KEY="ak_beta_..."

curl -X POST https://api.ayneye.com/api/videos/add   -H "Authorization: Bearer $AYNEYE_API_KEY"   -H "Content-Type: application/json"   -H "Idempotency-Key: eval-source-001"   -d '{"url":"https://example.com/sample.mp4","title":"first evaluation clip","source_type":"url"}'

curl -X POST https://api.ayneye.com/api/product/videos/$VIDEO_ID/execute   -H "Authorization: Bearer $AYNEYE_API_KEY"

curl https://api.ayneye.com/api/product/videos/$VIDEO_ID/scene-graph   -H "Authorization: Bearer $AYNEYE_API_KEY"

curl -X POST https://api.ayneye.com/api/product/videos/$VIDEO_ID/ask   -H "Authorization: Bearer $AYNEYE_API_KEY"   -H "Content-Type: application/json"   -d '{"question":"Which object entered the loading zone after 00:10?","require_evidence":true}'

Pass/fail checklist

Pass only when the artifact bundle exists, evidence references resolve, cost information is visible, and the user interface can explain waiting states and limits. Fail the evaluation if the response is plausible but untraceable, if the UI hides limit exhaustion, or if the agent is allowed to take unsafe actions from video evidence alone.

After the first run, repeat the same question with slightly different wording and compare the answer, evidence_refs, and review_required flag. Then ask one unsupported question. This verifies that the API is not simply producing plausible text; it is preserving a traceable evidence contract. Store the test video, artifact version, cost report, and final decision in your internal evaluation notes before requesting more capacity.

Do not skip this record-keeping step. It helps compare future runs, diagnose schema changes, and explain why a workload passed or failed. A buyer evaluating video AI wants repeatability more than impressive language. The quickstart therefore ends with concrete artifacts, evidence, cost, limits, and a written decision, not with a vague demo success.

If the output is incomplete, the correct next step is not to hide the result. Open the error page, inspect the detector report, compare the evidence bundle, and decide whether the issue is a limit, a hard video, a missing artifact, or an unsupported question. That troubleshooting loop is part of the quickstart.