Agentic media SLA

Summary

This service-level agreement governs the WAVE Agent Video Layer (AVL) endpoints under /api/v1/agentic-media/*, the WAVE Console multi-source query surface, and the AI inference billing layer. Public preview commitments apply during the preview period; GA commitments take effect when AVL graduates from public preview.

Availability commitments

Performance targets

• Discovery latency:p95 < 200 ms global edge
• Vault token verification:p95 < 50 ms
• Delivery URL issuance:p95 < 500 ms
• Receipt write to argus:p99 < 1 s
• Inference router (local Ollama):p95 < 5 s for 4096-token generation
• Inference router (cloud fallback): bounded by upstream provider SLA (Anthropic / OpenAI / Google)

Receipt retention

Every delivery and inference call emits a cryptographic receipt persisted to the WAVE Argus signal lake. Receipt records are retained for 5 years from creation, in compliance with EU AI Act Article 26 evidence-chain requirements. WAVE customers can export receipts via the /api/v1/agentic-media/receipts endpoint or the audit-export UI in /admin/argus.

Incident response

• P0 (service down): 15-minute first response, hourly updates
• P1 (major degradation): 1-hour first response, 4-hour update cadence
• P2 (minor degradation): 4-hour first response
• P3 (informational): next business day

Incident updates publish to status.wave.online.

Service credits (GA only)

When AVL graduates from public preview to GA, customers on paid plans become eligible for service credits if monthly availability falls below the GA target:

Exclusions

SLA commitments do not apply during scheduled maintenance windows (announced minimum 7 days in advance), force-majeure events, or customer-controlled misconfigurations (expired vault tokens, insufficient spend caps, unsigned webhook endpoints).

Contact

For SLA inquiries or service-credit requests, email support@wave.online with subject line "AVL SLA — <org_id>".