SLA — Our Promise
Clear guarantees, fast response, transparent remedies — engineered for real operations.
Service Level Guarantees
Four pillars of reliability — availability, performance, support, and data security
Availability
Production endpoints target 99.5% monthly uptime (measured monthly; excludes scheduled maintenance and force majeure).
Performance
Median API response < 1.5s; 95th percentile < 5s for RAG queries with cached vectors.
Support & Response
Sev1: initial response ≤ 1 hour, resolution target ≤ 8 hours. Support portal & escalation included in paid plans.
Data & Privacy
TLS in transit, AES-256 at rest. Sandbox uploads auto-delete after 30 days. Production residency by contract.
Incident Flow & Remedies
Incident Flow
Detect
Automated alert
Triage
Assign owner
Mitigate
Apply fix
Resolve
Deploy solution
Post-mortem
Document & prevent
Customers receive incident status updates and a post-mortem within 72 hours for Sev1 events.
Service Credits
Uptime < 99.5%
Proportional service credit
Missed Sev1 target
10% one-time credit (capped)
Claim window
Submit within 30 days
Service credits (not cash refunds) proportional to SLA shortfall. See full SLA for details.
Acceptance Tests
RAG Ingest
50 PDFs ingested within 10 minutes
Chat Accuracy
40 benchmark Qs, ≥80% correct citation
Agent Test
3 scenarios, ≥90% correct actions
Production SLAs activate after acceptance tests pass ✓
Response & Resolution Times
| Severity | Description | Response | Resolution Target |
|---|---|---|---|
| Sev1 | System down / critical | ≤ 1 hour | ≤ 8 hours |
| Sev2 | Major feature degraded | ≤ 4 hours | ≤ 48 hours |
| Sev3 | Partial impairment | ≤ 24 hours | ≤ 5 days |
| Sev4 | Minor / feature request | ≤ 3 days | Best effort |
"We activate production SLAs after you sign off on acceptance tests. Tests are objective, repeatable, and included with every deployment."
For production SLAs, choose a service appendix: RAG / Agents / Predictive Ops
Got a Bottleneck? Let's Solve It.
Tell us what's slowing you down — manual ops, data silos, or scaling challenges. Our team will propose a feasible, custom AI solution built for your enterprise.
Reach out to usteam@zaffarx.comWe typically respond within a few hours — no bots, no forms, just real engineers.