What we commit to on uptime, response time, and maintenance windows — and what you get when we miss. This SLA applies to every paid tier of the Quark Signal API.
This SLA covers the production Signal API served from
api.quarkresearch.cc for paying
subscribers on the Regime Alert, Market Pulse, Research Reports,
and Signal Feed tiers. It does not cover:
quarkresearch.cc) or the terminal UI at localhost:8000.We target 99.5% monthly uptime on the API surface, measured
as the fraction of minutes in a calendar month during which the API returns a
valid 2xx response to a synthetic probe against
GET /api/health. A minute counts as "down" only
when every probe in that minute fails.
| Availability | Permitted downtime / month | Service-credit tier |
|---|---|---|
| ≥ 99.5% | ≤ 3h 36m | No credit owed |
| 99.0% – 99.49% | 3h 37m – 7h 12m | 10% of monthly fee |
| 98.0% – 98.99% | 7h 13m – 14h 24m | 25% of monthly fee |
| < 98.0% | > 14h 24m | 50% of monthly fee |
The public status page at status.quarkresearch.cc shows probe history and any declared incidents for the current month. Credits are capped at one month's subscription fee per billing period; they do not roll over between tiers or become cash refunds.
Measured from the Cloudflare edge, warmed cache, during business hours (09:00 – 16:00 ET on US trading days):
| Endpoint class | p50 target | p99 target | Notes |
|---|---|---|---|
/api/signals/current, /api/signals/usage | < 500ms | < 5s | Cache-hit; refreshed post-close daily |
/api/signals/history, /api/signals/history.csv | < 1.5s | < 10s | Size-dependent; days=730 can hit p99 ceiling |
/api/alpha/attribution | < 1s | < 5s | Cached snapshot; 24h refresh |
| Webhook delivery attempt | < 10s from signal change | < 60s | First delivery; retries follow documented backoff |
Response-time misses outside the uptime calculation do not trigger credits on their own. Chronic breaches (> 10% of requests exceeding p99 across a full billing period) should be reported to support and will be investigated as an incident.
Planned maintenance follows two rules:
Planned-maintenance minutes, properly noticed, are excluded from the uptime calculation.
| Severity | Definition | Target response |
|---|---|---|
| SEV-1 | API fully down or returning wrong data (stale > 24h, wrong regime, etc.) | Acknowledged within 30 minutes, mitigation in progress within 2 hours |
| SEV-2 | Single endpoint down; cache stale; degraded p99 | Acknowledged within 2 hours, fix within 1 business day |
| SEV-3 | Documentation gaps, schema-discrepancy bugs, minor rate-limit miscounts | Acknowledged within 1 business day, fix in next regular release |
Every SEV-1 or SEV-2 incident gets a public post-mortem on the status page within 5 business days of resolution, including: what happened, why, how we fixed it, and what we're changing so it doesn't happen again.
Credits are the sole and exclusive remedy for any failure to meet this SLA. The credit schedule in §2 applies monthly and is capped at one month's subscription fee per billing period.
The uptime calculation excludes downtime caused by:
fallback: true and a
staleness_seconds field so you can decide whether to use it.SLA Credit Request — [YYYY-MM] — [account email].We may revise this SLA with 30 days' notice by email to the account contact on file and a prominent notice on the status page. Changes apply to billing periods starting after the notice period ends. Prior versions are archived in the API docs changelog at /api_docs.html#changelog.
Questions: hello@quarkresearch.cc.