This page sets out the service commitments EcoHash Technology LLC offers for its inference platform. Commitments differ by product: dedicated capacity carries availability and throughput guarantees, while shared serverless capacity does not. Support response targets apply to all paying customers.
This SLA supplements our Terms of Service and is incorporated into any Master Service Agreement (MSA) we sign. Where an executed MSA or order form states different terms, that agreement controls.
Serverless Inference API
Our multi-tenant serverless inference API is offered on a best-effort basis and carries no Service Level Agreement for latency or availability. Capacity is shared across all tenants, so throughput and response times vary with aggregate demand.
This is a deliberate trade-off rather than an omission. Serverless pricing reflects shared, elastic capacity with no reserved hardware. If your workload requires guaranteed availability or predictable latency, use a Dedicated Inference Instance Endpoint, described below.
Dedicated Inference Instance Endpoints
A Dedicated Inference Instance Endpoint reserves GPU capacity for your account alone. It carries the following commitments.
99% monthly reliability
At least 99% of eligible requests will complete successfully in each calendar month, measured per endpoint:
Monthly Reliability = (Eligible Requests − Failed Requests) ÷ Eligible Requests × 100
A failed requestis one that returns a 5xx response originating from the EcoHash platform, or that fails to return a response within the endpoint’s configured timeout. An eligible request is any request to the dedicated endpoint other than those excluded below.
The following are excluded from both counts:
- Customer-side errors, including 4xx responses, malformed or invalid payloads, unsupported parameters, and authentication failures.
- Requests rejected for exceeding your committed Tokens Per Minute (TPM) or other configured rate limits.
- Scheduled maintenance announced at least 48 hours in advance, and emergency maintenance required to preserve platform security or integrity.
- Failures caused by factors outside our reasonable control, including network faults beyond our edge, upstream model-provider outages for proxied third-party models, and force majeure events.
- Periods during which the endpoint is suspended for non-payment or for breach of the Terms of Service.
Throughput guarantee
Dedicated capacity is reserved for your committed Tokens Per Minute (TPM). That capacity is not shared with other tenants and remains available to you for the duration of the commitment, subject to the exclusions above. Sustained demand above your committed TPM may be rate-limited; committed throughput can be increased at any time by adding capacity to the endpoint.
Support response targets
Response times below are targets for our first substantive response, not for resolution. Priority is assigned by EcoHash based on the impact described when the case is raised; we may reclassify a case and will tell you when we do.
| Priority | Response time | Description | Examples |
|---|---|---|---|
| Urgent (P0) | Within 1 hour | Reserved for critical cases that break live production workflows. |
|
| High (P1) | Within 4 business hours | Problems that prevent regular platform usage but do not break live production. |
|
| Normal (P2) | Within 8 business hours | Requests for information, enhancements, or documentation clarification with no negative service impact. |
|
| Low (P3) | Within 2 business days | Any issues that do not fall into the P0, P1, or P2 categories. |
|
Business hours are 09:00–18:00 U.S. Pacific Time (UTC−8, or UTC−7 during daylight saving), Monday to Friday, excluding U.S. federal holidays. P0 response targets apply 24 hours a day, every day. Hours accrue only during business hours, so a P1 case raised at 17:00 on a Friday is answered by 13:00 the following Monday.
Raising a case
Open a support case from the console, or email info@ecohash.com. To let us classify a case correctly, include the affected endpoint or model, request IDs and timestamps where available, and the impact on your workload. For a suspected reliability shortfall, submit the claim within 30 days of the end of the affected month with the request IDs and timestamps supporting it.
Enterprise customers with an executed MSA may have different response targets, dedicated contacts, or escalation paths; those terms take precedence over this page. To discuss enterprise support, contact sales@ecohash.com.
Changes to this SLA
We may update this SLA from time to time. Material reductions in the commitments above will be announced at least 30 days before they take effect, and will not apply to the remainder of a commitment term already purchased. The “last updated” date at the top of this page reflects the most recent change.