API Terms of Service
Developer terms, rate limits, and SLA for API access.
Updated April 2026
1. Scope
These API Terms supplement the Terms of Service and govern use of DMS Labs AI's OpenAI-compatible inference endpoints and developer APIs.
2. API Keys
- Keys are personal to your account and must be stored securely (server-side, secrets manager).
- You must rotate keys promptly upon suspected compromise.
- Leaked keys used by third parties remain your responsibility until revoked.
3. Rate Limits and Quotas
- Limits are applied per key and per account based on your tier (Lite / Pro / Max / Enterprise).
- Enterprise limits are set in the order form.
- Burst traffic exceeding quota returns HTTP 429; retry with exponential backoff and respect Retry-After headers.
4. Compatibility and Stability
The API follows the OpenAI Chat Completions and Embeddings request/response shapes for supported endpoints. We may add optional fields without notice and will announce breaking changes at least 60 days in advance via release notes.
5. Model Outputs
Completions are provided as-is. You are responsible for safety filtering, attribution where required, and compliance with downstream platform policies (e.g., app stores).
6. Prohibited Technical Uses
- Denial-of-service, credential stuffing, or probing for vulnerabilities.
- Reselling raw API access without a written reseller agreement.
- Automated benchmark publication without prior approval.
- Training competing foundation models on our outputs.
7. SLA (Enterprise)
Enterprise tier: 99.99% monthly uptime on dedicated NVIDIA Blackwell-class capacity (primarily B200; RTX 5090, RTX 6000 Pro, or GB300 where the order form specifies). Service-credit remedies are the sole and exclusive remedy for availability breaches. Lite, Pro, and Max tiers are provided on a commercially reasonable basis.
8. Deprecation
Deprecated models remain available for at least 90 days after announcement, unless safety or legal constraints require shorter notice.
9. Support
Community support via documentation. Pro and above include email support; Enterprise includes named contacts and response-time targets per the order form.
