Skip to main content

Validation and Limits

To ensure security, reliability, and fairness, the Resultity API enforces strict validation and usage limits for all incoming inference requests.

Request Validation​

Each request passes through a validation pipeline:

  • Schema check — required fields (e.g. model, prompt, messages) and type validation;
  • Model availability — the requested model must be currently served by at least one node;
  • API key verification — checked against namespace, usage quota, and credit balance;
  • Limit enforcement — token, rate, and resource limits;
  • Secure modes (planned) — for some requests (e.g. in Space), cryptographic signatures may be required.

Invalid requests are rejected with detailed error messages and appropriate HTTP codes.

Rate Limits​

Each API key is subject to rate limiting to ensure fair resource sharing:

  • Requests per second (e.g. 30–120 RPS depending on plan and network load);
  • Burst limits — short spikes are allowed within tolerance windows;
  • Global and per-endpoint thresholds;
  • Namespace-level aggregation for enterprise setups.

Limits may be relaxed or adjusted for specific keys (e.g. testnet campaigns or high-volume plans).

Credit Limits​

Credits are required to execute jobs.
Before accepting a job:

  • A credit check is performed;
  • The corresponding amount is reserved (pre-charged);
  • After successful completion, credits are deducted;
  • If the job fails or times out, credits are refunded automatically.

These limits protect both the platform and its users by preventing abuse, ensuring predictability, and preserving availability.