TarseekPre-production

Tarseek API · Pre-production

Rate limits

Tarseek enforces project-level requests per minute (RPM), estimated tokens per minute (TPM), and concurrent-request limits before a provider receives the request. Production values are explicitly configured from the actual provider/account usage tiers and measured workload; API keys can have tighter overrides.

  • HTTP 429 rate_limit_exceeded
  • Headers: Retry-After, x-ratelimit-limit-requests, x-ratelimit-remaining-requests, token equivalents
  • A provider-side 429 returns provider_rate_limit; Tarseek does not switch to a secondary deployment or platform credential
  • Every project BYOK credential has its own capacity bucket. Tarseek admits at most 90% of configured upstream RPM, TPM, and concurrency.
Implemented
Project exhaustion returns HTTP 429 rate_limit_exceeded with retry and limit headers. Upstream capacity exhaustion returns retryable HTTP 503 provider_rate_limitwithout failover. Configure each BYOK credential from its provider dashboard.