Core API

Rate Limits & Concurrency

Review per-key, tier (free, pro, enterprise), per-model, and gateway rate limits plus active job concurrency limits (CONCURRENT_JOBS_EXCEEDED) and response headers on E2X.

Multi-Layer Rate Limiting Architecture

E2X enforces sliding-window rate limits backed by Redis across four layers on job submission, plus endpoint-specific sliding windows on pricing estimation and presigned uploads:

LayerScope & LimitsError Code
Per-API-Key LimitOptional requestsPerMinute (1–1,000), requestsPerHour, and requestsPerDay configured per API key.429 RATE_LIMIT_EXCEEDED
User Tier LimitAccount-level minute, hourly, daily, and concurrent job limits mapped from free, pro, or enterprise tiers.429 RATE_LIMIT_EXCEEDED
Model Rate & ConcurrencyModel-version per-minute/hourly caps plus concurrent active jobs (pending + processing) defaulting to 5 unless overridden by modelVersion.rateLimits.concurrentJobs.429 CONCURRENT_JOBS_EXCEEDED
Global Gateway CapPlatform-wide threshold capped at 10,000 requests/minute across the gateway.429 RATE_LIMIT_EXCEEDED

Fail-Closed Degraded Mode

If Redis is temporarily unreachable, the rate limiter fails closed using an in-memory fallback window of 5 requests per 60 seconds per bucket rather than opening unlimited traffic.

Account Tiers & Concurrency

TierRequests / MinRequests / HourRequests / DayConcurrent Jobs
free101001,0002 (user tier) / up to model cap
pro1005,00050,00010
enterprise1,00050,000500,00050

Endpoint-Specific Sliding Windows

EndpointWindow & LimitAdditional Constraints
POST /v1/jobs/estimate60 requests / 300 seconds (5 minutes) per userEmits X-RateLimit-Limit, X-RateLimit-Remaining, and Retry-After on 429.
POST /v1/uploads/presign60 requests / 600 seconds (10 minutes) per userMaximum 200 concurrent pending uploads per user (429 TOO_MANY_PENDING_UPLOADS).
POST /v1/jobs/submitTier, API key, model & global limitsReturns retryAfter seconds inside error.details when rate limited.

Headers & Handling HTTP 429

Sliding-window middleware routes (/v1/jobs/estimate and /v1/uploads/presign) include X-RateLimit-Limit and X-RateLimit-Remaining on normal responses, plus Retry-After when returning 429. Job submission 429 responses include retryAfter inside error.details:

JSONHTTP 429 Rate Limit Response (/v1/jobs/submit)
{
  "success": false,
  "error": {
    "code": "RATE_LIMIT_EXCEEDED",
    "message": "Rate limit exceeded",
    "details": {
      "retryAfter": 42
    }
  },
  "requestId": "req_rl_402"
}