Serverless Metrics API

Query per-account Serverless performance metrics for pass.wafer.ai.

Get per-account Serverless SLA metrics for traffic served through pass.wafer.ai. The response includes summary latency, TTFT, tokens-per-second, token counts, error counts, and a time series over the requested lookback window.

Serverless metrics are scoped to the account that owns the Serverless API key. They do not expose fleet-wide backend identities, live backend gauges, or other customers’ traffic.

Run

curl -s "https://api.wafer.ai/v1/endpoints/metrics?endpoint=pass.wafer.ai&range_minutes=60&model=<MODEL_ID>" \
  -H "Authorization: Bearer <SERVERLESS_API_KEY>"

Parameters

• endpoint (required): pass.wafer.ai
• range_minutes (optional): 5, 15, 30, 60, 360, 1440, 10080, 43200
• model (optional): filter to one resolved model, such as GLM-5.1 (see GET /v1/models for live IDs)
• default range_minutes: 60

Response Shape

{
  "endpoint": "pass.wafer.ai",
  "range_minutes": 60,
  "queried_at": "2026-05-22T19:55:19+00:00",
  "summary": {
    "total_requests": 42,
    "rps": 0.01,
    "ttft_p50_ms": 410.0,
    "ttft_p90_ms": 1250.0,
    "ttft_p95_ms": 2100.0,
    "ttft_p99_ms": 3200.0,
    "tps_p10": 52.0,
    "tps_p50": 92.0,
    "tps_p90": 155.0,
    "tps_p99": 180.0,
    "latency_p50_ms": 2100.0,
    "latency_p90_ms": 6400.0,
    "latency_p99_ms": 12000.0,
    "cache_hit_pct": 72.5,
    "total_input_tokens": 480000,
    "total_output_tokens": 42000,
    "total_cache_read_tokens": 348000,
    "count_2xx": 41,
    "count_4xx": 1,
    "count_5xx": 0,
    "error_rate_pct": 2.4,
    "concurrent_requests": 2,
    "active_accounts": null,
    "engine_running_requests": null,
    "engine_queue_depth": null,
    "engine_kv_cache_hit_rate_pct": null,
    "engine_kv_cache_usage_pct": null,
    "engine_preemptions_in_range": null
  },
  "backends": [],
  "timeseries": [
    {
      "time": "2026-05-22T18:55:00Z",
      "requests": 12,
      "tps_p10": 50.0,
      "tps_p50": 92.0,
      "tps_p90": 150.0,
      "tps_p99": 175.0,
      "ttft_p50_ms": 410.0,
      "ttft_p90_ms": 1200.0,
      "ttft_p95_ms": 1850.0,
      "ttft_p99_ms": 3000.0,
      "latency_p50_ms": 2100.0,
      "latency_p90_ms": 6400.0,
      "latency_p99_ms": 12000.0,
      "input_tokens": 48000,
      "cache_read_tokens": 34800,
      "error_count": 1
    }
  ]
}

Key Fields

• rps: average requests per second across the full window
• ttft_p50_ms, ttft_p90_ms, ttft_p95_ms, ttft_p99_ms: streaming time-to-first-token percentiles (null on non-streaming windows)
• tps_p10, tps_p50, tps_p90, tps_p99: output tokens-per-second percentiles
• latency_p50_ms, latency_p90_ms, latency_p99_ms: end-to-end latency percentiles
• cache_hit_pct: cache-read prompt tokens divided by total input tokens
• count_2xx, count_4xx, count_5xx: request count by status class
• error_rate_pct: 4xx + 5xx divided by total requests
• concurrent_requests: recent activity count for the scoped account
• timeseries[*]: per-bucket requests, the full TPS / TTFT / latency percentile set, plus input_tokens, cache_read_tokens, and error_count. The two token sums let you render a “cached vs total input” overlay without a second query. Bin size is derived from range_minutes.

Serverless Scoping

• Only Serverless API keys can query endpoint=pass.wafer.ai.
• Results are scoped to the owning Serverless account for the bearer key.
• model filters the resolved model that served the request.
• backends is empty for Serverless because backend identities are fleet-level data.
• Engine gauge fields and active_accounts are null for Serverless because they are fleet-level metrics.

Timeseries Bin Sizes

The timeseries bin width is derived from range_minutes:

range_minutes	Bin size
5, 15	1 minute
30	2 minutes
60	5 minutes
360 (6h)	15 minutes
1440 (24h)	30 minutes
10080 (7d)	3 hours
43200 (30d)	12 hours

Errors

• 401: missing or invalid API key
• 403: the API key is not a Serverless key for pass.wafer.ai
• 422: invalid range_minutes, endpoint, or model
• 502: upstream metrics query failed