ReportAiPerformance report.ai-performance@1

Generated at from openvibe-contracts v0.443.0 and openvibe-sdk v0.177.0.

Version
1.0.0
Owner
report
Visibility
internal
Status
active
Compatibility
backward
Schema
https://openvibe.network/contracts/report/ai-performance.v1.json

report.ai-performance@1 (ai-metrics.md): how AI calls performed over a range, for one group (a model, a provider, an operation, a system, a task). Percentiles come from log-scale histograms (within one bucket’s width). Public answers group by model, provider or operation only and show a group with at least 20 calls; system and task groups are for staff. Money is USD; tokens per second is the decode speed after the first token; cost per task sums the calls that share a task id.

Fields

FieldTypeRequiredDescriptionConstraints
groupenumyes
  • one of "model", "provider", "operation", "system", "task", "place", "harness"
keystringyes
  • maxLength 200
providerstring | null
  • maxLength 120
rangeobject
  • no other fields
range.fromstring
  • format date
range.tostring
  • format date
callsintegeryes
  • minimum 0
succeededinteger
  • minimum 0
errorsinteger
  • minimum 0
success_ratenumber | null
  • minimum 0
  • maximum 1
fallback_ratenumber | null
  • minimum 0
  • maximum 1
streamed_sharenumber | null
  • minimum 0
  • maximum 1
cache_hit_ratenumber | null
  • minimum 0
  • maximum 1
cache_read_sharenumber | nullCache-read tokens as a share of all input tokens.
  • minimum 0
  • maximum 1
ttft_msobject
  • no other fields
ttft_ms.p50number | null
  • minimum 0
ttft_ms.p90number | null
  • minimum 0
ttft_ms.p95number | null
  • minimum 0
ttft_ms.ninteger
  • minimum 0
latency_msobject
  • no other fields
latency_ms.p50number | null
  • minimum 0
latency_ms.p90number | null
  • minimum 0
latency_ms.p95number | null
  • minimum 0
latency_ms.ninteger
  • minimum 0
tpsobject
  • no other fields
tps.p50number | null
  • minimum 0
tps.p90number | null
  • minimum 0
tps.p95number | null
  • minimum 0
tps.ninteger
  • minimum 0
tokensobject
  • no other fields
tokens.inputinteger
  • minimum 0
tokens.outputinteger
  • minimum 0
tokens.reasoninginteger
  • minimum 0
tokens.cache_readinteger
  • minimum 0
tokens.cache_writeinteger
  • minimum 0
costobject
  • no other fields
cost.charged_usdnumber
  • minimum 0
cost.list_usdnumber
  • minimum 0
cost.cache_saved_usdnumber
  • minimum 0
cost.per_call_usdnumber | null
  • minimum 0
cost.per_million_output_usdnumber | null
  • minimum 0
tasksobject | null
top_errorsarray of object
  • maxItems 5
  • items: no other fields
top_errors[].error_classstringyes
  • maxLength 60
top_errors[].countintegeryes
  • minimum 0

Examples

From the contract's own test fixtures: valid ones validate, rejected ones must fail.

Valid: a-model-over-a-week
{
  "group": "model",
  "key": "deepseek-v4-flash",
  "provider": "deepseek",
  "range": {
    "from": "2026-11-05",
    "to": "2026-11-12"
  },
  "calls": 18420,
  "succeeded": 18311,
  "errors": 109,
  "success_rate": 0.9941,
  "fallback_rate": 0.012,
  "streamed_share": 0.71,
  "cache_hit_rate": 0.04,
  "cache_read_share": 0.63,
  "ttft_ms": {
    "p50": 540,
    "p90": 980,
    "p95": 1310,
    "n": 13080
  },
  "latency_ms": {
    "p50": 3900,
    "p90": 9100,
    "p95": 12800,
    "n": 18420
  },
  "tps": {
    "p50": 86,
    "p90": 121,
    "p95": 133,
    "n": 12990
  },
  "tokens": {
    "input": 210000000,
    "output": 9100000,
    "reasoning": 0,
    "cache_read": 132000000,
    "cache_write": 0
  },
  "cost": {
    "charged_usd": 61.2,
    "list_usd": 74.9,
    "cache_saved_usd": 19.8,
    "per_call_usd": 0.0033,
    "per_million_output_usd": 6.72
  },
  "tasks": {
    "count": 4120,
    "cost_per_task_usd": {
      "p50": 0.006,
      "p90": 0.031,
      "p95": 0.05,
      "n": 4120
    },
    "time_per_task_ms": {
      "p50": 14000,
      "p90": 61000,
      "p95": 98000,
      "n": 4120
    }
  },
  "top_errors": [
    {
      "error_class": "rate_limited",
      "count": 71
    },
    {
      "error_class": "timeout",
      "count": 30
    }
  ]
}
Rejected: a-group-nobody-has
{
  "group": "person",
  "key": "usr_x",
  "calls": 3
}

Validate

const contracts = require('openvibe-contracts');
contracts.validate('report.ai-performance@1', value);   // { valid, errors: [{ path, message }] }