List QA metrics

List the QA metric judges, their default thresholds, the extraction filter operators and QA pricing.

GEThttps://api.tried.com.au/qa/metrics

Needs keys Returns the catalog you need to build a cohort's metrics and extraction_filters. Each metric scores a call from 0 to 100 and passes at its threshold.

Request#

This endpoint takes no path, query or body parameters.

Example request#

curl https://api.tried.com.au/qa/metrics \
  -H "Authorization: Bearer $TRIED_API_KEY"

const res = await fetch("https://api.tried.com.au/qa/metrics", {
  headers: { Authorization: `Bearer ${process.env.TRIED_API_KEY}` },
});
const catalog = await res.json();

import os, requests

res = requests.get(
    "https://api.tried.com.au/qa/metrics",
    headers={"Authorization": f"Bearer {os.environ['TRIED_API_KEY']}"},
)
catalog = res.json()

Response#

200 OK

{
  "metrics": [
    {
      "metric": "hallucination",
      "default_threshold": 70
    },
    {
      "metric": "kb_recall",
      "default_threshold": 60
    },
    {
      "metric": "wer",
      "default_threshold": 80
    },
    {
      "metric": "overlapping_speech",
      "default_threshold": 70
    },
    {
      "metric": "sentiment",
      "default_threshold": 50
    },
    {
      "metric": "naturalness",
      "default_threshold": 60
    },
    {
      "metric": "tool_call_accuracy",
      "default_threshold": 80
    },
    {
      "metric": "transition_accuracy",
      "default_threshold": 80
    },
    {
      "metric": "transfer_success",
      "default_threshold": 100
    },
    {
      "metric": "transfer_wait_time",
      "default_threshold": 50
    },
    {
      "metric": "latency",
      "default_threshold": 60
    }
  ],
  "extraction_filter_ops": [
    "eq",
    "neq",
    "contains",
    "not_contains",
    "gt",
    "gte",
    "lt",
    "lte",
    "exists",
    "not_exists",
    "in"
  ],
  "pricing": {
    "free_minutes_per_month": 100,
    "price_per_minute": 0.15
  }
}

Metrics that read the transcript (hallucination, kb_recall, naturalness, tool_call_accuracy, transition_accuracy) are judged by OpenAI when a key is set and by deterministic heuristics otherwise, or when the model call fails. wer, overlapping_speech, sentiment, transfer_success, transfer_wait_time and latency are always computed from call data. A metric that does not apply to a call (for example kb_recall with no knowledge base lookups) gets a null score and is left out of the weighted score.

Field	Type	Description
`metrics[].default_threshold`	number	Score needed to pass when the cohort does not set `threshold`.
`extraction_filter_ops`	string[]	Operators for `extraction_filters[].op`.
`pricing.free_minutes_per_month`	integer	Free scored minutes per calendar month (UTC).
`pricing.price_per_minute`	number	AUD per scored minute after the free minutes.

Errors#

Status	When
`401`	Missing or invalid API key.
`403`	The key's IP allowlist or scopes reject the request. This route needs the `analytics` scope.
`429`	Rate limit exceeded. See Rate limits.

See Errors for the body format.

Retell compatibility#

Retell's public API reference has no AI QA endpoints, so this route has no Retell SDK equivalent.