Get run details

Endpoint: GET /runs/{run_id}

Description
Returns a single run you can access: yours or shared with your organization. The payload includes every evaluation and a compact representation of each record. If the id is not visible to your API key (including when the run is not assigned to the key's project), the server responds with 404 (same as “not found”).

run_type and run_source are human-readable labels (for example "Dataset" and "API Call"), not internal enum names. user is the owning account's email, or null for org-only runs. author_email is the email of whoever created the run.

The dataset field (dataset name) may be null even when the run returns 200 — for example when the backing dataset exists but is not accessible to your API key (same behaviour as the dashboard “Dataset unavailable”). The run remains visible; only the nested dataset name is omitted.

Each record inside evaluations is truncated to 100 characters per string field (prompt, input, context, output, golden_answer). external_id is not truncated. metric is the metric shortname (for example ans_corr).

Parameters

• Path: run_id — integer run id.

Error responses

• 401 — Authentication failed.
• 404 — Run with id run_id not found or not accessible to your API key (including not assigned to the key's project).
• 500 — Error retrieving the run.

Responses

• 200 — JSON object with the shape below.

Example response (200)

{
  "id": 991,
  "user": "analyst@acme.com",
  "author_email": "analyst@acme.com",
  "run_type": "Dataset",
  "run_source": "API Call",
  "dataset": "Support QA - March",
  "data_collection": "Customer Support",
  "org_id": null,
  "number_of_metrics": 2,
  "result": 83.5,
  "threshold": 70,
  "model_slug": "gpt-4o",
  "alias": "support-march",
  "aggregate_results": {
    "ans_corr": 86,
    "faith": 81
  },
  "total_cost": "0.001200000000000",
  "started_at": "2026-04-01T09:12:00Z",
  "finished_at": "2026-04-01T09:13:12Z",
  "is_gte_threshold": true,
  "evaluations": [
    {
      "id": 1542,
      "run_id": 991,
      "record_index": 0,
      "metric": "ans_corr",
      "record": {
        "id": 1001,
        "external_id": "row-42",
        "prompt": "What is 2+2?",
        "input": null,
        "context": null,
        "output": "4",
        "golden_answer": "4"
      },
      "result": 100,
      "threshold": 70,
      "explanation": "Exact match.",
      "model_slug": "gpt-4o",
      "metric_args": null,
      "started_at": "2026-04-01T09:12:01Z",
      "finished_at": "2026-04-01T09:12:02Z",
      "evaluation_cost": "0.0004",
      "is_success": true,
      "is_gte_threshold": true,
      "eval_metadata": null
    }
  ]
}

{
  "id": 0,
  "user": "string | null",
  "author_email": "string",
  "run_type": "string",
  "run_source": "string",
  "dataset": "string | null",
  "data_collection": "string | null",
  "org_id": "integer | null",
  "number_of_metrics": 0,
  "result": "number | null",
  "threshold": 0,
  "model_slug": "string | null",
  "alias": "string | null",
  "aggregate_results": "object | null",
  "total_cost": "number | string | null",
  "started_at": "date",
  "finished_at": "date | null",
  "is_gte_threshold": "boolean | null",
  "evaluations": []
}

Each element of evaluations has this shape:

{
  "id": 0,
  "run_id": 0,
  "record_index": 0,
  "metric": "string",
  "record": {
    "id": 0,
    "external_id": "string | null",
    "prompt": "string | null",
    "input": "string | null",
    "context": "string | null",
    "output": "string | null",
    "golden_answer": "string | null"
  },
  "result": "number | null",
  "threshold": 0,
  "explanation": "string | null",
  "model_slug": "string | null",
  "metric_args": "object | null",
  "started_at": "date",
  "finished_at": "date | null",
  "evaluation_cost": "string | null",
  "is_success": "boolean | null",
  "is_gte_threshold": "boolean | null",
  "eval_metadata": "object | null"
}

metric_args is the per-metric argument map this evaluation actually ran with (dataset-level defaults plus any per-run override applied at run creation). eval_metadata is a metric-specific JSON object with extra signals or null.

curl

curl "https://api.aegisevals.ai/api/v1/runs/1" \
  -H "Authorization: Bearer sk_00000000000000000000000000000000"