Prompt evaluations

PromptEvaluationsClient

adaline.prompts.evaluations kicks off evaluation runs against a prompt, inspects their status, and cancels them. Per-row results are accessed through the nested .results sub-client.

For long-running evaluations, prefer adaline.initEvaluationResults() — it wraps .results.list in a self-refreshing cache.

Access

import { Adaline } from '@adaline/client';

const adaline = new Adaline();
const evaluations = adaline.prompts.evaluations; // PromptEvaluationsClient

The class is also exported directly:

import { PromptEvaluationsClient } from '@adaline/client';

Sub-client

Property	Client	Covers
adaline.prompts.evaluations.results	EvaluationResultsClient	Paginated per-row evaluation results

Types from @adaline/api:

import type {
  Evaluation,
  EvaluationStatus,
  CreateEvaluationRequest,
  ListEvaluationsResponse,
  SortOrder,
} from '@adaline/api';

---

list()

List evaluations for a prompt (paginated). Filter by status, evaluator, or dataset.

list(options: {
  promptId: string;
  status?: EvaluationStatus;
  evaluatorId?: string;
  datasetId?: string;
  sort?: SortOrder;
  createdAfter?: number;
  createdBefore?: number;
  limit?: number;
  cursor?: string;
}): Promise<ListEvaluationsResponse>

Parameters

Name	Type	Required	Description
promptId	string	Yes	Prompt whose evaluations should be listed.
status	EvaluationStatus	No	Filter by lifecycle state: "queued", "running", "completed", "cancelled", "failed".
evaluatorId	string	No	Only return evaluations that included this evaluator.
datasetId	string	No	Only return evaluations that ran against this dataset.
sort	SortOrder	No	"createdAt:asc" or "createdAt:desc".
createdAfter / createdBefore	number	No	Unix millisecond bounds.
limit	number	No	Page size (default 50, max 200).
cursor	string	No	Cursor from a previous response.

Returns

Promise<ListEvaluationsResponse> with { data: Evaluation[]; pagination: Pagination }.

Example

const { data } = await adaline.prompts.evaluations.list({
  promptId: 'prompt_abc123',
  status: 'running',
  limit: 20,
});

---

create()

Start a new evaluation run. Runs asynchronously on the server — poll results with evaluations.results.list(...) or wrap with adaline.initEvaluationResults().

create(options: {
  promptId: string;
  evaluation: CreateEvaluationRequest;
}): Promise<Evaluation>

Parameters

Name	Type	Required	Description
promptId	string	Yes	Prompt to evaluate.
evaluation	CreateEvaluationRequest	Yes	datasetId, evaluatorIds, optional deploymentId or deploymentEnvironmentId, optional title.

Returns

Promise<Evaluation> — the queued evaluation.

Example

const evaluation = await adaline.prompts.evaluations.create({
  promptId: 'prompt_abc123',
  evaluation: {
    datasetId: 'dataset_abc123',
    evaluatorIds: ['evaluator_abc123', 'evaluator_xyz789'],
    deploymentEnvironmentId: 'environment_abc123',
    title: 'Nightly regression — 2026-04-21',
  },
});

console.log(evaluation.id, evaluation.status);

---

get()

Fetch a single evaluation by ID (metadata + aggregate scores, not per-row results).

get(options: {
  promptId: string;
  evaluationId: string;
}): Promise<Evaluation>

Example

const evaluation = await adaline.prompts.evaluations.get({
  promptId: 'prompt_abc123',
  evaluationId: 'eval_abc123',
});

console.log(evaluation.status, evaluation.summary);

---

cancel()

Cancel an in-flight evaluation. In-progress rows keep running to completion, but no new rows will start.

cancel(options: {
  promptId: string;
  evaluationId: string;
}): Promise<Evaluation>

Example

const cancelled = await adaline.prompts.evaluations.cancel({
  promptId: 'prompt_abc123',
  evaluationId: 'eval_abc123',
});

console.log(cancelled.status); // "cancelled"

---

See Also

• PromptsClient — parent client
• EvaluationResultsClient — .results sub-client
• Adaline class — initEvaluationResults() polling helper
• PromptEvaluatorsClient
• API reference: List evaluations · Create · Get · Cancel