client.reviews.reject()

import { createClient } from "@toloka-platform/sdk";

const client = createClient({
  baseUrl: "https://platform.toloka.ai",
  apiKey: process.env.TOLOKA_API_KEY!,
});

const review = await client.reviews.reject({
  reviewId: "trr.019a0000-0000-7000-8000-000000000001",
  criteria: [
    "example"
  ],
  comment: "example"
});
console.log(review);

{
  "id": "trr.019a0000-0000-7000-8000-000000000001",
  "runId": "plr.019a0000-0000-7000-8000-000000000001",
  "itemId": "plri.019a0000-0000-7000-8000-000000000001",
  "nodeId": "<string>",
  "nodeName": "<string>",
  "status": "needs_review",
  "submittedAt": "2026-01-01T00:00:00Z",
  "deadlineAt": "2026-01-01T00:00:00Z",
  "closedAt": null,
  "performer": {
    "type": "expert",
    "expertId": null
  },
  "taskContent": {
    "input": {},
    "output": {}
  },
  "qualityCriteria": [
    {
      "id": "<string>",
      "qualityElement": "<string>",
      "pass": "<string>",
      "fail": "<string>"
    }
  ],
  "verdicts": [
    {
      "id": "trv.019a0000-0000-7000-8000-000000000001",
      "decision": "accept",
      "forced": false,
      "criteria": [
        "<string>"
      ],
      "comment": "<string>",
      "source": "ui",
      "userId": "<string>",
      "scopedTokenId": null,
      "llmQaCheck": null,
      "createdAt": "2026-01-01T00:00:00Z"
    }
  ],
  "checkedRejectsLeft": 0
}

import { createClient } from "@toloka-platform/sdk";

const client = createClient({
  baseUrl: "https://platform.toloka.ai",
  apiKey: process.env.TOLOKA_API_KEY!,
});

const review = await client.reviews.reject({
  reviewId: "trr.019a0000-0000-7000-8000-000000000001",
  criteria: [
    "example"
  ],
  comment: "example"
});
console.log(review);

{
  "id": "trr.019a0000-0000-7000-8000-000000000001",
  "runId": "plr.019a0000-0000-7000-8000-000000000001",
  "itemId": "plri.019a0000-0000-7000-8000-000000000001",
  "nodeId": "<string>",
  "nodeName": "<string>",
  "status": "needs_review",
  "submittedAt": "2026-01-01T00:00:00Z",
  "deadlineAt": "2026-01-01T00:00:00Z",
  "closedAt": null,
  "performer": {
    "type": "expert",
    "expertId": null
  },
  "taskContent": {
    "input": {},
    "output": {}
  },
  "qualityCriteria": [
    {
      "id": "<string>",
      "qualityElement": "<string>",
      "pass": "<string>",
      "fail": "<string>"
    }
  ],
  "verdicts": [
    {
      "id": "trv.019a0000-0000-7000-8000-000000000001",
      "decision": "accept",
      "forced": false,
      "criteria": [
        "<string>"
      ],
      "comment": "<string>",
      "source": "ui",
      "userId": "<string>",
      "scopedTokenId": null,
      "llmQaCheck": null,
      "createdAt": "2026-01-01T00:00:00Z"
    }
  ],
  "checkedRejectsLeft": 0
}

Only teams with customer review enabled: other teams get 403 The "customerReview" feature is not enabled for this team. Reject the task under review, naming the quality criteria it fails and explaining what is wrong. The LLM QA check then checks this evidence: if it agrees, the reject is final, the item fails at this node, and nothing is billed for the task; otherwise the review returns with the check’s reason for a new decision, or expires if its deadline has passed. With force, the check is skipped and the reject is final at once; a forced reject is accepted only after the LLM QA check has returned a reject of this review, so reject without force first.

Signature

Method signature on client.reviews:

reject(args: {
  reviewId: string;
  criteria: string[];
  comment: string;
  force?: false | true;
}): Promise<Review>;

Parameters

string

required

Review ID with prefix trr..

string[]

required

IDs of the quality criteria the task fails: at least one, each named once.

string

required

What is wrong with the task, at most 10000 characters.

false | true

Skip the LLM QA check and reject at once. The reject is final and does not count toward the checked rejects. Allowed only after the LLM QA check has returned a reject of this review; otherwise the response is 409. Defaults to false.

Returns

Promise<Review> A customer review of one task, with its content, quality criteria and decisions.

ReviewId

required

Review ID with prefix trr..

RunId

required

Pipeline run ID with prefix plr..

RunItemId

required

Run item ID with prefix plri..

string

required

ID of the generation node whose task is under review.

string

required

Name of that generation node.

ReviewStatus

required

Review status. needs_review: waiting for the first decision. returned: the LLM QA check did not confirm the last reject, so the review waits for a new decision. checking: the LLM QA check is checking a reject, and decisions are refused until it answers. accepted: the team accepted the task. rejected: the reject is final, either forced or confirmed by the LLM QA check. expired: the deadline passed while the review was open, whether it waited for a decision or was checking a reject the check then did not confirm, so the task was accepted automatically.

string

required

Time the task was submitted, as an ISO 8601 timestamp.

string

required

Deadline for a decision, as an ISO 8601 timestamp. From then on the review takes no decisions, but it closes a little later: a review waiting for a decision expires shortly after the deadline, and its task is accepted automatically (the expert is paid and the task is billed); a review checking a reject closes when the LLM QA check answers.

string | null

required

Time the review closed, as an ISO 8601 timestamp, or null while it is open.

object

required

Who did the task.

Show Fields

"expert" | "synthetic"

required

expert for a human expert, synthetic for synthetic generation.

string | null

required

ID of the expert who did the task, or null for a synthetic task.

object

required

What the task showed and what was submitted. Project files appear as file proxy URLs.

Show Fields

object

required

The input values of the task, by field name.

Show Fields

unknown

object

required

The submitted answer, by field name.

Show Fields

unknown

ReviewQualityCriterion[]

required

The quality criteria a reject can name, as they were when the task reached review.

Show Fields

string

required

Criterion ID to name in a reject.

string

required

What the criterion checks.

string

required

What a task that passes this criterion looks like.

string

required

What a task that fails this criterion looks like.

ReviewVerdict[]

required

Decisions on this review, oldest first.

Show Fields

string

required

Verdict ID with prefix trv..

"accept" | "reject"

required

accept or reject.

boolean

required

True for a forced reject, which skipped the LLM QA check.

string[]

required

IDs of the quality criteria a reject named; empty for an accept.See the HTTP endpoint for the complete schema.

string

required

The comment of a reject; empty for an accept.

"ui" | "api"

required

ui for the platform UI, api for the API, the CLI or the SDK.

string

required

Platform user ID of the author.

string | null

required

ID of the scoped token the verdict was sent with, with prefix pst., or null.

object | null

required

Answer of the LLM QA check to a reject, or null for an accept, a forced reject, or a check still running.See the HTTP endpoint for the complete schema.

string

required

Time the verdict was sent, as an ISO 8601 timestamp.

number

required

How many more rejects the LLM QA check will check, out of 10. At 0, accept the task or send a forced reject.

Errors

HTTP errors throw PotatoApiError. The endpoint documents these error responses:

Status	Description
400	The evidence of a reject is invalid
403	Customer review is not enabled for the team of the run, or the scoped token lacks the needed `review` permission. Deciding needs `review:write`, which every preset except `read` and `labeling-read` includes.
404	Review not found
409	The review cannot take this decision now: the LLM QA check is still checking a reject (retry after it answers), the checked rejects are used up (accept or force the reject), a forced reject was sent before the LLM QA check returned any reject of this review (reject without force first), or the review is closed or past its deadline. Read the review to see which.
429	Rate limit exceeded

HTTP endpoint: POST /reviews/{reviewId}/reject.

import { createClient } from "@toloka-platform/sdk";

const client = createClient({
  baseUrl: "https://platform.toloka.ai",
  apiKey: process.env.TOLOKA_API_KEY!,
});

const review = await client.reviews.reject({
  reviewId: "trr.019a0000-0000-7000-8000-000000000001",
  criteria: [
    "example"
  ],
  comment: "example"
});
console.log(review);

{
  "id": "trr.019a0000-0000-7000-8000-000000000001",
  "runId": "plr.019a0000-0000-7000-8000-000000000001",
  "itemId": "plri.019a0000-0000-7000-8000-000000000001",
  "nodeId": "<string>",
  "nodeName": "<string>",
  "status": "needs_review",
  "submittedAt": "2026-01-01T00:00:00Z",
  "deadlineAt": "2026-01-01T00:00:00Z",
  "closedAt": null,
  "performer": {
    "type": "expert",
    "expertId": null
  },
  "taskContent": {
    "input": {},
    "output": {}
  },
  "qualityCriteria": [
    {
      "id": "<string>",
      "qualityElement": "<string>",
      "pass": "<string>",
      "fail": "<string>"
    }
  ],
  "verdicts": [
    {
      "id": "trv.019a0000-0000-7000-8000-000000000001",
      "decision": "accept",
      "forced": false,
      "criteria": [
        "<string>"
      ],
      "comment": "<string>",
      "source": "ui",
      "userId": "<string>",
      "scopedTokenId": null,
      "llmQaCheck": null,
      "createdAt": "2026-01-01T00:00:00Z"
    }
  ],
  "checkedRejectsLeft": 0
}

Reference

TypeScript SDK

CLI

HTTP API

Signature

Parameters

Returns

Errors