Skip to content
Qualia Docs

Checks_ListDatasetChecks

GET
/v1/accounts/{account}/datasets/{dataset}/checks
 
curl --request GET \
  --url https://junction-api.qualiastudios.dev/v1/accounts/example/datasets/example/checks \
  --header 'X-API-Key: <X-API-Key>'

Lists the catalog’s checks as one dataset sees them: every active check version, grouped by category in the order returned, each with whether it can run on this dataset now and its latest completed result here. A check that cannot run is listed with the reason, never left out. NOT_FOUND outside the caller’s workspace.

Authorizations

Section titled “Authorizations ”

Parameters

Section titled “ Parameters ”

Path Parameters

Section titled “Path Parameters ”
account
required
string

The account id.

dataset
required
string

The dataset id.

Query Parameters

Section titled “Query Parameters ”
pageSize
integer format: int32

At most this many; 0 means 100, the maximum is 1000.

pageToken
string

From a previous response’s next_page_token.

Responses

Section titled “ Responses ”

200

Section titled “200 ”

OK

Media type application/json

Response for Checks.ListDatasetChecks.

object
datasetChecks

One page of checks, by category then check id.

Array<object>

A check in the platform’s catalog, as one dataset sees it. A check is a reviewed, versioned unit of verification that yields a verdict and findings per episode; it is not a model call or a pipeline template. Its definition is published on deploy and never changes under a version.

Served by qds-api.

object
name

Accounts/{account}/datasets/{dataset}/checks/{check}; {check} is the check id, e.g. task.graph.

string
checkId

The stable, namespaced check id: <category>.<name>.

string
version

The active version.

integer format: int32
displayName

A short display name.

string
category

The category it groups under, e.g. task or quality.

string
kind

How it reads an episode.

string format: enum
Allowed values: KIND_UNSPECIFIED VLM HEURISTIC
description

One sentence: what it checks.

string
requiredInputs

What an episode needs for the check to read it, e.g. task.revision, episode.media.

Array<string>
runnable

Whether RunDatasetChecks can select it on this dataset now.

boolean
unavailableReason

Why not, when it is not runnable; empty otherwise.

string
lastResult

The latest completed result here; unset when it never ran here.

object
checkRun

Accounts/{account}/checks/vlm/runs/{run}: the run.

string
completeTime

When it ended.

string format: date-time
version

The check version it ran.

integer format: int32
selectedCount

Episodes the run selected (after its filter and sample), so a partial sample is not read as full coverage.

integer format: int32
skippedCount

Of those, episodes the check could not read (ineligible or unread).

integer format: int32
partial

Whether the run read a filtered or sampled subset of the dataset.

boolean
passCount

Findings it wrote, by verdict.

integer format: int32
failCount

Failing findings.

integer format: int32
unknownCount

Findings that could not tell.

integer format: int32
issueCount

Of its findings, those raised as issues (a confident fail).

integer format: int32
noteCount

Of its findings, those raised as notes (a less confident fail, or unknown).

integer format: int32
assessedCount

Episodes the check wholly assessed: every result for the episode a pass or a fail. Less than selected_count means some episodes were not wholly assessed. Zero means none was wholly assessed; classify as not assessed only when pass_count and fail_count are also both zero (a camera can fail while another is unassessed). Set for heuristic checks; unset for a VLM check or a run recorded before the count was kept.

integer format: int32
unassessedReasons

Why results were not assessed, most frequent first, as the check said (e.g. missing or invalid metric: jerk_mean). Empty when every result was assessed, for a VLM check, or for a run recorded before reasons were kept.

Array<object>

One reason a check could not assess results, and how many it covers.

object
reason

The check’s own words, at most 200 characters.

string
count

Results not assessed for this reason.

integer format: int32
runningCheckRun

Accounts/{account}/checks/{check}/runs/{run}: the newest run of this check still running on this dataset, whoever started it; empty when none is. A person’s RunDatasetChecks selecting it is refused until it ends.

string
nextPageToken

Empty on the last page.

string
Example 
{
  "datasetChecks": [
    {
      "kind": "KIND_UNSPECIFIED"
    }
  ]
}

default

Section titled “default ”

Default error response

Media type application/json
One of:
API error

A refusal from the API itself.

object
code
required

The gRPC code name in upper case, e.g. NOT_FOUND.

string
message
required
string
retryable
required

Whether the same call may succeed if retried.

boolean
details
required

Empty, except when a dataset copy or move is refused: then one google.rpc.ErrorInfo with the refusal’s reason code, its domain and, in metadata, the datasets it names.

Array<object>
object
Example generated
{
  "code": "example",
  "message": "example",
  "retryable": true,
  "details": [
    {}
  ]
}