Hugging Face Dataset Filter Rows

arazzo: 1.0.1
info:
  title: Hugging Face Dataset Filter Rows
  summary: Confirm a dataset supports filtering, resolve its split, then apply a SQL-like filter.
  description: >-
    A targeted querying flow over the Dataset Viewer API. The workflow checks the
    dataset's validity to confirm the filter feature is available, branching to
    stop early when it is not, resolves the first subset and split, and then
    applies a SQL-like WHERE and ORDER BY filter to return only the matching rows
    in the requested order. Every step spells out its request inline so the flow
    can be read and executed without opening the underlying OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: datasetInfoApi
  url: ../openapi/hugging-face-dataset-info-api-openapi.yml
  type: openapi
- name: searchFilterApi
  url: ../openapi/hugging-face-search-filter-api-openapi.yml
  type: openapi
workflows:
- workflowId: dataset-filter-rows
  summary: Apply a SQL-like filter to a dataset split after confirming filter support.
  description: >-
    Confirms filtering is available for a dataset, resolves its first split, and
    runs a WHERE/ORDER BY filter against that split.
  inputs:
    type: object
    required:
    - hfToken
    - dataset
    - where
    properties:
      hfToken:
        type: string
        description: Hugging Face access token used as a Bearer credential.
      dataset:
        type: string
        description: The dataset id on the Hugging Face Hub.
      where:
        type: string
        description: SQL-like WHERE clause for filtering (e.g. "label = 1").
      orderby:
        type: string
        description: Optional SQL-like ORDER BY clause for sorting (e.g. "score DESC").
      length:
        type: integer
        description: Number of rows to return (max 100).
        default: 100
  steps:
  - stepId: checkFilterSupport
    description: >-
      Check the dataset's validity and confirm the filter feature is available
      before issuing a filter query.
    operationId: isValid
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.hfToken
    - name: dataset
      in: query
      value: $inputs.dataset
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      filter: $response.body#/filter
    onSuccess:
    - name: filterable
      type: goto
      stepId: resolveSplit
      criteria:
      - context: $response.body
        condition: $.filter == true
        type: jsonpath
  - stepId: resolveSplit
    description: >-
      Resolve the dataset's first subset and split to use as the filter target.
    operationId: getSplits
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.hfToken
    - name: dataset
      in: query
      value: $inputs.dataset
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      config: $response.body#/splits/0/config
      split: $response.body#/splits/0/split
  - stepId: filterSplit
    description: >-
      Apply the SQL-like WHERE and optional ORDER BY clauses to the resolved
      split and return the matching rows.
    operationId: filterRows
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.hfToken
    - name: dataset
      in: query
      value: $inputs.dataset
    - name: config
      in: query
      value: $steps.resolveSplit.outputs.config
    - name: split
      in: query
      value: $steps.resolveSplit.outputs.split
    - name: where
      in: query
      value: $inputs.where
    - name: orderby
      in: query
      value: $inputs.orderby
    - name: length
      in: query
      value: $inputs.length
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      rows: $response.body#/rows
      numRowsTotal: $response.body#/num_rows_total
  outputs:
    config: $steps.resolveSplit.outputs.config
    split: $steps.resolveSplit.outputs.split
    rows: $steps.filterSplit.outputs.rows
    numRowsTotal: $steps.filterSplit.outputs.numRowsTotal