Hyperbrowser Extract Structured Data

arazzo: 1.0.1
info:
  title: Hyperbrowser Extract Structured Data
  summary: Start an extract job with a prompt and schema, poll status, then fetch data.
  description: >-
    Turns one or more pages into structured JSON using an LLM prompt and an
    optional output schema. The workflow starts the extract job over a list of
    URLs, polls the lightweight status endpoint until the job reaches a terminal
    state, and fetches the structured data object once it completes. 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: extractApi
  url: ../openapi/hyperbrowser-extract-api-openapi.yml
  type: openapi
workflows:
- workflowId: extract-structured-data
  summary: Extract structured JSON from pages using a prompt and optional schema.
  description: >-
    Submits an extract job, waits for completion by polling status, and returns
    the structured data once the job finishes.
  inputs:
    type: object
    required:
    - apiKey
    - urls
    - prompt
    properties:
      apiKey:
        type: string
        description: Hyperbrowser account API key sent in the x-api-key header.
      urls:
        type: array
        description: The list of URLs to extract structured data from.
        items:
          type: string
      prompt:
        type: string
        description: The natural-language extraction prompt.
      schema:
        type: object
        description: Optional JSON schema describing the structured output shape.
  steps:
  - stepId: startExtract
    description: >-
      Submit an extract job over the supplied URLs using the prompt and optional
      output schema, capturing the returned jobId.
    operationId: post-api-extract
    parameters:
    - name: x-api-key
      in: header
      value: $inputs.apiKey
    requestBody:
      contentType: application/json
      payload:
        urls: $inputs.urls
        prompt: $inputs.prompt
        schema: $inputs.schema
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      jobId: $response.body#/jobId
  - stepId: pollStatus
    description: >-
      Poll the extract job status. The status moves through pending and running
      before reaching completed, failed, or stopped; loop back while still in
      progress and branch out on a terminal state.
    operationId: get-api-extract-id-status
    parameters:
    - name: x-api-key
      in: header
      value: $inputs.apiKey
    - name: id
      in: path
      value: $steps.startExtract.outputs.jobId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      status: $response.body#/status
    onSuccess:
    - name: extractComplete
      type: goto
      stepId: getResult
      criteria:
      - context: $response.body
        condition: $.status == "completed"
        type: jsonpath
    - name: extractRunning
      type: goto
      stepId: pollStatus
      criteria:
      - context: $response.body
        condition: $.status == "pending" || $.status == "running"
        type: jsonpath
  - stepId: getResult
    description: >-
      Fetch the completed extract job to return the structured data object the
      LLM produced from the pages.
    operationId: get-api-extract-id
    parameters:
    - name: x-api-key
      in: header
      value: $inputs.apiKey
    - name: id
      in: path
      value: $steps.startExtract.outputs.jobId
    successCriteria:
    - condition: $statusCode == 200
    - context: $response.body
      condition: $.status == "completed"
      type: jsonpath
    outputs:
      status: $response.body#/status
      data: $response.body#/data
  outputs:
    jobId: $steps.startExtract.outputs.jobId
    data: $steps.getResult.outputs.data