OpenAI Batch Job

arazzo: 1.0.1
info:
  title: OpenAI Batch Job
  summary: Upload a batch input file, create a batch, poll until complete, and read the output file id.
  description: >-
    Uploads a JSONL batch input file with the batch purpose, creates a batch
    against a target endpoint, polls the batch until it reaches a terminal
    status, and on completion returns the output file id holding the results.
    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: openaiApi
  url: ../openapi/openai-openapi-master.yml
  type: openapi
workflows:
- workflowId: batch-job
  summary: Upload batch input, run a batch, poll to completion, and return output file id.
  description: >-
    Uploads the batch input file, creates the batch, polls its status, and on
    completion returns the output and error file identifiers.
  inputs:
    type: object
    required:
    - apiKey
    - batchInputFile
    - endpoint
    properties:
      apiKey:
        type: string
        description: OpenAI API key used as a Bearer token.
      batchInputFile:
        type: string
        description: The JSONL batch input file contents to upload (binary/multipart field).
      endpoint:
        type: string
        description: The endpoint the batch runs against (e.g. /v1/chat/completions).
  steps:
  - stepId: uploadInput
    description: Upload the JSONL batch input file with the batch purpose.
    operationId: createFile
    parameters:
    - name: Authorization
      in: header
      value: "Bearer $inputs.apiKey"
    requestBody:
      contentType: multipart/form-data
      payload:
        file: $inputs.batchInputFile
        purpose: batch
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      inputFileId: $response.body#/id
  - stepId: createBatch
    description: Create a batch using the uploaded input file and target endpoint.
    operationId: createBatch
    parameters:
    - name: Authorization
      in: header
      value: "Bearer $inputs.apiKey"
    requestBody:
      contentType: application/json
      payload:
        input_file_id: $steps.uploadInput.outputs.inputFileId
        endpoint: $inputs.endpoint
        completion_window: 24h
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      batchId: $response.body#/id
      status: $response.body#/status
  - stepId: pollBatch
    description: >-
      Poll the batch until it reaches a terminal status, branching to the result
      step when completed and looping while it is still processing.
    operationId: retrieveBatch
    parameters:
    - name: Authorization
      in: header
      value: "Bearer $inputs.apiKey"
    - name: batch_id
      in: path
      value: $steps.createBatch.outputs.batchId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      status: $response.body#/status
    onSuccess:
    - name: batchCompleted
      type: goto
      stepId: getResult
      criteria:
      - context: $response.body
        condition: $.status == "completed"
        type: jsonpath
    - name: batchProcessing
      type: goto
      stepId: pollBatch
      criteria:
      - context: $response.body
        condition: $.status == "validating" || $.status == "in_progress" || $.status == "finalizing"
        type: jsonpath
  - stepId: getResult
    description: Retrieve the completed batch and return its output and error file ids.
    operationId: retrieveBatch
    parameters:
    - name: Authorization
      in: header
      value: "Bearer $inputs.apiKey"
    - name: batch_id
      in: path
      value: $steps.createBatch.outputs.batchId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      outputFileId: $response.body#/output_file_id
      errorFileId: $response.body#/error_file_id
  outputs:
    batchId: $steps.createBatch.outputs.batchId
    outputFileId: $steps.getResult.outputs.outputFileId