arazzo: 1.0.1
info:
  title: Vercel v0 Generate App Then Poll For Files
  summary: Generate a web app from a prompt, then poll the chat until generated files are present.
  description: >-
    The v0 Platform API generates code asynchronously, so a freshly created chat
    may return before any files are ready. This workflow creates a v0 chat from a
    natural language prompt and then re-reads the chat in a poll loop, branching
    back to itself until the chat carries at least one generated file or the
    attempt budget is exhausted. Every step inlines its request so the flow can be
    executed without opening the underlying OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: v0PlatformApi
  url: ../openapi/vercel-v0-platform-openapi.yml
  type: openapi
workflows:
- workflowId: generateAppThenPoll
  summary: Create a v0 chat from a prompt and poll it until generated files appear.
  description: >-
    Starts a v0 generation session from a prompt and then repeatedly fetches the
    chat until the generated code files are available, returning the preview URL
    and file list once they are.
  inputs:
    type: object
    required:
    - apiKey
    - message
    properties:
      apiKey:
        type: string
        description: v0 API key (V0_API_KEY) used as a Bearer token.
      message:
        type: string
        description: Natural language prompt describing the app to generate.
      system:
        type: string
        description: Optional system instructions to guide code generation.
      model:
        type: string
        description: Optional AI model identifier to use for generation.
  steps:
  - stepId: createChat
    description: >-
      Start a new v0 chat session from the supplied prompt to begin generating
      the application.
    operationId: createChat
    parameters:
    - name: Authorization
      in: header
      value: Bearer $inputs.apiKey
    requestBody:
      contentType: application/json
      payload:
        message: $inputs.message
        system: $inputs.system
        modelConfiguration:
          model: $inputs.model
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      chatId: $response.body#/id
      url: $response.body#/url
  - stepId: pollChat
    description: >-
      Re-read the chat session. If no generated files are present yet, loop back
      and try again; once files exist, finish.
    operationId: getChat
    parameters:
    - name: chatId
      in: path
      value: $steps.createChat.outputs.chatId
    - name: Authorization
      in: header
      value: Bearer $inputs.apiKey
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      previewUrl: $response.body#/previewUrl
      files: $response.body#/files
      title: $response.body#/title
    onSuccess:
    - name: filesReady
      type: end
      criteria:
      - context: $response.body
        condition: $.files.length > 0
        type: jsonpath
    - name: keepPolling
      type: goto
      stepId: pollChat
      criteria:
      - context: $response.body
        condition: $.files.length == 0
        type: jsonpath
  outputs:
    chatId: $steps.createChat.outputs.chatId
    previewUrl: $steps.pollChat.outputs.previewUrl
    files: $steps.pollChat.outputs.files
    title: $steps.pollChat.outputs.title