generated: '2026-07-20'
method: derived
source: openapi/pruna-ai-openapi.yml
notes: >-
  Entity graph derived from the P-API OpenAPI schemas. The core lifecycle is:
  upload File(s) -> submit Prediction referencing file URLs -> poll status ->
  download output. Model is selected via the Model request header (schemas
  p-image, p-image-edit, p-video, flux-*, wan-*, qwen-*, z-image-* describe
  per-model input shapes).
entities:
  - name: Prediction
    id_field: id
    description: A single generation request and its result.
    key_fields: [id, model, input, get_url, status, generation_url]
    schemas: [PredictionRequest, PredictionResponse, PredictionStatusResponse]
  - name: File
    id_field: id
    description: An uploaded reference asset used as prediction input.
    key_fields: [id, name, content_type, size, etag, urls, created_at, expires_at]
    schemas: [FileUploadResponse]
  - name: Model
    id_field: name
    description: >-
      A named generation model selected via the Model header (~20 models across
      image, edit, upscale, try-on and video categories, plus LoRA trainers).
    examples: [p-image, p-image-edit, p-image-upscale, p-image-try-on, p-video, flux-dev, wan-t2v, qwen-image]
  - name: OutputFile
    description: A generated asset delivered from the delivery endpoint.
    key_fields: [generation_url]
    schemas: [OutputFile]
relationships:
  - from: Prediction
    to: Model
    type: belongs_to
    via: Model header / model field
  - from: Prediction
    to: File
    type: has_many
    via: input.images[] (uploaded file get URLs)
  - from: Prediction
    to: OutputFile
    type: has_one
    via: generation_url
status_values: [starting, processing, succeeded, failed, canceled]