generated: '2026-06-20'
method: derived
source: openapi/tensorflow-serving-openapi.yml
note: >-
  Derived from the TensorFlow Serving OpenAPI component schemas. The Serving API is inference-oriented
  rather than a CRUD resource graph, so entities relate by conceptual containment (a model has many
  versions; inference requests/responses reference a model) rather than by stored id-foreign-keys.
entities:
  - name: Model
    identified_by: model_name
    description: A served SavedModel, addressed by name in the ModelServer.
    relationships:
      - type: has_many
        target: ModelVersion
        via: versions
      - type: has_many
        target: ModelLabel
        via: labels
  - name: ModelVersion
    identified_by: version
    description: A specific numbered version of a model; carries a lifecycle state (LOADING/AVAILABLE/UNLOADING...).
    source_schema: ModelStatusResponse
    relationships:
      - type: belongs_to
        target: Model
        via: model_name
  - name: ModelLabel
    identified_by: label
    description: A named alias (e.g. stable, canary) pointing at a model version.
    relationships:
      - type: belongs_to
        target: Model
        via: model_name
  - name: ModelMetadata
    source_schema: ModelMetadataResponse
    description: Signature definitions and model_spec (name + version) for a served model.
    relationships:
      - type: belongs_to
        target: ModelVersion
        via: model_spec.version
  - name: InferenceRequest
    source_schemas: [ClassifyRequest, RegressRequest, PredictRequest]
    description: Classify/Regress/Predict input referencing a model signature.
    relationships:
      - type: belongs_to
        target: Model
        via: model_name
  - name: InferenceResponse
    source_schemas: [ClassifyResponse, RegressResponse, PredictResponse]
    description: Inference output (result / predictions / outputs).
render: null