openapi: 3.1.0
info:
  title: Google TensorFlow TensorFlow Serving REST API
  description: >-
    TensorFlow Serving exposes a REST API for running inference on trained
    TensorFlow models. It supports predict, classify, and regress endpoints
    with model versioning and lifecycle management.
  version: '1'
  contact:
    name: TensorFlow Community
    url: https://www.tensorflow.org/community
  license:
    name: Apache 2.0
    url: https://www.apache.org/licenses/LICENSE-2.0
externalDocs:
  description: TensorFlow Serving REST API Documentation
  url: https://www.tensorflow.org/tfx/serving/api_rest
servers:
  - url: http://localhost:8501
    description: Default TensorFlow Serving REST endpoint
tags:
  - name: Model Status
    description: Model metadata and status operations
  - name: Prediction
    description: Model inference operations
paths:
  /v1/models/{model_name}:
    get:
      operationId: getModelStatus
      summary: Google TensorFlow Get model status
      description: Returns the status of a model including available versions.
      tags:
        - Model Status
      parameters:
        - $ref: '#/components/parameters/model_name'
      responses:
        '200':
          description: Model status
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ModelStatusResponse'
        '404':
          description: Model not found
  /v1/models/{model_name}/metadata:
    get:
      operationId: getModelMetadata
      summary: Google TensorFlow Get model metadata
      description: >-
        Returns metadata for a model including signature definitions that
        describe the model's inputs and outputs.
      tags:
        - Model Status
      parameters:
        - $ref: '#/components/parameters/model_name'
      responses:
        '200':
          description: Model metadata
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ModelMetadataResponse'
  /v1/models/{model_name}:predict:
    post:
      operationId: predict
      summary: Google TensorFlow Predict
      description: >-
        Runs prediction (inference) on a model. Accepts input tensors and
        returns output tensors from the model's serving signature.
      tags:
        - Prediction
      parameters:
        - $ref: '#/components/parameters/model_name'
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/PredictRequest'
      responses:
        '200':
          description: Prediction results
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/PredictResponse'
        '400':
          description: Invalid input
        '404':
          description: Model not found
  /v1/models/{model_name}:classify:
    post:
      operationId: classify
      summary: Google TensorFlow Classify
      description: >-
        Runs classification on a model. Returns class labels and scores
        for the given input examples.
      tags:
        - Prediction
      parameters:
        - $ref: '#/components/parameters/model_name'
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/ClassifyRequest'
      responses:
        '200':
          description: Classification results
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/ClassifyResponse'
  /v1/models/{model_name}:regress:
    post:
      operationId: regress
      summary: Google TensorFlow Regress
      description: >-
        Runs regression on a model. Returns regression values for the
        given input examples.
      tags:
        - Prediction
      parameters:
        - $ref: '#/components/parameters/model_name'
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/RegressRequest'
      responses:
        '200':
          description: Regression results
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/RegressResponse'
components:
  parameters:
    model_name:
      name: model_name
      in: path
      required: true
      description: The name of the model
      schema:
        type: string
  schemas:
    PredictRequest:
      type: object
      properties:
        signature_name:
          type: string
          description: The serving signature to use (defaults to serving_default)
        instances:
          type: array
          description: Input instances for row-format requests
          items: {}
        inputs:
          description: Input tensors for columnar-format requests
          oneOf:
            - type: object
              additionalProperties: true
            - type: array
              items: {}
    PredictResponse:
      type: object
      properties:
        predictions:
          type: array
          description: Output predictions in row format
          items: {}
        outputs:
          description: Output tensors in columnar format
    ClassifyRequest:
      type: object
      properties:
        signature_name:
          type: string
        context:
          type: object
          additionalProperties: true
        examples:
          type: array
          items:
            type: object
            additionalProperties: true
    ClassifyResponse:
      type: object
      properties:
        results:
          type: array
          items:
            type: object
            properties:
              classes:
                type: array
                items:
                  type: object
                  properties:
                    label:
                      type: string
                    score:
                      type: number
    RegressRequest:
      type: object
      properties:
        signature_name:
          type: string
        context:
          type: object
          additionalProperties: true
        examples:
          type: array
          items:
            type: object
            additionalProperties: true
    RegressResponse:
      type: object
      properties:
        results:
          type: array
          items:
            type: object
            properties:
              value:
                type: number
    ModelStatusResponse:
      type: object
      properties:
        model_version_status:
          type: array
          items:
            type: object
            properties:
              version:
                type: string
              state:
                type: string
                enum: [START, LOADING, AVAILABLE, UNLOADING, END]
              status:
                type: object
                properties:
                  error_code:
                    type: string
                  error_message:
                    type: string
    ModelMetadataResponse:
      type: object
      properties:
        model_spec:
          type: object
          properties:
            name:
              type: string
            version:
              type: string
        metadata:
          type: object
          additionalProperties: true