TensorFlow Model Inference

naftiko: "1.0.0-alpha1"

info:
  label: "TensorFlow Model Inference"
  description: >-
    Workflow capability for running ML model inference using TensorFlow Serving.
    Combines model management and inference operations to support MLOps workflows
    including model health monitoring, metadata inspection, and running
    classification, regression, and prediction tasks in production environments.
  tags:
    - TensorFlow
    - Machine Learning
    - Model Serving
    - Inference
    - MLOps
    - AI
  created: "2026-05-03"
  modified: "2026-05-03"

binds:
  - namespace: env
    keys:
      TENSORFLOW_SERVING_HOST: TENSORFLOW_SERVING_HOST
      TENSORFLOW_SERVING_PORT: TENSORFLOW_SERVING_PORT

capability:
  consumes:
    - import: tensorflow-serving
      location: ./shared/tensorflow-serving.yaml

  exposes:
    - type: rest
      port: 8080
      namespace: tensorflow-inference-api
      description: "Unified REST API for TensorFlow model inference and management."
      resources:
        - path: /v1/models/{model_name}
          name: model-status
          description: "Model status and health information"
          operations:
            - method: GET
              name: get-model-status
              description: "Returns the current status of a model in the ModelServer"
              call: "tensorflow-serving.get-model-status"
              with:
                model_name: "rest.model_name"
              outputParameters:
                - type: object
                  mapping: "$."
        - path: /v1/models/{model_name}/metadata
          name: model-metadata
          description: "Model signature and schema metadata"
          operations:
            - method: GET
              name: get-model-metadata
              description: "Returns metadata and signature definitions for a model"
              call: "tensorflow-serving.get-model-metadata"
              with:
                model_name: "rest.model_name"
              outputParameters:
                - type: object
                  mapping: "$."
        - path: /v1/models/{model_name}/versions/{version}
          name: model-version-status
          description: "Status for a specific model version"
          operations:
            - method: GET
              name: get-model-version-status
              description: "Returns status for a specific version of the model"
              call: "tensorflow-serving.get-model-version-status"
              with:
                model_name: "rest.model_name"
                version: "rest.version"
              outputParameters:
                - type: object
                  mapping: "$."
        - path: /v1/models/{model_name}/versions/{version}/metadata
          name: model-version-metadata
          description: "Metadata for a specific model version"
          operations:
            - method: GET
              name: get-model-version-metadata
              description: "Returns metadata for a specific version of a model"
              call: "tensorflow-serving.get-model-version-metadata"
              with:
                model_name: "rest.model_name"
                version: "rest.version"
              outputParameters:
                - type: object
                  mapping: "$."
        - path: /v1/models/{model_name}/classify
          name: classify
          description: "Classification inference endpoint"
          operations:
            - method: POST
              name: classify-model
              description: "Run classification inference on the model"
              call: "tensorflow-serving.classify-model"
              with:
                model_name: "rest.model_name"
              outputParameters:
                - type: object
                  mapping: "$."
        - path: /v1/models/{model_name}/regress
          name: regress
          description: "Regression inference endpoint"
          operations:
            - method: POST
              name: regress-model
              description: "Run regression inference on the model"
              call: "tensorflow-serving.regress-model"
              with:
                model_name: "rest.model_name"
              outputParameters:
                - type: object
                  mapping: "$."
        - path: /v1/models/{model_name}/predict
          name: predict
          description: "Prediction inference endpoint"
          operations:
            - method: POST
              name: predict-model
              description: "Run prediction inference on the model"
              call: "tensorflow-serving.predict-model"
              with:
                model_name: "rest.model_name"
              outputParameters:
                - type: object
                  mapping: "$."

    - type: mcp
      port: 9090
      namespace: tensorflow-inference-mcp
      transport: http
      description: "MCP server for AI-assisted TensorFlow model inference and management."
      tools:
        - name: get-model-status
          description: "Check the health and availability status of a TensorFlow model"
          hints:
            readOnly: true
            openWorld: false
          call: "tensorflow-serving.get-model-status"
          with:
            model_name: "tools.model_name"
          outputParameters:
            - type: object
              mapping: "$."
        - name: get-model-version-status
          description: "Check status of a specific version of a TensorFlow model"
          hints:
            readOnly: true
            openWorld: false
          call: "tensorflow-serving.get-model-version-status"
          with:
            model_name: "tools.model_name"
            version: "tools.version"
          outputParameters:
            - type: object
              mapping: "$."
        - name: get-model-metadata
          description: "Retrieve signature definitions and schema metadata for a TensorFlow model"
          hints:
            readOnly: true
            openWorld: false
          call: "tensorflow-serving.get-model-metadata"
          with:
            model_name: "tools.model_name"
          outputParameters:
            - type: object
              mapping: "$."
        - name: get-model-version-metadata
          description: "Retrieve metadata for a specific version of a TensorFlow model"
          hints:
            readOnly: true
            openWorld: false
          call: "tensorflow-serving.get-model-version-metadata"
          with:
            model_name: "tools.model_name"
            version: "tools.version"
          outputParameters:
            - type: object
              mapping: "$."
        - name: classify-with-model
          description: "Run classification inference on a TensorFlow model with input examples"
          hints:
            readOnly: true
            openWorld: true
          call: "tensorflow-serving.classify-model"
          with:
            model_name: "tools.model_name"
          outputParameters:
            - type: object
              mapping: "$."
        - name: regress-with-model
          description: "Run regression inference on a TensorFlow model with input examples"
          hints:
            readOnly: true
            openWorld: true
          call: "tensorflow-serving.regress-model"
          with:
            model_name: "tools.model_name"
          outputParameters:
            - type: object
              mapping: "$."
        - name: predict-with-model
          description: "Run prediction inference on a TensorFlow model using row or column format inputs"
          hints:
            readOnly: true
            openWorld: true
          call: "tensorflow-serving.predict-model"
          with:
            model_name: "tools.model_name"
          outputParameters:
            - type: object
              mapping: "$."