Merge pull request #476 from dba/develop

Release 2.8.0

.whitesource

@@ -11,5 +11,11 @@
       "patch"
     ],
     "automerge": true
-  }]
-}
+  }],
+      "scanSettings": {
+        "baseBranches": ["develop"],
+        "configMode": "LOCAL",
+        "configExternalURL": "",
+        "projectToken": ""
+      }
+}

CHANGELOG.md

@@ -1,5 +1,9 @@
 # Open Prediction Service HUB
 
+## 2.8.0 _2022/11/28_
+
+* [DBACLD-47417] Extend OPS to support downloading of Ruleset Models
+
 ## 2.7.1 _2022/09/15_
 
 * [DBACLD-58320] Fix ads ml service rule import

open-prediction-service.yaml

@@ -1,7 +1,7 @@
 openapi: 3.0.3
 info:
   title: Open Prediction Service
-  version: 2.7.4-SNAPSHOT
+  version: 2.8.0-SNAPSHOT
   description: The Open Prediction Service API is an effort to provide an Open API that enables unsupported native ML Providers in Decision Designer or Decision Runtime.
 tags:
   - name: info
@@ -213,6 +213,47 @@ paths:
             application/json:
               schema:
                 $ref: '#/components/schemas/Error'
+  /models/{model_id}/download:
+    get:
+      tags:
+        - discover
+      summary: Download model binary
+      operationId: get_binary_by_id
+      parameters:
+        - $ref: '#/components/parameters/ModelIDParam'
+      responses:
+        '200':
+          description: Successful Response
+          content:
+            # A binary file:
+            application/octet-stream: {}
+        default:
+          description: Error
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/Error'
+  /models/{model_id}/metadata:
+    get:
+      tags:
+        - discover
+      summary: Get parsed model metadata
+      operationId: get_metadata_by_id
+      parameters:
+        - $ref: '#/components/parameters/ModelIDParam'
+      responses:
+        '200':
+          description: Successful Response
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/AdditionalModelInfo'
+        default:
+          description: Error
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/Error'
   /endpoints:
     get:
       tags:
@@ -823,13 +864,15 @@ components:
             unknown_file_size:
               type: boolean
               default: false
-              description: Whether the user can upload a file whose length is unknown at the time of the request.
+              description: Whether the user can upload a file whose length(Content-Length in HTTP header) is unknown at the time of the request.
       example:
         capabilities:
           - info
           - discover
           - manage
           - prediction
+          - download
+          - metadata
         managed_capabilities:
           supported_input_data_structure:
             - "auto"
@@ -854,6 +897,10 @@ components:
         - discover
         - manage
         - run
+        # Download model binary
+        - download
+        # Inspect model specific metadata, for example for PMML model return subType (Scorecard, Ruleset)
+        - metadata
     Parameter:
       title: Parameter
       description: Parameter for ml model invocation
@@ -907,6 +954,21 @@ components:
         target:
           - rel: endpoint
             href: 'http://open-prediction-service.org/endpoints/8c2af534-cdce-11ea-87d0-0242ac130003'
+    AdditionalModelInfo:
+      type: object
+      title: Additional Model Information
+      required:
+        - modelPackage
+        - modelType
+      properties:
+        modelPackage:
+          type: string
+          description: The file format of binary model
+          example: pmml
+        modelType:
+          type: string
+          description: Model type of binary model
+          example: RuleSetModel
   parameters:
     ModelIDParam:
       in: path

ops-client-sdk/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.ibm.decision</groupId>
     <artifactId>ops-client-sdk</artifactId>
-    <version>2.7.4-SNAPSHOT</version>
+    <version>2.8.0-SNAPSHOT</version>
     <inceptionYear>2020</inceptionYear>
     <build>
         <plugins>
@@ -224,7 +224,7 @@ limitations under the License.IBM Confidential]]>
         <maven.compiler.source>${java.version}</maven.compiler.source>
         <maven.compiler.target>${java.version}</maven.compiler.target>
         <gson-fire-version>1.8.5</gson-fire-version>
-        <swagger-core-version>1.6.8</swagger-core-version>
+        <swagger-core-version>1.6.9</swagger-core-version>
         <okhttp-version>4.10.0</okhttp-version>
         <gson-version>2.10</gson-version>
         <commons-lang3-version>3.12.0</commons-lang3-version>

ops-implementations/ads-ml-service/Dockerfile

@@ -1,4 +1,4 @@
-FROM --platform=${BUILDPLATFORM} python:3.8-slim AS python-base
+FROM --platform=${BUILDPLATFORM} python:3.11-slim AS python-base
 
 ENV \
     OPS_HOME="/usr/src/ads-ml-service" \
@@ -15,6 +15,7 @@ FROM python-base AS codegen-image
 
 COPY app/gen/. app/gen/
 RUN \
+    apt-get -q update && apt-get -q install --assume-yes build-essential && \
     python3 -m pip install -q --no-cache-dir --upgrade pip && \
     python3 -m pip install -q datamodel-code-generator~=0.11.19 && \
     datamodel-codegen \
@@ -30,8 +31,9 @@ COPY requirements.txt ./
 COPY logging.yaml /etc/ads-ml-service/logging/logging.yaml
 RUN \
     python3 -m venv /usr/src/ads-ml-service/venv && \
+    apt-get -q update && apt-get -q install --assume-yes build-essential libpq-dev python3-numpy && \
     python3 -m pip install -q --no-cache-dir --upgrade pip && \
-    python3 -m pip install -q --no-cache-dir --requirement requirements.txt && \
+    python3 -m pip install -q --no-cache-dir --prefer-binary --requirement requirements.txt && \
     mkdir -p /var/log/ads-ml-service && \
     mkdir -p /var/lib/ads-ml-service/models
 

ops-implementations/ads-ml-service/app/api/api_v2/endpoints/capabilities.py

@@ -35,7 +35,9 @@ async def server_capabilities() -> typing.Dict[typing.Text, typing.Any]:
             ops_model.Capability.info,
             ops_model.Capability.discover,
             ops_model.Capability.manage,
-            ops_model.Capability.run
+            ops_model.Capability.run,
+            ops_model.Capability.download,
+            ops_model.Capability.metadata
         ],
         'managed_capabilities': {
             'supported_input_data_structure': ['auto', 'DataFrame', 'ndarray', 'DMatrix', 'list'],

ops-implementations/ads-ml-service/app/api/api_v2/endpoints/models.py

@@ -17,6 +17,7 @@
 
 import logging
 import typing
+import io
 
 import fastapi
 import fastapi.encoders as encoders
@@ -28,10 +29,12 @@
 import app.crud as crud
 import app.gen.schemas.ops_schemas as ops_schemas
 import app.runtime.model_upload as app_model_upload
+import app.runtime.inspection as app_runtime_inspection
 import app.schemas as schemas
 import app.schemas.binary_config as app_binary_config
 import app.schemas.impl as impl
 import app.core.configuration as app_conf
+import app.models as models
 
 router = fastapi.APIRouter()
 LOGGER = logging.getLogger(__name__)
@@ -182,3 +185,76 @@ async def add_binary(
         model_id=model_id
     )
     return impl.EndpointImpl.from_database(crud.endpoint.get(db, id=m))
+
+
+@router.get(
+    path='/models/{model_id}/metadata',
+    response_model=ops_schemas.AdditionalModelInfo,
+    tags=['discover'])
+def get_model_metadata(
+        model_id: int,
+        db: saorm.Session = fastapi.Depends(deps.get_db)):
+    LOGGER.info('Retrieving model metadata for id: %s', model_id)
+    match crud.binary_ml_model.get(db=db, id=model_id):
+        case None:
+            raise fastapi.HTTPException(
+                status_code=status.HTTP_404_NOT_FOUND, detail=f'Model with id {model_id} is not found')
+        case models.BinaryMlModel(format=app_binary_config.ModelWrapper.PICKLE | app_binary_config.ModelWrapper.JOBLIB as format):
+            return ops_schemas.AdditionalModelInfo(
+                modelPackage=format.value,
+                modelType='other')
+        case models.BinaryMlModel(format=app_binary_config.ModelWrapper.PMML, model_b64=binary):
+            match app_runtime_inspection.inspect_pmml_subtype(binary):
+                case 'Scorecard' | 'RuleSetModel' as typ:
+                    return ops_schemas.AdditionalModelInfo(
+                        modelPackage='pmml',
+                        modelType=typ)
+                case _:
+                    return ops_schemas.AdditionalModelInfo(
+                        modelPackage='pmml',
+                        modelType='other')
+        case _:
+            return ops_schemas.AdditionalModelInfo(
+                modelPackage='other',
+                modelType='other')
+
+
+@router.get(
+    path='/models/{model_id}/download',
+    response_class=responses.StreamingResponse,
+    tags=['discover'])
+def get_model_binary(
+        model_id: int,
+        db: saorm.Session = fastapi.Depends(deps.get_db)):
+    LOGGER.info('Retrieving model binary for id: %s', model_id)
+
+    model = crud.model.get(db=db, id=model_id)
+
+    if model is None:
+        raise fastapi.HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail=f'Model with id {model_id} is not found')
+
+    filename = model.config.configuration['name']
+
+    try:
+        binary = model.endpoint.binary
+    except AttributeError:
+        raise fastapi.HTTPException(
+            status_code=status.HTTP_404_NOT_FOUND, detail=f'Model binary with id {model_id} is not found')
+
+    if binary.format == app_binary_config.ModelWrapper.PMML:
+        file_extension = 'pmml'
+    elif binary.format == app_binary_config.ModelWrapper.PICKLE:
+        file_extension = 'pickle'
+    elif binary.format == app_binary_config.ModelWrapper.JOBLIB:
+        file_extension = 'joblib'
+    else:
+        file_extension = 'bin'
+
+    LOGGER.info('Downloading model binary with name %s and format %s', filename, file_extension)
+
+    return responses.StreamingResponse(
+        content=io.BytesIO(binary.model_b64),
+        media_type='application/octet-stream',
+        headers={'Content-Disposition': 'attachment; filename="{basename}.{extension}"'.format(
+            basename=filename, extension=file_extension)})

ops-implementations/ads-ml-service/app/api/api_v2/endpoints/upload.py

@@ -77,7 +77,7 @@ async def upload(
         model_binary,
         input_data_structure,
         output_data_structure,
-        format_,
+        file_format,
         name=model_name
     )
 

ops-implementations/ads-ml-service/app/db/session.py

@@ -43,7 +43,11 @@ def get_db_opts() -> typing.Dict[str, typing.Any]:
     return opt
 
 
+def get_engine():
+    return create_engine(get_db_url(), **get_db_opts())
+
+
 SessionLocal = sessionmaker(
         autocommit=False,
         autoflush=False,
-        bind=create_engine(get_db_url(), **get_db_opts()))
+        bind=get_engine())