Merge pull request #122 from SAP/enable-inference-with-ext-url

Enable Inference With External URL

Author: SreevishnuAB

CHANGELOG.md

@@ -6,6 +6,14 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.10.0]
+
+### Added
+
+* Support for inference using external URL in `create_inference_request_with_url`
+
+[#122]: https://github.com/SAP/data-attribute-recommendation-python-sdk/pull/122
+
 ## [0.9.2]
 
 ### Added
@@ -17,7 +25,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 [#116]: https://github.com/SAP/data-attribute-recommendation-python-sdk/pull/116
 
-
 ## [0.9.1]
 
 ### Added
@@ -237,7 +244,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 * First public release
 
-[Unreleased]: https://github.com/SAP/data-attribute-recommendation-python-sdk/compare/rel/0.9.1...HEAD
+[Unreleased]: https://github.com/SAP/data-attribute-recommendation-python-sdk/compare/rel/0.10.0...HEAD
+[0.10.0]: https://github.com/SAP/data-attribute-recommendation-python-sdk/compare/rel/0.9.2...rel/0.10.0
+[0.9.2]: https://github.com/SAP/data-attribute-recommendation-python-sdk/compare/rel/0.9.1...rel/0.9.2
 [0.9.1]: https://github.com/SAP/data-attribute-recommendation-python-sdk/compare/rel/0.9.0...rel/0.9.1
 [0.9.0]: https://github.com/SAP/data-attribute-recommendation-python-sdk/compare/rel/0.8.2...rel/0.9.0
 [0.8.2]: https://github.com/SAP/data-attribute-recommendation-python-sdk/compare/rel/0.8.1...rel/0.8.2

sap/aibus/dar/client/dar_session.py

@@ -158,6 +158,21 @@ def post_data_to_endpoint(
         self._check_status_code(response, url)
         return response
 
+    def post_to_url(self, url: str, payload: dict, retry: bool = False) -> Response:
+        """
+        Performs **POST** request against fully-qualified URL
+
+        :param url: a fully-qualified inference URL
+        :param payload: request body
+        :param retry: enables retrying a failed request
+        """
+        connection = self.http
+        if retry:
+            connection = self.http_post_retry
+        response = connection.post(url, headers=self._get_headers(), json=payload)
+        self._check_status_code(response, url)
+        return response
+
 
 def _get_requests_version():
     requests_version = None

sap/aibus/dar/client/inference_client.py

@@ -121,3 +121,35 @@ def do_bulk_inference(
             )
             result.extend(response["predictions"])
         return result
+
+    def create_inference_request_with_url(
+        self,
+        url: str,
+        objects: List[dict],
+        top_n: int = TOP_N,
+        retry: bool = False,
+    ) -> dict:
+        """
+        Performs inference for the given *objects* against fully-qualified URL.
+        A complete inference URL can be the passed to the method inference, instead
+        of constructing URL from using base url and model name
+
+        :param url: fully-qualified inference URL
+        :param objects: Objects to be classified
+        :param top_n: How many predictions to return per object
+        :param retry: whether to retry on errors. Default: False
+        :return: API response
+        """
+        self.log.debug(
+            "Submitting Inference request with '%s'"
+            " objects and top_n '%s' to url %s",
+            len(objects),
+            top_n,
+            url,
+        )
+        response = self.session.post_to_url(
+            url, payload={"topN": top_n, "objects": objects}, retry=retry
+        )
+        as_json = response.json()
+        self.log.debug("Inference response ID: %s", as_json["id"])
+        return as_json

system_tests/workflow/test_end_to_end.py

@@ -1,9 +1,11 @@
 import logging
 import uuid
 from io import BytesIO
+import os
 
 import pytest
 
+from sap.aibus.dar.client.inference_constants import InferencePaths
 from sap.aibus.dar.client.data_manager_client import DataManagerClient
 from sap.aibus.dar.client.exceptions import DARHTTPException, ModelAlreadyExists
 from sap.aibus.dar.client.inference_client import InferenceClient
@@ -257,6 +259,18 @@ def _assert_inference_works(self, inference_client, model_name):
         )
         assert len(response) == 123
 
+        url = os.environ["DAR_URL"]
+        if url[-1] == "/":
+            url = url[:-1]
+        url = url + InferencePaths.format_inference_endpoint_by_name(model_name)
+        response = inference_client.create_inference_request_with_url(
+            url=url, objects=to_be_classified
+        )
+        logger.info("Inference with URL done. API response: %s", response)
+        print(response)
+        # One object has been classified
+        assert len(response["predictions"]) == 1
+
     def _assert_deployment_exists(self, deployment_id, model_manager_client):
         # Look at individual resource
         read_response = model_manager_client.read_deployment_by_id(deployment_id)

tests/sap/aibus/dar/client/test_dar_session.py

@@ -217,6 +217,45 @@ def test_post_data_to_endpoint_error_handling(self):
         assert exc.url == self.dar_url[:-1] + endpoint
         assert exc.response == sess.http.post()
 
+    def test_post_to_url(self):
+        for allowed_status_code in range(200, 300):
+            sess = self._prepare()
+            sess.http.post.return_value.status_code = allowed_status_code
+
+            url = self.dar_url[:-1] + "/data-manager/api/v3/datasetSchemas/"
+
+            payload = {"a": 1, "b": "ok!"}
+
+            response = sess.post_to_url(url=url, payload=payload)
+
+            expected_http_call = call(
+                url,
+                headers=self.expected_headers,
+                json=payload,
+            )
+            assert getattr(sess.http, "post").call_args_list == [expected_http_call]
+            assert response == sess.http.post.return_value
+
+    def test_post_to_url_with_retry(self):
+        for allowed_status_code in range(200, 300):
+            sess = self._prepare()
+            sess.http_post_retry.post.return_value.status_code = allowed_status_code
+
+            url = self.dar_url[:-1] + "/data-manager/api/v3/datasetSchemas/"
+
+            payload = {"a": 1, "b": "ok!"}
+
+            response = sess.post_to_url(url=url, payload=payload, retry=True)
+            expected_http_call = call(
+                url,
+                headers=self.expected_headers,
+                json=payload,
+            )
+            assert getattr(sess.http_post_retry, "post").call_args_list == [
+                expected_http_call
+            ]
+            assert response == sess.http_post_retry.post.return_value
+
     def _assert(self, sess, method, endpoint):
         # Validate test-internal assumption
         assert self.dar_url[-1] == "/"

tests/sap/aibus/dar/client/test_data_manager_client.py

@@ -679,7 +679,7 @@ def prepare_client(dar_url: str, clazz):
     mock_session = create_autospec(DARSession, instance=True)
     mock_session.get_from_endpoint.return_value = mock_response
     mock_session.post_to_endpoint.return_value = mock_response
-
+    mock_session.post_to_url.return_value = mock_response
     client = clazz.construct_from_jwt(dar_url, "abcd")
     client.session = mock_session
     return client

tests/sap/aibus/dar/client/test_inference_client.py

@@ -203,3 +203,50 @@ def _assert_bulk_inference_works(
             inference_client.session.post_to_endpoint.call_args_list
             == expected_calls_to_post
         )
+
+    def test_create_inference_with_url_works(self, inference_client: InferenceClient):
+        """
+        Checks inference call.
+        """
+        url = DAR_URL + "inference/api/v3/models/my-model/versions/1"
+        response = inference_client.create_inference_request_with_url(
+            url, objects=self.objects
+        )
+
+        expected_call = call(
+            url,
+            payload={"topN": 1, "objects": self.objects},
+            retry=False,
+        )
+
+        assert inference_client.session.post_to_url.call_args_list == [expected_call]
+
+        assert (
+            inference_client.session.post_to_url.return_value.json.return_value
+            == response
+        )
+
+    def test_create_inference_request_with_url_retry_enabled(
+        self, inference_client: InferenceClient
+    ):
+        """
+        Checks if retry parameter is passed correctly.
+        """
+        url = DAR_URL + "inference/api/v3/models/my-model/versions/1"
+
+        response = inference_client.create_inference_request_with_url(
+            url=url, objects=self.objects, retry=True
+        )
+
+        expected_call = call(
+            url,
+            payload={"topN": 1, "objects": self.objects},
+            retry=True,
+        )
+
+        assert inference_client.session.post_to_url.call_args_list == [expected_call]
+
+        assert (
+            inference_client.session.post_to_url.return_value.json.return_value
+            == response
+        )

version.txt

@@ -1 +1 @@
-0.9.1
+0.10.0