feat: Support filters in matching engine vector matching (#1608)

jaycee-li · nayaknishant · web-flow · commit d591d3e36b76 · 2022-08-25T19:47:22.000-07:00
* feat: support filter in index_enpoint.match()

* fix type error

* Add unit test for index_endpoint.match()

* update docstring example

* Update docstring

Co-authored-by: nayaknishant &lt;nishantnayak@google.com&gt;
diff --git a/google/cloud/aiplatform/matching_engine/matching_engine_index_endpoint.py b/google/cloud/aiplatform/matching_engine/matching_engine_index_endpoint.py
@@ -15,7 +15,7 @@
 # limitations under the License.
 #
 
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Dict, List, Optional, Sequence, Tuple
 
 from google.auth import credentials as auth_credentials
@@ -51,6 +51,25 @@ class MatchNeighbor:
     distance: float
 
 
+@dataclass
+class Namespace:
+    """Namespace specifies the rules for determining the datapoints that are eligible for each matching query, overall query is an AND across namespaces.
+    Args:
+        name (str):
+            Required. The name of this Namespace.
+        allow_tokens (List(str)):
+            Optional. The allowed tokens in the namespace.
+        deny_tokens (List(str)):
+            Optional. The denied tokens in the namespace. When a token is denied, then matches will be excluded whenever the other datapoint has that token.
+            For example, if a query specifies [Namespace("color", ["red","blue"], ["purple"])], then that query will match datapoints that are red or blue,
+            but if those points are also purple, then they will be excluded even if they are red/blue.
+    """
+
+    name: str
+    allow_tokens: list = field(default_factory=list)
+    deny_tokens: list = field(default_factory=list)
+
+
 class MatchingEngineIndexEndpoint(base.VertexAiResourceNounWithFutureManager):
     """Matching Engine index endpoint resource for Vertex AI."""
 
@@ -796,7 +815,11 @@ def description(self) -> str:
         return self._gca_resource.description
 
     def match(
-        self, deployed_index_id: str, queries: List[List[float]], num_neighbors: int = 1
+        self,
+        deployed_index_id: str,
+        queries: List[List[float]],
+        num_neighbors: int = 1,
+        filter: Optional[List[Namespace]] = [],
     ) -> List[List[MatchNeighbor]]:
         """Retrieves nearest neighbors for the given embedding queries on the specified deployed index.
 
@@ -808,6 +831,11 @@ def match(
             num_neighbors (int):
                 Required. The number of nearest neighbors to be retrieved from database for
                 each query.
+            filter (List[Namespace]):
+                Optional. A list of Namespaces for filtering the matching results.
+                For example, [Namespace("color", ["red"], []), Namespace("shape", [], ["squared"])] will match datapoints
+                that satisfy "red color" but not include datapoints with "squared shape".
+                Please refer to https://cloud.google.com/vertex-ai/docs/matching-engine/filtering#json for more detail.
 
         Returns:
             List[List[MatchNeighbor]] - A list of nearest neighbors for each query.
@@ -836,16 +864,22 @@ def match(
             match_service_pb2.BatchMatchRequest.BatchMatchRequestPerIndex()
         )
         batch_request_for_index.deployed_index_id = deployed_index_id
-        batch_request_for_index.requests.extend(
-            [
-                match_service_pb2.MatchRequest(
-                    num_neighbors=num_neighbors,
-                    deployed_index_id=deployed_index_id,
-                    float_val=query,
-                )
-                for query in queries
-            ]
-        )
+        requests = []
+        for query in queries:
+            request = match_service_pb2.MatchRequest(
+                num_neighbors=num_neighbors,
+                deployed_index_id=deployed_index_id,
+                float_val=query,
+            )
+            for namespace in filter:
+                restrict = match_service_pb2.Namespace()
+                restrict.name = namespace.name
+                restrict.allow_tokens.extend(namespace.allow_tokens)
+                restrict.deny_tokens.extend(namespace.deny_tokens)
+                request.restricts.append(restrict)
+            requests.append(request)
+
+        batch_request_for_index.requests.extend(requests)
         batch_request.requests.append(batch_request_for_index)
 
         # Perform the request
diff --git a/tests/system/aiplatform/test_matching_engine_index.py b/tests/system/aiplatform/test_matching_engine_index.py
@@ -18,6 +18,9 @@
 import uuid
 
 from google.cloud import aiplatform
+from google.cloud.aiplatform.matching_engine.matching_engine_index_endpoint import (
+    Namespace,
+)
 
 from tests.system.aiplatform import e2e_base
 
@@ -161,6 +164,8 @@
     -0.021106,
 ]
 
+_TEST_FILTER = [Namespace("name", ["allow_token"], ["deny_token"])]
+
 
 class TestMatchingEngine(e2e_base.TestEndToEnd):
 
@@ -283,6 +288,16 @@ def test_create_get_list_matching_engine_index(self, shared_state):
 
         # assert results[0][0].id == 870
 
+        # TODO: Test `my_index_endpoint.match` with filter.
+        # This requires uploading a new content of the Matching Engine Index to Cloud Storage.
+        # results = my_index_endpoint.match(
+        #     deployed_index_id=_TEST_DEPLOYED_INDEX_ID,
+        #     queries=[_TEST_MATCH_QUERY],
+        #     num_neighbors=1,
+        #     filter=_TEST_FILTER,
+        # )
+        # assert results[0][0].id == 9999
+
         # Undeploy index
         my_index_endpoint = my_index_endpoint.undeploy_index(
             deployed_index_id=deployed_index.id
diff --git a/tests/unit/aiplatform/test_matching_engine_index_endpoint.py b/tests/unit/aiplatform/test_matching_engine_index_endpoint.py
@@ -24,6 +24,10 @@
 from google.cloud import aiplatform
 from google.cloud.aiplatform import base
 from google.cloud.aiplatform import initializer
+from google.cloud.aiplatform.matching_engine._protos import match_service_pb2
+from google.cloud.aiplatform.matching_engine.matching_engine_index_endpoint import (
+    Namespace,
+)
 from google.cloud.aiplatform.compat.types import (
     matching_engine_deployed_index_ref as gca_matching_engine_deployed_index_ref,
     index_endpoint as gca_index_endpoint,
@@ -37,6 +41,8 @@
 
 from google.protobuf import field_mask_pb2
 
+import grpc
+
 import pytest
 
 # project
@@ -210,6 +216,9 @@
     ]
 ]
 _TEST_NUM_NEIGHBOURS = 1
+_TEST_FILTER = [
+    Namespace(name="class", allow_tokens=["token_1"], deny_tokens=["token_2"])
+]
 
 
 def uuid_mock():
@@ -380,6 +389,33 @@ def create_index_endpoint_mock():
         yield create_index_endpoint_mock
 
 
+@pytest.fixture
+def index_endpoint_match_queries_mock():
+    with patch.object(
+        grpc._channel._UnaryUnaryMultiCallable,
+        "__call__",
+    ) as index_endpoint_match_queries_mock:
+        index_endpoint_match_queries_mock.return_value = (
+            match_service_pb2.BatchMatchResponse(
+                responses=[
+                    match_service_pb2.BatchMatchResponse.BatchMatchResponsePerIndex(
+                        deployed_index_id="1",
+                        responses=[
+                            match_service_pb2.MatchResponse(
+                                neighbor=[
+                                    match_service_pb2.MatchResponse.Neighbor(
+                                        id="1", distance=0.1
+                                    )
+                                ]
+                            )
+                        ],
+                    )
+                ]
+            )
+        )
+        yield index_endpoint_match_queries_mock
+
+
 @pytest.mark.usefixtures("google_auth_mock")
 class TestMatchingEngineIndexEndpoint:
     def setup_method(self):
@@ -617,3 +653,42 @@ def test_delete_index_endpoint_with_force(
         delete_index_endpoint_mock.assert_called_once_with(
             name=_TEST_INDEX_ENDPOINT_NAME
         )
+
+    @pytest.mark.usefixtures("get_index_endpoint_mock")
+    def test_index_endpoint_match_queries(self, index_endpoint_match_queries_mock):
+        aiplatform.init(project=_TEST_PROJECT)
+
+        my_index_endpoint = aiplatform.MatchingEngineIndexEndpoint(
+            index_endpoint_name=_TEST_INDEX_ENDPOINT_ID
+        )
+
+        my_index_endpoint.match(
+            deployed_index_id=_TEST_DEPLOYED_INDEX_ID,
+            queries=_TEST_QUERIES,
+            num_neighbors=_TEST_NUM_NEIGHBOURS,
+            filter=_TEST_FILTER,
+        )
+
+        batch_request = match_service_pb2.BatchMatchRequest(
+            requests=[
+                match_service_pb2.BatchMatchRequest.BatchMatchRequestPerIndex(
+                    deployed_index_id=_TEST_DEPLOYED_INDEX_ID,
+                    requests=[
+                        match_service_pb2.MatchRequest(
+                            num_neighbors=_TEST_NUM_NEIGHBOURS,
+                            deployed_index_id=_TEST_DEPLOYED_INDEX_ID,
+                            float_val=_TEST_QUERIES[0],
+                            restricts=[
+                                match_service_pb2.Namespace(
+                                    name="class",
+                                    allow_tokens=["token_1"],
+                                    deny_tokens=["token_2"],
+                                )
+                            ],
+                        )
+                    ],
+                )
+            ]
+        )
+
+        index_endpoint_match_queries_mock.assert_called_with(batch_request)
