cmem-plugin-pgvector 0.6.3__tar.gz → 0.7.0__tar.gz

{cmem_plugin_pgvector-0.6.3 → cmem_plugin_pgvector-0.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: cmem-plugin-pgvector
-Version: 0.6.3
+Version: 0.7.0
 Summary: Store and search for embedding vectors in a Postgres vector store.
 License: Apache-2.0
 Keywords: eccenca Corporate Memory,plugin

{cmem_plugin_pgvector-0.6.3 → cmem_plugin_pgvector-0.7.0}/cmem_plugin_pgvector/search_task.py

@@ -13,10 +13,11 @@ from cmem_plugin_base.dataintegration.ports import (
     FixedNumberOfInputs,
     FixedSchemaPort,
 )
-from cmem_plugin_base.dataintegration.types import IntParameterType
+from cmem_plugin_base.dataintegration.types import EnumParameterType, IntParameterType
 from langchain_core.documents import Document
 from langchain_core.embeddings import Embeddings
 from langchain_postgres import PGVector
+from langchain_postgres.vectorstores import DistanceStrategy
 
 from cmem_plugin_pgvector.commons import DatabaseParams
 
@@ -86,6 +87,14 @@ The results in this output are structured like this:
             default_value=10,
             param_type=IntParameterType(),
         ),
+        PluginParameter(
+            name="distance_strategy",
+            label="Distance Strategy",
+            description="The distance strategy to use. (default: COSINE)",
+            param_type=EnumParameterType(enum_type=DistanceStrategy),
+            default_value=DistanceStrategy.COSINE,
+            advanced=True,
+        ),
     ],
 )
 class PGVectorSearchPlugin(WorkflowPlugin):
@@ -105,6 +114,7 @@ class PGVectorSearchPlugin(WorkflowPlugin):
     report: ExecutionReport
     search_result_path: str
     top_k: int
+    distance_strategy: DistanceStrategy
 
     def __init__(  # noqa: PLR0913
         self,
@@ -117,6 +127,7 @@ class PGVectorSearchPlugin(WorkflowPlugin):
         search_result_path: str = "_search_result",
         embedding_query_path: str = "_embedding",
         top_k: int = 10,
+        distance_strategy: DistanceStrategy = DistanceStrategy.COSINE,
     ) -> None:
         self.collection_name = collection_name
         self.user = user
@@ -126,6 +137,7 @@ class PGVectorSearchPlugin(WorkflowPlugin):
         self.embedding_query_path = embedding_query_path
         self.search_result_path = search_result_path
         self.top_k = top_k
+        self.distance_strategy = distance_strategy
 
         str_password = self.password = password if isinstance(password, str) else password.decrypt()
         self.connection_string = (
@@ -142,6 +154,7 @@ class PGVectorSearchPlugin(WorkflowPlugin):
             embeddings=DummyEmbeddings(),
             use_jsonb=True,
             pre_delete_collection=False,
+            distance_strategy=distance_strategy,
         )
         self._setup_ports()
 

{cmem_plugin_pgvector-0.6.3 → cmem_plugin_pgvector-0.7.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "cmem-plugin-pgvector"
-version = "0.6.3"
+version = "0.7.0"
 license = "Apache-2.0"
 description = "Store and search for embedding vectors in a Postgres vector store."
 authors = ["eccenca GmbH <cmempy-developer@eccenca.com>"]