pgvector-haystack 6.6.0__tar.gz → 7.0.0__tar.gz

{pgvector_haystack-6.6.0 → pgvector_haystack-7.0.0}/CHANGELOG.md

@@ -1,5 +1,33 @@
 # Changelog
 
+## [integrations/pgvector-v6.6.1] - 2026-09-25
+
+### 🐛 Bug Fixes
+
+- *(pgvector)* Raise FilterError on empty logical conditions (#3946)
+- *(pgvector)* Count every written document in write_documents_async (#3964)
+
+### ⚙️ CI
+
+- Improve changelog generation; fix existing changelogs (#3883)
+
+
+## [integrations/pgvector-v6.6.0] - 2026-08-26
+
+### 🐛 Bug Fixes
+
+- PGVector adding optional parameter `filters` to `get_metadata_field_unique_values()` (#3735)
+- Standardize license files (#3771)
+
+### 🧪 Testing
+
+- `PGVectorDocumentStore` removing duplicated tests already covered by haystack Mixin tests (#3783)
+
+### 🧹 Chores
+
+- Re-standardize READMEs (#3768)
+
+
 ## [integrations/pgvector-v6.5.0] - 2026-08-03
 
 ### 🚀 Features
@@ -147,9 +175,14 @@
 
 ### 🧹 Chores
 
-- Remove black (#1985)
 - Pgvector - update docker image in examples (#1995)
 
+## [integrations/pgvector-v5.2.0] - 2025-06-26
+
+### 🧹 Chores
+
+- Remove black (#1985)
+
 
 ## [integrations/pgvector-v3.4.1] - 2025-06-20
 
@@ -253,48 +286,75 @@
 
 ### 🚀 Features
 
-- Add filter_policy to pgvector integration (#820)
 - Add schema support to pgvector document store. (#1095)
 - Pgvector - recreate the connection if it is no longer valid (#1202)
 
-### 🐛 Bug Fixes
-
-- `PgVector` - Fallback to default filter policy when deserializing retrievers without the init parameter (#900)
-
 ### 📚 Documentation
 
 - Explain different connection string formats in the docstring (#1132)
 
+### ⚙️ CI
+
+- Adopt uv as installer (#1142)
+
+### 🧹 Chores
+
+- Update changelog after removing legacy filters (#1083)
+- Update ruff linting scripts and settings (#1105)
+
+## [integrations/pgvector-v1.0.0] - 2024-09-12
+
 ### 🧪 Testing
 
 - Do not retry tests in `hatch run test` command (#954)
 
+### 🧹 Chores
+
+- PgVector - remove legacy filter support (#1068)
+
+### 🌀 Miscellaneous
+
+- Chore: Update pgvector test for the new `apply_filter_policy` usage (#970)
+- Chore: pgvector ruff update, don't ruff tests (#984)
+
+## [integrations/pgvector-v0.5.1] - 2024-07-17
+
+### 🐛 Bug Fixes
+
+- `PgVector` - Fallback to default filter policy when deserializing retrievers without the init parameter (#900)
+
+## [integrations/pgvector-v0.5.0] - 2024-07-15
+
+### 🚀 Features
+
+- Add filter_policy to pgvector integration (#820)
+
 ### ⚙️ CI
 
 - Retry tests to reduce flakyness (#836)
-- Adopt uv as installer (#1142)
 
 ### 🧹 Chores
 
 - Update ruff invocation to include check parameter (#853)
-- PgVector - remove legacy filter support (#1068)
-- Update changelog after removing legacy filters (#1083)
-- Update ruff linting scripts and settings (#1105)
 
 ### 🌀 Miscellaneous
 
 - Ci: install `pytest-rerunfailures` where needed; add retry config to `test-cov` script (#845)
 - Chore: Minor retriever pydoc fix (#884)
-- Chore: Update pgvector test for the new `apply_filter_policy` usage (#970)
-- Chore: pgvector ruff update, don't ruff tests (#984)
 
 ## [integrations/pgvector-v0.4.0] - 2024-06-20
 
 ### 🚀 Features
 
-- Defer the database connection to when it's needed (#773)
 - Add customizable index names for pgvector (#818)
 
+
+## [integrations/pgvector-v0.3.0] - 2024-05-30
+
+### 🚀 Features
+
+- Defer the database connection to when it's needed (#773)
+
 ### 🌀 Miscellaneous
 
 - Docs: add missing api references (#728)
@@ -328,6 +388,14 @@
 
 ## [integrations/pgvector-v0.1.0] - 2024-02-14
 
+### 🌀 Miscellaneous
+
+- Pgvector: generate API docs (#325)
+- Pgvector: add an example (#334)
+- Adopt `Secret` to pgvector (#402)
+
+## [integrations/pgvector-v0.0.1] - 2024-02-02
+
 ### 🐛 Bug Fixes
 
 - Pgvector: fix linting (#328)
@@ -338,8 +406,5 @@
 - Pgvector - filters (#257)
 - Pgvector - embedding retrieval (#298)
 - Pgvector - Embedding Retriever (#320)
-- Pgvector: generate API docs (#325)
-- Pgvector: add an example (#334)
-- Adopt `Secret` to pgvector (#402)
 
 <!-- generated by git-cliff -->

{pgvector_haystack-6.6.0 → pgvector_haystack-7.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: pgvector-haystack
-Version: 6.6.0
+Version: 7.0.0
 Summary: An integration of pgvector (vector search extension for Postgres) with Haystack
 Project-URL: Source, https://github.com/deepset-ai/haystack-core-integrations
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/blob/main/integrations/pgvector/README.md

{pgvector_haystack-6.6.0 → pgvector_haystack-7.0.0}/src/haystack_integrations/document_stores/pgvector/document_store.py

@@ -9,6 +9,7 @@ from haystack import default_from_dict, default_to_dict, logging
 from haystack.dataclasses.document import Document
 from haystack.document_stores.errors import DocumentStoreError, DuplicateDocumentError
 from haystack.document_stores.types import DuplicatePolicy
+from haystack.errors import FilterError
 from haystack.utils.auth import Secret, deserialize_secrets_inplace
 from psycopg import AsyncConnection, Connection, Cursor, Error, IntegrityError
 from psycopg.cursor_async import AsyncCursor
@@ -923,7 +924,7 @@ class PgvectorDocumentStore:
             )
             raise DocumentStoreError(error_msg) from e
 
-        # get the number of the inserted documents, inspired by psycopg3 docs
+        # executemany(returning=True) yields one result set per document, so walk them all with nextset()
         # https://www.psycopg.org/psycopg3/docs/api/cursors.html#psycopg.Cursor.executemany
         written_docs = 0
         while True:
@@ -980,9 +981,14 @@ class PgvectorDocumentStore:
             )
             raise DocumentStoreError(error_msg) from e
 
+        # executemany(returning=True) yields one result set per document, so walk them all with nextset()
+        # https://www.psycopg.org/psycopg3/docs/api/cursors.html#psycopg.Cursor.executemany
         written_docs = 0
-        async for _ in self._async_cursor:
-            written_docs += 1
+        while True:
+            if await self._async_cursor.fetchone():
+                written_docs += 1
+            if not self._async_cursor.nextset():
+                break
 
         return written_docs
 
@@ -1073,18 +1079,21 @@ class PgvectorDocumentStore:
         :param filters: The filters to apply to select documents for deletion.
             For filter syntax, see [Haystack metadata filtering](https://docs.haystack.deepset.ai/docs/metadata-filtering)
         :returns: The number of documents deleted.
+        :raises FilterError: If `filters` is empty. Use `delete_all_documents()` to delete everything.
         """
         _validate_filters(filters)
+        if not filters:
+            # An empty filter would compile to an unqualified DELETE and empty the table
+            msg = "delete_by_filter requires a non-empty filter. Use delete_all_documents() to delete every document."
+            raise FilterError(msg)
 
         delete_sql = SQL("DELETE FROM {schema_name}.{table_name}").format(
             schema_name=Identifier(self.schema_name),
             table_name=Identifier(self.table_name),
         )
 
-        params = ()
-        if filters:
-            sql_where_clause, params = _convert_filters_to_where_clause_and_params(filters)
-            delete_sql += sql_where_clause
+        sql_where_clause, params = _convert_filters_to_where_clause_and_params(filters)
+        delete_sql += sql_where_clause
 
         self._ensure_db_setup()
         assert self._cursor is not None
@@ -1115,18 +1124,24 @@ class PgvectorDocumentStore:
         :param filters: The filters to apply to select documents for deletion.
             For filter syntax, see [Haystack metadata filtering](https://docs.haystack.deepset.ai/docs/metadata-filtering)
         :returns: The number of documents deleted.
+        :raises FilterError: If `filters` is empty. Use `delete_all_documents_async()` to delete everything.
         """
         _validate_filters(filters)
+        if not filters:
+            # An empty filter would compile to an unqualified DELETE and empty the table
+            msg = (
+                "delete_by_filter_async requires a non-empty filter. "
+                "Use delete_all_documents_async() to delete every document."
+            )
+            raise FilterError(msg)
 
         delete_sql = SQL("DELETE FROM {schema_name}.{table_name}").format(
             schema_name=Identifier(self.schema_name),
             table_name=Identifier(self.table_name),
         )
 
-        params = ()
-        if filters:
-            sql_where_clause, params = _convert_filters_to_where_clause_and_params(filters)
-            delete_sql += sql_where_clause
+        sql_where_clause, params = _convert_filters_to_where_clause_and_params(filters)
+        delete_sql += sql_where_clause
 
         await self._ensure_db_setup_async()
         assert self._async_cursor is not None

{pgvector_haystack-6.6.0 → pgvector_haystack-7.0.0}/src/haystack_integrations/document_stores/pgvector/filters.py

@@ -71,6 +71,10 @@ def _parse_logical_condition(condition: dict[str, Any]) -> tuple[Composed, list[
         msg = f"Unknown logical operator '{operator}'. Valid operators are: 'AND', 'OR'"
         raise FilterError(msg)
 
+    if not condition["conditions"]:
+        msg = f"'conditions' must not be empty in {condition}"
+        raise FilterError(msg)
+
     # logical conditions can be nested, so we need to parse them recursively
     conditions = []
     for c in condition["conditions"]:

{pgvector_haystack-6.6.0 → pgvector_haystack-7.0.0}/tests/test_document_store.py

@@ -9,6 +9,7 @@ import pytest
 from haystack.dataclasses.document import ByteStream, Document
 from haystack.document_stores.errors import DocumentStoreError, DuplicateDocumentError
 from haystack.document_stores.types import DuplicatePolicy
+from haystack.errors import FilterError
 from haystack.testing.document_store import (
     CountDocumentsByFilterTest,
     CountDocumentsTest,
@@ -66,6 +67,20 @@ class TestDocumentStore(
         with pytest.raises(DuplicateDocumentError):
             document_store.write_documents(docs, DuplicatePolicy.FAIL)
 
+    def test_write_documents_counts_every_document(self, document_store: PgvectorDocumentStore):
+        docs = [Document(id="1"), Document(id="2"), Document(id="3")]
+        assert document_store.write_documents(docs) == 3
+
+    def test_write_documents_skip_counts_only_new_documents(self, document_store: PgvectorDocumentStore):
+        document_store.write_documents([Document(id="1")])
+        docs = [Document(id="1"), Document(id="2"), Document(id="3")]
+        assert document_store.write_documents(docs, DuplicatePolicy.SKIP) == 2
+
+    def test_write_documents_overwrite_counts_every_document(self, document_store: PgvectorDocumentStore):
+        document_store.write_documents([Document(id="1", content="old")])
+        docs = [Document(id="1", content="new"), Document(id="2"), Document(id="3")]
+        assert document_store.write_documents(docs, DuplicatePolicy.OVERWRITE) == 3
+
     def test_get_metadata_field_unique_values_distinct_types(self, document_store: PgvectorDocumentStore):
         """
         Override: the base mixin test stores int, float, str and bool under the *same* metadata field
@@ -151,6 +166,12 @@ class TestDocumentStore(
         assert document_store._connection is not None
 
 
+def test_delete_by_filter_rejects_an_empty_filter(mock_store):
+    """An empty filter would compile to an unqualified DELETE and empty the table."""
+    with pytest.raises(FilterError, match="non-empty filter"):
+        mock_store.delete_by_filter({})
+
+
 @pytest.mark.usefixtures("patches_for_unit_tests")
 def test_init(monkeypatch):
     monkeypatch.setenv("PG_CONN_STR", "some_connection_string")

{pgvector_haystack-6.6.0 → pgvector_haystack-7.0.0}/tests/test_document_store_async.py

@@ -9,6 +9,7 @@ import pytest
 from haystack.dataclasses.document import ByteStream, Document
 from haystack.document_stores.errors import DocumentStoreError, DuplicateDocumentError
 from haystack.document_stores.types import DuplicatePolicy
+from haystack.errors import FilterError
 from haystack.testing.document_store_async import (
     CountDocumentsAsyncTest,
     CountDocumentsByFilterAsyncTest,
@@ -71,6 +72,20 @@ class TestDocumentStoreAsync(
         with pytest.raises(DuplicateDocumentError):
             await document_store.write_documents_async(docs, DuplicatePolicy.FAIL)
 
+    async def test_write_documents_async_counts_every_document(self, document_store: PgvectorDocumentStore):
+        docs = [Document(id="1"), Document(id="2"), Document(id="3")]
+        assert await document_store.write_documents_async(docs) == 3
+
+    async def test_write_documents_async_skip_counts_only_new_documents(self, document_store: PgvectorDocumentStore):
+        await document_store.write_documents_async([Document(id="1")])
+        docs = [Document(id="1"), Document(id="2"), Document(id="3")]
+        assert await document_store.write_documents_async(docs, DuplicatePolicy.SKIP) == 2
+
+    async def test_write_documents_async_overwrite_counts_every_document(self, document_store: PgvectorDocumentStore):
+        await document_store.write_documents_async([Document(id="1", content="old")])
+        docs = [Document(id="1", content="new"), Document(id="2"), Document(id="3")]
+        assert await document_store.write_documents_async(docs, DuplicatePolicy.OVERWRITE) == 3
+
     async def test_get_metadata_field_unique_values_distinct_types_async(self, document_store: PgvectorDocumentStore):
         """
         Override: the base mixin test stores int, float, str and bool under the *same* metadata field
@@ -353,6 +368,13 @@ async def test_write_documents_async_rejects_non_document_items(mock_store):
         await mock_store.write_documents_async([{"not": "a document"}])
 
 
+@pytest.mark.asyncio
+async def test_delete_by_filter_async_rejects_an_empty_filter(mock_store):
+    """Same contract as the sync path, and the message must name the async methods."""
+    with pytest.raises(FilterError, match=r"delete_by_filter_async requires .* delete_all_documents_async\(\)"):
+        await mock_store.delete_by_filter_async({})
+
+
 @pytest.mark.asyncio
 async def test_count_unique_metadata_by_filter_async_rejects_empty_fields(mock_store):
     with pytest.raises(ValueError, match="metadata_fields must be a non-empty list"):

{pgvector_haystack-6.6.0 → pgvector_haystack-7.0.0}/tests/test_filters.py

@@ -483,3 +483,21 @@ def test_validate_filters():
     invalid_filters = {"field": "meta.number", "value": "100"}
     with pytest.raises(ValueError):
         _validate_filters(invalid_filters)
+
+
+@pytest.mark.parametrize("operator", ["AND", "OR"])
+def test_logical_condition_empty_conditions_raises(operator):
+    with pytest.raises(FilterError):
+        _parse_logical_condition({"operator": operator, "conditions": []})
+
+
+def test_logical_condition_nested_empty_conditions_raises():
+    filters = {
+        "operator": "AND",
+        "conditions": [
+            {"field": "meta.type", "operator": "==", "value": "article"},
+            {"operator": "OR", "conditions": []},
+        ],
+    }
+    with pytest.raises(FilterError):
+        _convert_filters_to_where_clause_and_params(filters)