Implementation:Langchain ai Langchain VectorStoreIntegrationTests

Knowledge Sources	Langchain_ai_Langchain
Domains	Testing, Vector Store, Standard Tests
Last Updated	2026-02-11 00:00 GMT

Overview

Standard integration test suite for validating VectorStore implementations, covering document addition, similarity search, deletion, idempotency, and get_by_ids in both sync and async modes.

Description

VectorStoreIntegrationTests is an abstract test suite class in the langchain-tests (standard-tests) package that provides comprehensive integration tests for VectorStore implementations from langchain-core. The suite tests document addition, similarity search correctness, document deletion (single and bulk), idempotent additions, mutation via overwrite, and retrieval by IDs. It includes both synchronous and asynchronous variants of each test. The suite uses DeterministicFakeEmbedding from langchain-core for deterministic, reproducible embeddings during testing.

Usage

Import VectorStoreIntegrationTests when writing integration tests for a custom VectorStore implementation to ensure it conforms to the standard vector store contract.

Code Reference

Source Location

• Repository: Langchain_ai_Langchain
• File: libs/standard-tests/langchain_tests/integration_tests/vectorstores.py
• Lines: 1-842

Signature

class VectorStoreIntegrationTests(BaseStandardTests):
    @abstractmethod
    @pytest.fixture
    def vectorstore(self) -> VectorStore: ...

    @property
    def has_sync(self) -> bool: ...

    @property
    def has_async(self) -> bool: ...

    @property
    def has_get_by_ids(self) -> bool: ...

    @staticmethod
    def get_embeddings() -> Embeddings: ...

Import

from langchain_tests.integration_tests.vectorstores import VectorStoreIntegrationTests

I/O Contract

Inputs (Fixtures)

Name	Type	Required	Description
vectorstore	VectorStore	Yes	An empty VectorStore instance to test. Must be empty at the start of each test.

Configuration Properties

Name	Type	Default	Description
has_sync	bool	True	Whether to run synchronous tests.
has_async	bool	True	Whether to run asynchronous tests.
has_get_by_ids	bool	True	Whether the VectorStore supports get_by_ids.

Outputs

Name	Type	Description
Test results	pytest outcomes	Pass/fail results for each standard test method.

Test Methods (Sync)

Test Method	Description
test_vectorstore_is_empty	Verifies the store starts empty (similarity_search returns []).
test_add_documents	Tests adding documents and verifying similarity search order.
test_vectorstore_still_empty	Verifies the fixture resets the store to empty between tests.
test_deleting_documents	Tests deletion of a single document by ID.
test_deleting_bulk_documents	Tests bulk deletion of multiple documents.
test_delete_missing_content	Verifies deleting non-existent IDs does not raise exceptions.
test_add_documents_with_ids_is_idempotent	Verifies adding the same documents twice with same IDs does not duplicate.
test_add_documents_by_id_with_mutation	Tests overwriting a document by re-adding with the same ID.
test_get_by_ids	Tests retrieval of documents by their IDs.
test_get_by_ids_missing	Tests that get_by_ids with non-existent IDs returns empty list.
test_add_documents_documents	Tests add_documents with auto-generated IDs and get_by_ids verification.
test_add_documents_with_existing_ids	Tests that Document.id field is preserved during add_documents.

Test Methods (Async)

The async suite provides equivalent tests using aadd_documents, asimilarity_search, adelete, and aget_by_ids methods:

• test_vectorstore_is_empty_async
• test_add_documents_async
• test_vectorstore_still_empty_async
• test_deleting_documents_async
• test_deleting_bulk_documents_async
• test_delete_missing_content_async
• test_add_documents_with_ids_is_idempotent_async
• test_add_documents_by_id_with_mutation_async
• test_get_by_ids_async
• test_get_by_ids_missing_async
• test_add_documents_documents_async
• test_add_documents_with_existing_ids_async

Usage Examples

Basic Usage

from typing import Generator

import pytest
from langchain_core.vectorstores import VectorStore
from langchain_tests.integration_tests.vectorstores import VectorStoreIntegrationTests

from langchain_chroma import Chroma


class TestChromaStandard(VectorStoreIntegrationTests):
    @pytest.fixture()
    def vectorstore(self) -> Generator[VectorStore, None, None]:
        """Get an empty VectorStore for tests."""
        store = Chroma(embedding_function=self.get_embeddings())
        try:
            yield store
        finally:
            store.delete_collection()

Disabling Async Tests

class TestMyVectorStore(VectorStoreIntegrationTests):
    @pytest.fixture()
    def vectorstore(self) -> Generator[VectorStore, None, None]:
        ...

    @property
    def has_async(self) -> bool:
        return False

Embedding Configuration

The suite provides a get_embeddings() static method that returns a DeterministicFakeEmbedding with size 6. This produces deterministic but meaningless embeddings based on a hash of the input text. All test implementations should use this method to ensure reproducibility:

store = MyVectorStore(embedding_function=self.get_embeddings())