test_as_retriever_sparse_mode_with_search_kwargs() — langchain Function Reference

Source Code

libs/partners/qdrant/tests/integration_tests/qdrant_vector_store/test_search.py lines 386–413

def test_as_retriever_sparse_mode_with_search_kwargs(location: str) -> None:
    """Test as_retriever() with custom search_kwargs in SPARSE mode."""
    # Use from_texts to create the vectorstore, which handles collection creation
    texts = ["test document"]
    vectorstore = QdrantVectorStore.from_texts(
        texts,
        embedding=None,  # No dense embedding for SPARSE mode
        location=location,
        retrieval_mode=RetrievalMode.SPARSE,
        sparse_embedding=ConsistentFakeSparseEmbeddings(),
        sparse_vector_name="sparse",
    )

    # Add test documents
    docs = [
        Document(page_content="Python programming", metadata={"topic": "programming"}),
        Document(page_content="Machine learning", metadata={"topic": "AI"}),
        Document(page_content="Data analysis", metadata={"topic": "data"}),
    ]
    vectorstore.add_documents(docs)

    # Test with custom search_kwargs
    retriever = vectorstore.as_retriever(search_kwargs={"k": 1})
    results = retriever.invoke("programming")

    # Should return exactly 1 document
    assert len(results) == 1
    assert isinstance(results[0], Document)