Skip to content

Graph-Native Infrastructure for Context and Accountable AI Systems

Merge remote-tracking branch 'semantica-agi/main' into fix/843-vectorstore-persistent-backend-accessors

Sameer6305f75469f4720f
diff --cc tests/vector_store/test_decision_embedding_pipeline.py
index 90e6d494,4fb69280..17d47ddc
--- a/tests/vector_store/test_decision_embedding_pipeline.py
+++ b/tests/vector_store/test_decision_embedding_pipeline.py
@@@ -394,52 -393,95 +394,142 @@@ class TestDecisionEmbeddingPipelineEdge
assert all("decision_data" in result for result in results)
assert all("vector_id" in result for result in results)
+ def test_find_similar_decisions_real_faiss_backend(self):
+ """Regression test for FAISS backend crashing due to .vectors access (issue #839)"""
+ # Skip if faiss is not installed
+ pytest.importorskip("faiss")
+
+ try:
+ from semantica.vector_store import VectorStore
+ except ImportError:
+ pytest.skip("VectorStore not available")
+
+ # Create a real VectorStore with FAISS backend
+ vs = VectorStore(backend="faiss", config={"dimension": 384})
+ # Force fallback random embeddings of the exact dimension (384) to avoid sentence-transformers
+ # dependency or default 128-d mismatches.
+ vs.embedder = None
+ vs.initialize_decision_pipeline()
+
+ # Store a couple of dummy decisions to ensure index has data
+ vs.store_decision(scenario="A test decision 1", category="test")
+ vs.store_decision(scenario="A test decision 2", category="test")
+
+ # This will call _get_candidate_embeddings without a mock
+ # Before the fix, this crashed with AttributeError: 'VectorStore' object has no attribute 'vectors'
+ results = vs.search_decisions("test query", limit=2)
+
+ # Assert it returns results and doesn't crash
+ assert isinstance(results, list)
+ assert len(results) <= 2
+
+
+ def test_find_similar_decisions_real_inmemory_backend(self):
+ """Regression test for inmemory backend preserving behavior after issue #839 fix"""
+ try:
+ from semantica.vector_store import VectorStore
+ except ImportError:
+ pytest.skip("VectorStore not available")
+
+ vs = VectorStore(backend="inmemory", config={"dimension": 384})
+ vs.embedder = None
+ vs.initialize_decision_pipeline()
+
+ vs.store_decision(scenario="Apple product launch", category="tech")
+ vs.store_decision(scenario="Microsoft earnings report", category="tech")
+ vs.store_decision(scenario="Local bakery opens", category="food")
+
+ results = vs.search_decisions("software company", limit=2)
+
+ assert isinstance(results, list)
+ assert len(results) <= 2
+
+ if len(results) > 0:
+ assert "similarity" in results[0]
+ assert "semantic_similarity" in results[0]
+ assert "structural_similarity" in results[0]
+
+ def test_get_candidate_embeddings_returns_partial_matches_when_pool_exhausted(self):
+ """Regression test: _get_candidate_embeddings must not discard matches found
+ in its final retry iteration when the expand-and-retry loop exhausts max_k
+ without ever crossing `limit` matches or getting a short page back."""
+ pipeline = DecisionEmbeddingPipeline.__new__(DecisionEmbeddingPipeline)
+ pipeline.vector_store = Mock()
+ pipeline.embedding_dimension = 384
+ pipeline.node_embedding_dimension = 128
+
+ rare_indices = {10, 30, 50, 70}
+
+ def fake_search_vectors(query_vector, k=10, filter=None):
+ # Always returns a full page, so the backend never hits the
+ # "len(results) < current_k" stop condition.
+ return [
+ {
+ "id": f"v{i}",
+ "vector": np.random.rand(384),
+ "metadata": {"category": "rare" if i in rare_indices else "common"},
+ "score": 1.0 - i / 1000.0,
+ }
+ for i in range(k)
+ ]
+
+ pipeline.vector_store.search_vectors.side_effect = fake_search_vectors
+
+ result = pipeline._get_candidate_embeddings(
+ np.random.rand(384), limit=10, filters={"category": "rare"}
+ )
+
+ assert len(result["embeddings"]) == len(rare_indices)
+ assert len(result["metadata"]) == len(rare_indices)
+ assert len(result["scores"]) == len(rare_indices)
+
+
+class TestVectorStoreRetrieval:
+ """Test get_vector and get_metadata on real backends."""
+
+ def test_inmemory_retrieval(self):
+ """Test exact dict behavior for inmemory backend."""
+ vs = VectorStore(backend="inmemory")
+ vs.store_vectors([np.array([0.1, 0.2, 0.3], dtype=np.float32)], ids=["test1"], metadata=[{"foo": "bar"}])
+
+ vec = vs.get_vector("vec_0")
+ meta = vs.get_metadata("vec_0")
+
+ assert vec is not None
+ np.testing.assert_array_almost_equal(vec, np.array([0.1, 0.2, 0.3], dtype=np.float32))
+ assert meta == {"foo": "bar"}
+
+ def test_faiss_retrieval(self):
+ """Test reconstruction from FAISS."""
+ try:
+ import faiss
+ except ImportError:
+ pytest.skip("FAISS not installed")
+
+ vs = VectorStore(backend="faiss", config={"dimension": 3})
+ vs.store_vectors([np.array([0.1, 0.2, 0.3], dtype=np.float32)], ids=["test1"], metadata=[{"foo": "faiss_bar"}])
+
+ vec = vs.get_vector("test1")
+ assert vec is not None
+ np.testing.assert_array_almost_equal(vec, np.array([0.1, 0.2, 0.3], dtype=np.float32))
+
+ meta = vs.get_metadata("test1")
+ assert meta == {"foo": "faiss_bar"}
+
+ def test_cloud_backends_untested(self):
+ """
+ Note: The following backends are not tested locally as they require
+ live external services (Docker containers or API keys):
+ - QdrantStore
+ - PineconeStore
+ - MilvusStore
+ - WeaviateStore
+ - PgVectorStore
+
+ Their implementations rely directly on official client SDKs (e.g. client.retrieve,
+ index.fetch) to ensure correctness in production.
+ """
+ pass
+
if __name__ == "__main__":
pytest.main([__file__])