chore: Updating documentation and adding exception handling for Vector Stores in RAG Tool and updating inference to use openai and updating memory implementation to use existing libraries

Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>

tests/unit/providers/utils/memory/test_vector_store.py

@@ -178,3 +178,41 @@ def test_content_from_data_and_mime_type_both_encodings_fail():
         # Should raise an exception instead of returning empty string
         with pytest.raises(UnicodeDecodeError):
             content_from_data_and_mime_type(data, mime_type)
+
+
+async def test_memory_tool_error_handling():
+    """Test that memory tool handles various failures gracefully without crashing."""
+    from llama_stack.providers.inline.tool_runtime.rag.config import RagToolRuntimeConfig
+    from llama_stack.providers.inline.tool_runtime.rag.memory import MemoryToolRuntimeImpl
+
+    config = RagToolRuntimeConfig()
+    memory_tool = MemoryToolRuntimeImpl(
+        config=config,
+        vector_io_api=AsyncMock(),
+        inference_api=AsyncMock(),
+        files_api=AsyncMock(),
+    )
+
+    docs = [
+        RAGDocument(document_id="good_doc", content="Good content", metadata={}),
+        RAGDocument(document_id="bad_url_doc", content=URL(uri="https://bad.url"), metadata={}),
+        RAGDocument(document_id="another_good_doc", content="Another good content", metadata={}),
+    ]
+
+    mock_file1 = MagicMock()
+    mock_file1.id = "file_good1"
+    mock_file2 = MagicMock()
+    mock_file2.id = "file_good2"
+    memory_tool.files_api.openai_upload_file.side_effect = [mock_file1, mock_file2]
+
+    with patch("httpx.AsyncClient") as mock_client:
+        mock_instance = AsyncMock()
+        mock_instance.get.side_effect = Exception("Bad URL")
+        mock_client.return_value.__aenter__.return_value = mock_instance
+
+        # won't raise exception despite one document failing
+        await memory_tool.insert(docs, "vector_store_123")
+
+    # processed 2 documents successfully, skipped 1
+    assert memory_tool.files_api.openai_upload_file.call_count == 2
+    assert memory_tool.vector_io_api.openai_attach_file_to_vector_store.call_count == 2

tests/unit/rag/test_rag_query.py

@@ -81,3 +81,58 @@ class TestRagQuery:
         # Test that invalid mode raises an error
         with pytest.raises(ValueError):
             RAGQueryConfig(mode="wrong_mode")
+
+    async def test_query_adds_vector_db_id_to_chunk_metadata(self):
+        rag_tool = MemoryToolRuntimeImpl(
+            config=MagicMock(),
+            vector_io_api=MagicMock(),
+            inference_api=MagicMock(),
+            files_api=MagicMock(),
+        )
+
+        vector_db_ids = ["db1", "db2"]
+
+        # Fake chunks from each DB
+        chunk_metadata1 = ChunkMetadata(
+            document_id="doc1",
+            chunk_id="chunk1",
+            source="test_source1",
+            metadata_token_count=5,
+        )
+        chunk1 = Chunk(
+            content="chunk from db1",
+            metadata={"vector_db_id": "db1", "document_id": "doc1"},
+            stored_chunk_id="c1",
+            chunk_metadata=chunk_metadata1,
+        )
+
+        chunk_metadata2 = ChunkMetadata(
+            document_id="doc2",
+            chunk_id="chunk2",
+            source="test_source2",
+            metadata_token_count=5,
+        )
+        chunk2 = Chunk(
+            content="chunk from db2",
+            metadata={"vector_db_id": "db2", "document_id": "doc2"},
+            stored_chunk_id="c2",
+            chunk_metadata=chunk_metadata2,
+        )
+
+        rag_tool.vector_io_api.query_chunks = AsyncMock(
+            side_effect=[
+                QueryChunksResponse(chunks=[chunk1], scores=[0.9]),
+                QueryChunksResponse(chunks=[chunk2], scores=[0.8]),
+            ]
+        )
+
+        result = await rag_tool.query(content="test", vector_db_ids=vector_db_ids)
+        returned_chunks = result.metadata["chunks"]
+        returned_scores = result.metadata["scores"]
+        returned_doc_ids = result.metadata["document_ids"]
+        returned_vector_db_ids = result.metadata["vector_db_ids"]
+
+        assert returned_chunks == ["chunk from db1", "chunk from db2"]
+        assert returned_scores == (0.9, 0.8)
+        assert returned_doc_ids == ["doc1", "doc2"]
+        assert returned_vector_db_ids == ["db1", "db2"]