Some tests

tests/integration/test_nlp_query_integration.py

@@ -0,0 +1,540 @@
+"""
+Integration tests for NLP Query Service
+
+These tests verify the end-to-end functionality of the NLP query service,
+testing service coordination, prompt service integration, and schema processing.
+Following the TEST_STRATEGY.md approach for integration testing.
+"""
+
+import pytest
+import json
+from unittest.mock import AsyncMock, MagicMock, patch
+
+from trustgraph.schema import (
+    QuestionToStructuredQueryRequest, QuestionToStructuredQueryResponse,
+    PromptRequest, PromptResponse, Error, RowSchema, Field as SchemaField
+)
+from trustgraph.retrieval.nlp_query.service import Processor
+
+
+@pytest.mark.integration
+class TestNLPQueryServiceIntegration:
+    """Integration tests for NLP query service coordination"""
+
+    @pytest.fixture
+    def sample_schemas(self):
+        """Sample schemas for testing"""
+        return {
+            "customers": RowSchema(
+                name="customers",
+                description="Customer data with contact information",
+                fields=[
+                    SchemaField(name="id", type="string", primary=True),
+                    SchemaField(name="name", type="string"),
+                    SchemaField(name="email", type="string"),
+                    SchemaField(name="state", type="string"),
+                    SchemaField(name="phone", type="string")
+                ]
+            ),
+            "orders": RowSchema(
+                name="orders",
+                description="Customer order transactions",
+                fields=[
+                    SchemaField(name="order_id", type="string", primary=True),
+                    SchemaField(name="customer_id", type="string"),
+                    SchemaField(name="total", type="float"),
+                    SchemaField(name="status", type="string"),
+                    SchemaField(name="order_date", type="datetime")
+                ]
+            ),
+            "products": RowSchema(
+                name="products",
+                description="Product catalog information",
+                fields=[
+                    SchemaField(name="product_id", type="string", primary=True),
+                    SchemaField(name="name", type="string"),
+                    SchemaField(name="category", type="string"),
+                    SchemaField(name="price", type="float"),
+                    SchemaField(name="in_stock", type="boolean")
+                ]
+            )
+        }
+
+    @pytest.fixture
+    def integration_processor(self, sample_schemas):
+        """Create processor with realistic configuration"""
+        proc = Processor(
+            taskgroup=MagicMock(),
+            pulsar_client=AsyncMock(),
+            config_type="schema",
+            schema_selection_template="schema-selection-v1",
+            graphql_generation_template="graphql-generation-v1"
+        )
+        
+        # Set up schemas
+        proc.schemas = sample_schemas
+        
+        # Mock the client method
+        proc.client = MagicMock()
+        
+        return proc
+
+    @pytest.mark.asyncio
+    async def test_end_to_end_nlp_query_processing(self, integration_processor):
+        """Test complete NLP query processing pipeline"""
+        # Arrange - Create realistic query request
+        request = QuestionToStructuredQueryRequest(
+            question="Show me customers from California who have placed orders over $500",
+            max_results=50
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "integration-test-001"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock Phase 1 - Schema Selection Response
+        phase1_response = PromptResponse(
+            text=json.dumps(["customers", "orders"]),
+            error=None
+        )
+        
+        # Mock Phase 2 - GraphQL Generation Response  
+        expected_graphql = """
+        query GetCaliforniaCustomersWithLargeOrders($min_total: Float!) {
+          customers(where: {state: {eq: "California"}}) {
+            id
+            name
+            email
+            state
+            orders(where: {total: {gt: $min_total}}) {
+              order_id
+              total
+              status
+              order_date
+            }
+          }
+        }
+        """
+        
+        phase2_response = PromptResponse(
+            text=json.dumps({
+                "query": expected_graphql.strip(),
+                "variables": {"min_total": 500.0},
+                "confidence": 0.92
+            }),
+            error=None
+        )
+        
+        # Set up mock to return different responses for each call
+        integration_processor.client.return_value.request = AsyncMock(
+            side_effect=[phase1_response, phase2_response]
+        )
+        
+        # Act - Process the message
+        await integration_processor.on_message(msg, consumer, flow)
+        
+        # Assert - Verify the complete pipeline
+        assert integration_processor.client.return_value.request.call_count == 2
+        flow_response.send.assert_called_once()
+        
+        # Verify response structure and content
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert isinstance(response, QuestionToStructuredQueryResponse)
+        assert response.error is None
+        assert "customers" in response.graphql_query
+        assert "orders" in response.graphql_query
+        assert "California" in response.graphql_query
+        assert response.detected_schemas == ["customers", "orders"]
+        assert response.confidence == 0.92
+        assert response.variables["min_total"] == 500.0
+
+    @pytest.mark.asyncio
+    async def test_complex_multi_table_query_integration(self, integration_processor):
+        """Test integration with complex multi-table queries"""
+        # Arrange
+        request = QuestionToStructuredQueryRequest(
+            question="Find all electronic products under $100 that are in stock, along with any recent orders",
+            max_results=25
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "multi-table-test"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock responses
+        phase1_response = PromptResponse(
+            text=json.dumps(["products", "orders"]),
+            error=None
+        )
+        
+        phase2_response = PromptResponse(
+            text=json.dumps({
+                "query": "query { products(where: {category: {eq: \"Electronics\"}, price: {lt: 100}, in_stock: {eq: true}}) { product_id name price orders { order_id total } } }",
+                "variables": {},
+                "confidence": 0.88
+            }),
+            error=None
+        )
+        
+        integration_processor.client.return_value.request = AsyncMock(
+            side_effect=[phase1_response, phase2_response]
+        )
+        
+        # Act
+        await integration_processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert response.detected_schemas == ["products", "orders"]
+        assert "Electronics" in response.graphql_query
+        assert "price: {lt: 100}" in response.graphql_query
+        assert "in_stock: {eq: true}" in response.graphql_query
+
+    @pytest.mark.asyncio
+    async def test_schema_configuration_integration(self, integration_processor):
+        """Test integration with dynamic schema configuration"""
+        # Arrange - New schema configuration
+        new_schema_config = {
+            "schema": {
+                "inventory": json.dumps({
+                    "name": "inventory",
+                    "description": "Product inventory tracking",
+                    "fields": [
+                        {"name": "sku", "type": "string", "primary_key": True},
+                        {"name": "quantity", "type": "integer"},
+                        {"name": "warehouse_location", "type": "string"}
+                    ]
+                })
+            }
+        }
+        
+        # Act - Update configuration
+        await integration_processor.on_schema_config(new_schema_config, "v2")
+        
+        # Arrange - Test query using new schema
+        request = QuestionToStructuredQueryRequest(
+            question="Show inventory levels for all products in warehouse A",
+            max_results=100
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "schema-config-test"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock responses that use the new schema
+        phase1_response = PromptResponse(
+            text=json.dumps(["inventory"]),
+            error=None
+        )
+        
+        phase2_response = PromptResponse(
+            text=json.dumps({
+                "query": "query { inventory(where: {warehouse_location: {eq: \"A\"}}) { sku quantity warehouse_location } }",
+                "variables": {},
+                "confidence": 0.85
+            }),
+            error=None
+        )
+        
+        integration_processor.client.return_value.request = AsyncMock(
+            side_effect=[phase1_response, phase2_response]
+        )
+        
+        # Act
+        await integration_processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        assert "inventory" in integration_processor.schemas
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        assert response.detected_schemas == ["inventory"]
+        assert "inventory" in response.graphql_query
+
+    @pytest.mark.asyncio
+    async def test_prompt_service_error_recovery_integration(self, integration_processor):
+        """Test integration with prompt service error scenarios"""
+        # Arrange
+        request = QuestionToStructuredQueryRequest(
+            question="Show me customer data",
+            max_results=10
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "error-recovery-test"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock Phase 1 error
+        phase1_error_response = PromptResponse(
+            text="",
+            error=Error(type="template-not-found", message="Schema selection template not available")
+        )
+        
+        integration_processor.client.return_value.request = AsyncMock(
+            return_value=phase1_error_response
+        )
+        
+        # Act
+        await integration_processor.on_message(msg, consumer, flow)
+        
+        # Assert - Error is properly handled and propagated
+        flow_response.send.assert_called_once()
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert isinstance(response, QuestionToStructuredQueryResponse)
+        assert response.error is not None
+        assert response.error.type == "nlp-query-error"
+        assert "Prompt service error" in response.error.message
+
+    @pytest.mark.asyncio
+    async def test_template_parameter_integration(self, sample_schemas):
+        """Test integration with different template configurations"""
+        # Test with custom templates
+        custom_processor = Processor(
+            taskgroup=MagicMock(),
+            pulsar_client=AsyncMock(),
+            config_type="schema",
+            schema_selection_template="custom-schema-selector",
+            graphql_generation_template="custom-graphql-generator"
+        )
+        
+        custom_processor.schemas = sample_schemas
+        custom_processor.client = MagicMock()
+        
+        request = QuestionToStructuredQueryRequest(
+            question="Test query",
+            max_results=5
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "template-test"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock responses
+        phase1_response = PromptResponse(text=json.dumps(["customers"]), error=None)
+        phase2_response = PromptResponse(
+            text=json.dumps({
+                "query": "query { customers { id name } }",
+                "variables": {},
+                "confidence": 0.9
+            }),
+            error=None
+        )
+        
+        custom_processor.client.return_value.request = AsyncMock(
+            side_effect=[phase1_response, phase2_response]
+        )
+        
+        # Act
+        await custom_processor.on_message(msg, consumer, flow)
+        
+        # Assert - Verify custom templates are used
+        assert custom_processor.schema_selection_template == "custom-schema-selector"
+        assert custom_processor.graphql_generation_template == "custom-graphql-generator"
+        
+        # Verify the calls were made
+        assert custom_processor.client.return_value.request.call_count == 2
+
+    @pytest.mark.asyncio 
+    async def test_large_schema_set_integration(self, integration_processor):
+        """Test integration with large numbers of schemas"""
+        # Arrange - Add many schemas
+        large_schema_set = {}
+        for i in range(20):
+            schema_name = f"table_{i:02d}"
+            large_schema_set[schema_name] = RowSchema(
+                name=schema_name,
+                description=f"Test table {i} with sample data",
+                fields=[
+                    SchemaField(name="id", type="string", primary=True),
+                    SchemaField(name=f"field_{j}", type="string") for j in range(5)
+                ]
+            )
+        
+        integration_processor.schemas.update(large_schema_set)
+        
+        request = QuestionToStructuredQueryRequest(
+            question="Show me data from table_05 and table_12",
+            max_results=20
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "large-schema-test"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock responses
+        phase1_response = PromptResponse(
+            text=json.dumps(["table_05", "table_12"]),
+            error=None
+        )
+        
+        phase2_response = PromptResponse(
+            text=json.dumps({
+                "query": "query { table_05 { id field_0 } table_12 { id field_1 } }",
+                "variables": {},
+                "confidence": 0.87
+            }),
+            error=None
+        )
+        
+        integration_processor.client.return_value.request = AsyncMock(
+            side_effect=[phase1_response, phase2_response]
+        )
+        
+        # Act
+        await integration_processor.on_message(msg, consumer, flow)
+        
+        # Assert - Should handle large schema sets efficiently
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert response.detected_schemas == ["table_05", "table_12"]
+        assert "table_05" in response.graphql_query
+        assert "table_12" in response.graphql_query
+
+    @pytest.mark.asyncio
+    async def test_concurrent_request_handling_integration(self, integration_processor):
+        """Test integration with concurrent request processing"""
+        # Arrange - Multiple concurrent requests
+        requests = []
+        messages = []
+        flows = []
+        
+        for i in range(5):
+            request = QuestionToStructuredQueryRequest(
+                question=f"Query {i}: Show me data",
+                max_results=10
+            )
+            
+            msg = MagicMock()
+            msg.value.return_value = request
+            msg.properties.return_value = {"id": f"concurrent-test-{i}"}
+            
+            flow = MagicMock()
+            flow_response = MagicMock()
+            flow.return_value = flow_response
+            
+            requests.append(request)
+            messages.append(msg)
+            flows.append(flow)
+        
+        # Mock responses for all requests
+        mock_responses = []
+        for i in range(10):  # 2 calls per request (phase1 + phase2)
+            if i % 2 == 0:  # Phase 1 responses
+                mock_responses.append(PromptResponse(
+                    text=json.dumps(["customers"]),
+                    error=None
+                ))
+            else:  # Phase 2 responses
+                mock_responses.append(PromptResponse(
+                    text=json.dumps({
+                        "query": f"query {{ customers {{ id name }} }}",
+                        "variables": {},
+                        "confidence": 0.9
+                    }),
+                    error=None
+                ))
+        
+        integration_processor.client.return_value.request = AsyncMock(
+            side_effect=mock_responses
+        )
+        
+        # Act - Process all messages concurrently
+        import asyncio
+        consumer = MagicMock()
+        
+        tasks = []
+        for msg, flow in zip(messages, flows):
+            task = integration_processor.on_message(msg, consumer, flow)
+            tasks.append(task)
+        
+        await asyncio.gather(*tasks)
+        
+        # Assert - All requests should be processed
+        assert integration_processor.client.return_value.request.call_count == 10
+        for flow in flows:
+            flow.return_value.send.assert_called_once()
+
+    @pytest.mark.asyncio
+    async def test_performance_timing_integration(self, integration_processor):
+        """Test performance characteristics of the integration"""
+        # Arrange
+        request = QuestionToStructuredQueryRequest(
+            question="Performance test query",
+            max_results=100
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "performance-test"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock fast responses
+        phase1_response = PromptResponse(text=json.dumps(["customers"]), error=None)
+        phase2_response = PromptResponse(
+            text=json.dumps({
+                "query": "query { customers { id } }",
+                "variables": {},
+                "confidence": 0.9
+            }),
+            error=None
+        )
+        
+        integration_processor.client.return_value.request = AsyncMock(
+            side_effect=[phase1_response, phase2_response]
+        )
+        
+        # Act
+        import time
+        start_time = time.time()
+        
+        await integration_processor.on_message(msg, consumer, flow)
+        
+        end_time = time.time()
+        execution_time = end_time - start_time
+        
+        # Assert
+        assert execution_time < 1.0  # Should complete quickly with mocked services
+        flow_response.send.assert_called_once()
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        assert response.error is None

tests/unit/test_retrieval/test_nlp_query.py

@@ -0,0 +1,356 @@
+"""
+Unit tests for NLP Query service
+Following TEST_STRATEGY.md approach for service testing
+"""
+
+import pytest
+import json
+from unittest.mock import AsyncMock, MagicMock, patch
+from typing import Dict, Any
+
+from trustgraph.schema import (
+    QuestionToStructuredQueryRequest, QuestionToStructuredQueryResponse,
+    PromptRequest, PromptResponse, Error, RowSchema, Field as SchemaField
+)
+from trustgraph.retrieval.nlp_query.service import Processor
+
+
+@pytest.fixture
+def mock_prompt_client():
+    """Mock prompt service client"""
+    return AsyncMock()
+
+
+@pytest.fixture
+def mock_pulsar_client():
+    """Mock Pulsar client"""
+    return AsyncMock()
+
+
+@pytest.fixture
+def sample_schemas():
+    """Sample schemas for testing"""
+    return {
+        "customers": RowSchema(
+            name="customers",
+            description="Customer data",
+            fields=[
+                SchemaField(name="id", type="string", primary=True),
+                SchemaField(name="name", type="string"),
+                SchemaField(name="email", type="string"),
+                SchemaField(name="state", type="string")
+            ]
+        ),
+        "orders": RowSchema(
+            name="orders", 
+            description="Order data",
+            fields=[
+                SchemaField(name="order_id", type="string", primary=True),
+                SchemaField(name="customer_id", type="string"),
+                SchemaField(name="total", type="float"),
+                SchemaField(name="status", type="string")
+            ]
+        )
+    }
+
+
+@pytest.fixture
+def processor(mock_pulsar_client, sample_schemas):
+    """Create processor with mocked dependencies"""
+    proc = Processor(
+        taskgroup=MagicMock(),
+        pulsar_client=mock_pulsar_client,
+        config_type="schema"
+    )
+    
+    # Set up schemas
+    proc.schemas = sample_schemas
+    
+    # Mock the client method
+    proc.client = MagicMock()
+    
+    return proc
+
+
+@pytest.mark.asyncio
+class TestNLPQueryProcessor:
+    """Test NLP Query service processor"""
+
+    async def test_phase1_select_schemas_success(self, processor, mock_prompt_client):
+        """Test successful schema selection (Phase 1)"""
+        # Arrange
+        question = "Show me customers from California"
+        expected_schemas = ["customers"]
+        
+        mock_response = PromptResponse(
+            text=json.dumps(expected_schemas),
+            error=None
+        )
+        
+        processor.client.return_value.request = AsyncMock(return_value=mock_response)
+        
+        # Act
+        result = await processor.phase1_select_schemas(question)
+        
+        # Assert
+        assert result == expected_schemas
+        processor.client.assert_called_once_with("prompt-request")
+
+    async def test_phase1_select_schemas_prompt_error(self, processor):
+        """Test schema selection with prompt service error"""
+        # Arrange
+        question = "Show me customers"
+        error = Error(type="prompt-error", message="Template not found")
+        mock_response = PromptResponse(text="", error=error)
+        
+        processor.client.return_value.request = AsyncMock(return_value=mock_response)
+        
+        # Act & Assert
+        with pytest.raises(Exception, match="Prompt service error"):
+            await processor.phase1_select_schemas(question)
+
+    async def test_phase2_generate_graphql_success(self, processor):
+        """Test successful GraphQL generation (Phase 2)"""
+        # Arrange
+        question = "Show me customers from California"
+        selected_schemas = ["customers"]
+        expected_result = {
+            "query": "query { customers(where: {state: {eq: \"California\"}}) { id name email state } }",
+            "variables": {},
+            "confidence": 0.95
+        }
+        
+        mock_response = PromptResponse(
+            text=json.dumps(expected_result),
+            error=None
+        )
+        
+        processor.client.return_value.request = AsyncMock(return_value=mock_response)
+        
+        # Act
+        result = await processor.phase2_generate_graphql(question, selected_schemas)
+        
+        # Assert
+        assert result == expected_result
+        processor.client.assert_called_once_with("prompt-request")
+
+    async def test_phase2_generate_graphql_prompt_error(self, processor):
+        """Test GraphQL generation with prompt service error"""
+        # Arrange
+        question = "Show me customers"
+        selected_schemas = ["customers"]
+        error = Error(type="prompt-error", message="Generation failed")
+        mock_response = PromptResponse(text="", error=error)
+        
+        processor.client.return_value.request = AsyncMock(return_value=mock_response)
+        
+        # Act & Assert
+        with pytest.raises(Exception, match="Prompt service error"):
+            await processor.phase2_generate_graphql(question, selected_schemas)
+
+    async def test_on_message_full_flow_success(self, processor):
+        """Test complete message processing flow"""
+        # Arrange
+        request = QuestionToStructuredQueryRequest(
+            question="Show me customers from California", 
+            max_results=100
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-123"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock Phase 1 response
+        phase1_response = PromptResponse(
+            text=json.dumps(["customers"]),
+            error=None
+        )
+        
+        # Mock Phase 2 response
+        phase2_response = PromptResponse(
+            text=json.dumps({
+                "query": "query { customers(where: {state: {eq: \"California\"}}) { id name email } }",
+                "variables": {},
+                "confidence": 0.9
+            }),
+            error=None
+        )
+        
+        # Set up mock to return different responses for each call
+        processor.client.return_value.request = AsyncMock(
+            side_effect=[phase1_response, phase2_response]
+        )
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        assert processor.client.return_value.request.call_count == 2
+        flow_response.send.assert_called_once()
+        
+        # Verify response structure
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]  # First argument is the response object
+        
+        assert isinstance(response, QuestionToStructuredQueryResponse)
+        assert response.error is None
+        assert "customers" in response.graphql_query
+        assert response.detected_schemas == ["customers"]
+        assert response.confidence == 0.9
+
+    async def test_on_message_phase1_error(self, processor):
+        """Test message processing with Phase 1 failure"""
+        # Arrange
+        request = QuestionToStructuredQueryRequest(
+            question="Show me customers", 
+            max_results=100
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-123"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = MagicMock()
+        flow.return_value = flow_response
+        
+        # Mock Phase 1 error
+        phase1_response = PromptResponse(
+            text="",
+            error=Error(type="template-error", message="Template not found")
+        )
+        
+        processor.client.return_value.request = AsyncMock(return_value=phase1_response)
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        flow_response.send.assert_called_once()
+        
+        # Verify error response
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert isinstance(response, QuestionToStructuredQueryResponse)
+        assert response.error is not None
+        assert response.error.type == "nlp-query-error"
+        assert "Prompt service error" in response.error.message
+
+    async def test_schema_config_loading(self, processor):
+        """Test schema configuration loading"""
+        # Arrange
+        config = {
+            "schema": {
+                "test_schema": json.dumps({
+                    "name": "test_schema",
+                    "description": "Test schema", 
+                    "fields": [
+                        {
+                            "name": "id",
+                            "type": "string",
+                            "primary_key": True,
+                            "required": True
+                        },
+                        {
+                            "name": "name", 
+                            "type": "string",
+                            "description": "User name"
+                        }
+                    ]
+                })
+            }
+        }
+        
+        # Act
+        await processor.on_schema_config(config, "v1")
+        
+        # Assert
+        assert "test_schema" in processor.schemas
+        schema = processor.schemas["test_schema"]
+        assert schema.name == "test_schema"
+        assert schema.description == "Test schema"
+        assert len(schema.fields) == 2
+        assert schema.fields[0].name == "id"
+        assert schema.fields[0].primary == True
+        assert schema.fields[1].name == "name"
+
+    async def test_schema_config_loading_invalid_json(self, processor):
+        """Test schema configuration loading with invalid JSON"""
+        # Arrange
+        config = {
+            "schema": {
+                "bad_schema": "invalid json{"
+            }
+        }
+        
+        # Act
+        await processor.on_schema_config(config, "v1")
+        
+        # Assert - bad schema should be ignored
+        assert "bad_schema" not in processor.schemas
+
+    def test_processor_initialization(self, mock_pulsar_client):
+        """Test processor initialization with correct specifications"""
+        # Act
+        processor = Processor(
+            taskgroup=MagicMock(),
+            pulsar_client=mock_pulsar_client,
+            schema_selection_template="custom-schema-select",
+            graphql_generation_template="custom-graphql-gen"
+        )
+        
+        # Assert
+        assert processor.schema_selection_template == "custom-schema-select"
+        assert processor.graphql_generation_template == "custom-graphql-gen"
+        assert processor.config_key == "schema"
+        assert processor.schemas == {}
+
+    def test_add_args(self):
+        """Test command-line argument parsing"""
+        import argparse
+        
+        parser = argparse.ArgumentParser()
+        Processor.add_args(parser)
+        
+        # Test default values
+        args = parser.parse_args([])
+        assert args.config_type == "schema"
+        assert args.schema_selection_template == "schema-selection"
+        assert args.graphql_generation_template == "graphql-generation"
+        
+        # Test custom values
+        args = parser.parse_args([
+            "--config-type", "custom",
+            "--schema-selection-template", "my-selector",
+            "--graphql-generation-template", "my-generator"
+        ])
+        assert args.config_type == "custom"
+        assert args.schema_selection_template == "my-selector"
+        assert args.graphql_generation_template == "my-generator"
+
+
+@pytest.mark.unit
+class TestNLPQueryHelperFunctions:
+    """Test helper functions and data transformations"""
+    
+    def test_schema_info_formatting(self, sample_schemas):
+        """Test schema info formatting for prompts"""
+        # This would test any helper functions for formatting schema data
+        # Currently the formatting is inline, but good to test if extracted
+        
+        customers_schema = sample_schemas["customers"]
+        expected_fields = ["id", "name", "email", "state"]
+        
+        actual_fields = [f.name for f in customers_schema.fields]
+        assert actual_fields == expected_fields
+        
+        # Test primary key detection
+        primary_fields = [f.name for f in customers_schema.fields if f.primary]
+        assert primary_fields == ["id"]

tests/unit/test_retrieval/test_structured_query.py

@@ -0,0 +1,517 @@
+"""
+Unit tests for Structured Query Service
+Following TEST_STRATEGY.md approach for service testing
+"""
+
+import pytest
+import json
+from unittest.mock import AsyncMock, MagicMock, patch
+
+from trustgraph.schema import (
+    StructuredQueryRequest, StructuredQueryResponse,
+    QuestionToStructuredQueryRequest, QuestionToStructuredQueryResponse,
+    ObjectsQueryRequest, ObjectsQueryResponse,
+    Error, GraphQLError
+)
+from trustgraph.retrieval.structured_query.service import Processor
+
+
+@pytest.fixture
+def mock_pulsar_client():
+    """Mock Pulsar client"""
+    return AsyncMock()
+
+
+@pytest.fixture
+def processor(mock_pulsar_client):
+    """Create processor with mocked dependencies"""
+    proc = Processor(
+        taskgroup=MagicMock(),
+        pulsar_client=mock_pulsar_client
+    )
+    
+    # Mock the client method
+    proc.client = MagicMock()
+    
+    return proc
+
+
+@pytest.mark.asyncio
+class TestStructuredQueryProcessor:
+    """Test Structured Query service processor"""
+
+    async def test_successful_end_to_end_query(self, processor):
+        """Test successful end-to-end query processing"""
+        # Arrange
+        request = StructuredQueryRequest(
+            question="Show me all customers from New York"
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-123"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = AsyncMock()
+        flow.return_value = flow_response
+        
+        # Mock NLP query service response
+        nlp_response = QuestionToStructuredQueryResponse(
+            error=None,
+            graphql_query='query { customers(where: {state: {eq: "NY"}}) { id name email } }',
+            variables={"state": "NY"},
+            detected_schemas=["customers"],
+            confidence=0.95
+        )
+        
+        # Mock objects query service response
+        objects_response = ObjectsQueryResponse(
+            error=None,
+            data='{"customers": [{"id": "1", "name": "John", "email": "john@example.com"}]}',
+            errors=None
+        )
+        
+        # Set up mock clients
+        mock_nlp_client = AsyncMock()
+        mock_nlp_client.request.return_value = nlp_response
+        
+        mock_objects_client = AsyncMock()
+        mock_objects_client.request.return_value = objects_response
+        
+        processor.client.side_effect = lambda name: (
+            mock_nlp_client if name == "nlp-query-request" else mock_objects_client
+        )
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        # Verify NLP query service was called correctly
+        mock_nlp_client.request.assert_called_once()
+        nlp_call_args = mock_nlp_client.request.call_args[0][0]
+        assert isinstance(nlp_call_args, QuestionToStructuredQueryRequest)
+        assert nlp_call_args.question == "Show me all customers from New York"
+        assert nlp_call_args.max_results == 100
+        
+        # Verify objects query service was called correctly
+        mock_objects_client.request.assert_called_once()
+        objects_call_args = mock_objects_client.request.call_args[0][0]
+        assert isinstance(objects_call_args, ObjectsQueryRequest)
+        assert objects_call_args.query == 'query { customers(where: {state: {eq: "NY"}}) { id name email } }'
+        assert objects_call_args.variables == {"state": "NY"}
+        assert objects_call_args.user == "default"
+        assert objects_call_args.collection == "default"
+        
+        # Verify response
+        flow_response.send.assert_called_once()
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert isinstance(response, StructuredQueryResponse)
+        assert response.error is None
+        assert response.data == '{"customers": [{"id": "1", "name": "John", "email": "john@example.com"}]}'
+        assert len(response.errors) == 0
+
+    async def test_nlp_query_service_error(self, processor):
+        """Test handling of NLP query service errors"""
+        # Arrange
+        request = StructuredQueryRequest(
+            question="Invalid query"
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-error"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = AsyncMock()
+        flow.return_value = flow_response
+        
+        # Mock NLP query service error response
+        nlp_response = QuestionToStructuredQueryResponse(
+            error=Error(type="nlp-query-error", message="Failed to parse question"),
+            graphql_query="",
+            variables={},
+            detected_schemas=[],
+            confidence=0.0
+        )
+        
+        mock_nlp_client = AsyncMock()
+        mock_nlp_client.request.return_value = nlp_response
+        
+        processor.client.return_value = mock_nlp_client
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        flow_response.send.assert_called_once()
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert isinstance(response, StructuredQueryResponse)
+        assert response.error is not None
+        assert response.error.type == "structured-query-error"
+        assert "NLP query service error" in response.error.message
+
+    async def test_empty_graphql_query_error(self, processor):
+        """Test handling of empty GraphQL query from NLP service"""
+        # Arrange
+        request = StructuredQueryRequest(
+            question="Ambiguous question"
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-empty"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = AsyncMock()
+        flow.return_value = flow_response
+        
+        # Mock NLP query service response with empty query
+        nlp_response = QuestionToStructuredQueryResponse(
+            error=None,
+            graphql_query="",  # Empty query
+            variables={},
+            detected_schemas=[],
+            confidence=0.1
+        )
+        
+        mock_nlp_client = AsyncMock()
+        mock_nlp_client.request.return_value = nlp_response
+        
+        processor.client.return_value = mock_nlp_client
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        flow_response.send.assert_called_once()
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert response.error is not None
+        assert "empty GraphQL query" in response.error.message
+
+    async def test_objects_query_service_error(self, processor):
+        """Test handling of objects query service errors"""
+        # Arrange
+        request = StructuredQueryRequest(
+            question="Show me customers"
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-objects-error"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = AsyncMock()
+        flow.return_value = flow_response
+        
+        # Mock successful NLP response
+        nlp_response = QuestionToStructuredQueryResponse(
+            error=None,
+            graphql_query='query { customers { id name } }',
+            variables={},
+            detected_schemas=["customers"],
+            confidence=0.9
+        )
+        
+        # Mock objects query service error
+        objects_response = ObjectsQueryResponse(
+            error=Error(type="graphql-execution-error", message="Table 'customers' not found"),
+            data=None,
+            errors=None
+        )
+        
+        mock_nlp_client = AsyncMock()
+        mock_nlp_client.request.return_value = nlp_response
+        
+        mock_objects_client = AsyncMock()
+        mock_objects_client.request.return_value = objects_response
+        
+        processor.client.side_effect = lambda name: (
+            mock_nlp_client if name == "nlp-query-request" else mock_objects_client
+        )
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        flow_response.send.assert_called_once()
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert response.error is not None
+        assert "Objects query service error" in response.error.message
+        assert "Table 'customers' not found" in response.error.message
+
+    async def test_graphql_errors_handling(self, processor):
+        """Test handling of GraphQL validation/execution errors"""
+        # Arrange
+        request = StructuredQueryRequest(
+            question="Show invalid field"
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-graphql-errors"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = AsyncMock()
+        flow.return_value = flow_response
+        
+        # Mock successful NLP response
+        nlp_response = QuestionToStructuredQueryResponse(
+            error=None,
+            graphql_query='query { customers { invalid_field } }',
+            variables={},
+            detected_schemas=["customers"],
+            confidence=0.8
+        )
+        
+        # Mock objects response with GraphQL errors
+        graphql_errors = [
+            GraphQLError(
+                message="Cannot query field 'invalid_field' on type 'Customer'",
+                path=["customers", 0, "invalid_field"]
+            )
+        ]
+        
+        objects_response = ObjectsQueryResponse(
+            error=None,
+            data=None,
+            errors=graphql_errors
+        )
+        
+        mock_nlp_client = AsyncMock()
+        mock_nlp_client.request.return_value = nlp_response
+        
+        mock_objects_client = AsyncMock()
+        mock_objects_client.request.return_value = objects_response
+        
+        processor.client.side_effect = lambda name: (
+            mock_nlp_client if name == "nlp-query-request" else mock_objects_client
+        )
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        flow_response.send.assert_called_once()
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert response.error is None
+        assert len(response.errors) == 1
+        assert "Cannot query field 'invalid_field'" in response.errors[0]
+        assert "customers" in response.errors[0]
+
+    async def test_complex_query_with_variables(self, processor):
+        """Test processing complex queries with variables"""
+        # Arrange
+        request = StructuredQueryRequest(
+            question="Show customers with orders over $100 from last month"
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-complex"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = AsyncMock()
+        flow.return_value = flow_response
+        
+        # Mock NLP response with complex query and variables
+        nlp_response = QuestionToStructuredQueryResponse(
+            error=None,
+            graphql_query='''
+            query GetCustomersWithLargeOrders($minTotal: Float!, $startDate: String!) {
+                customers {
+                    id
+                    name
+                    orders(where: {total: {gt: $minTotal}, date: {gte: $startDate}}) {
+                        id
+                        total
+                        date
+                    }
+                }
+            }
+            ''',
+            variables={
+                "minTotal": 100.0,
+                "startDate": "2024-01-01"
+            },
+            detected_schemas=["customers", "orders"],
+            confidence=0.88
+        )
+        
+        # Mock objects response
+        objects_response = ObjectsQueryResponse(
+            error=None,
+            data='{"customers": [{"id": "1", "name": "Alice", "orders": [{"id": "100", "total": 150.0}]}]}',
+            errors=None
+        )
+        
+        mock_nlp_client = AsyncMock()
+        mock_nlp_client.request.return_value = nlp_response
+        
+        mock_objects_client = AsyncMock()
+        mock_objects_client.request.return_value = objects_response
+        
+        processor.client.side_effect = lambda name: (
+            mock_nlp_client if name == "nlp-query-request" else mock_objects_client
+        )
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        # Verify variables were passed correctly
+        objects_call_args = mock_objects_client.request.call_args[0][0]
+        assert objects_call_args.variables["minTotal"] == 100.0
+        assert objects_call_args.variables["startDate"] == "2024-01-01"
+        
+        # Verify response
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        assert response.error is None
+        assert "Alice" in response.data
+
+    async def test_null_data_handling(self, processor):
+        """Test handling of null/empty data responses"""
+        # Arrange
+        request = StructuredQueryRequest(
+            question="Show nonexistent data"
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-null"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = AsyncMock()
+        flow.return_value = flow_response
+        
+        # Mock responses
+        nlp_response = QuestionToStructuredQueryResponse(
+            error=None,
+            graphql_query='query { customers { id } }',
+            variables={},
+            detected_schemas=["customers"],
+            confidence=0.9
+        )
+        
+        objects_response = ObjectsQueryResponse(
+            error=None,
+            data=None,  # Null data
+            errors=None
+        )
+        
+        mock_nlp_client = AsyncMock()
+        mock_nlp_client.request.return_value = nlp_response
+        
+        mock_objects_client = AsyncMock()
+        mock_objects_client.request.return_value = objects_response
+        
+        processor.client.side_effect = lambda name: (
+            mock_nlp_client if name == "nlp-query-request" else mock_objects_client
+        )
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert response.error is None
+        assert response.data == "null"  # Should convert None to "null" string
+
+    async def test_exception_handling(self, processor):
+        """Test general exception handling"""
+        # Arrange
+        request = StructuredQueryRequest(
+            question="Test exception"
+        )
+        
+        msg = MagicMock()
+        msg.value.return_value = request
+        msg.properties.return_value = {"id": "test-exception"}
+        
+        consumer = MagicMock()
+        flow = MagicMock()
+        flow_response = AsyncMock()
+        flow.return_value = flow_response
+        
+        # Mock client to raise exception
+        mock_client = AsyncMock()
+        mock_client.request.side_effect = Exception("Network timeout")
+        processor.client.return_value = mock_client
+        
+        # Act
+        await processor.on_message(msg, consumer, flow)
+        
+        # Assert
+        flow_response.send.assert_called_once()
+        response_call = flow_response.send.call_args
+        response = response_call[0][0]
+        
+        assert response.error is not None
+        assert response.error.type == "structured-query-error"
+        assert "Network timeout" in response.error.message
+        assert response.data == "null"
+        assert len(response.errors) == 0
+
+    def test_processor_initialization(self, mock_pulsar_client):
+        """Test processor initialization with correct specifications"""
+        # Act
+        processor = Processor(
+            taskgroup=MagicMock(),
+            pulsar_client=mock_pulsar_client
+        )
+        
+        # Assert - Test default ID
+        assert processor.id == "structured-query"
+        
+        # Verify specifications were registered (we can't directly access them,
+        # but we know they were registered if initialization succeeded)
+        assert processor is not None
+
+    def test_add_args(self):
+        """Test command-line argument parsing"""
+        import argparse
+        
+        parser = argparse.ArgumentParser()
+        Processor.add_args(parser)
+        
+        # Test that it doesn't crash (no additional args)
+        args = parser.parse_args([])
+        # No specific assertions since no custom args are added
+        assert args is not None
+
+
+@pytest.mark.unit
+class TestStructuredQueryHelperFunctions:
+    """Test helper functions and data transformations"""
+    
+    def test_service_logging_integration(self):
+        """Test that logging is properly configured"""
+        # Import the logger
+        from trustgraph.retrieval.structured_query.service import logger
+        
+        assert logger.name == "trustgraph.retrieval.structured_query.service"
+        
+    def test_default_values(self):
+        """Test default configuration values"""
+        from trustgraph.retrieval.structured_query.service import default_ident
+        
+        assert default_ident == "structured-query"

trustgraph-flow/trustgraph/retrieval/structured_query/service.py

@@ -13,7 +13,7 @@ from ...schema import QuestionToStructuredQueryRequest, QuestionToStructuredQuer
 from ...schema import ObjectsQueryRequest, ObjectsQueryResponse
 from ...schema import Error
 
-from ...base import FlowProcessor, ConsumerSpec, ProducerSpec, ClientSpec
+from ...base import FlowProcessor, ConsumerSpec, ProducerSpec, RequestResponseSpec
 
 # Module logger
 logger = logging.getLogger(__name__)
@@ -49,7 +49,7 @@ class Processor(FlowProcessor):
         
         # Client spec for calling NLP query service
         self.register_specification(
-            ClientSpec(
+            RequestResponseSpec(
                 request_name = "nlp-query-request",
                 response_name = "nlp-query-response", 
                 request_schema = QuestionToStructuredQueryRequest,
@@ -59,7 +59,7 @@ class Processor(FlowProcessor):
         
         # Client spec for calling objects query service
         self.register_specification(
-            ClientSpec(
+            RequestResponseSpec(
                 request_name = "objects-query-request",
                 response_name = "objects-query-response",
                 request_schema = ObjectsQueryRequest,