dependency_injection

Dependency Injection System

This document describes the dependency injection (DI) system implemented in the MultiModal RAG project using the dependency-injector library.

Overview

The dependency injection container provides centralized configuration and dependency management for the entire application, making it easier to:

• Configure services through environment variables or configuration files
• Mock dependencies for testing
• Manage singleton instances of expensive resources
• Maintain clean separation of concerns

Usage Examples

TestContainer

A separate test container provides mocked dependencies for testing:

from retail_shelf_monitoring.containers import TestContainer

test_container = TestContainer()
# All dependencies are mocked for isolated testing

Basic Usage with Injection

from dependency_injector.wiring import Provide, inject
from retail_shelf_monitoring.containers import ApplicationContainer
from retail_shelf_monitoring.usecases.document_indexing import DocumentIndexingUseCase

@inject
async def index_document(
    indexing_use_case: DocumentIndexingUseCase = Provide[ApplicationContainer.document_indexing_use_case],
    document: DoclingDocument,
    document_id: str
) -> None:
    result = await indexing_use_case.index_document(document, document_id)
    print(f"Indexed: {result.success}")

# Setup container and wire
container = ApplicationContainer()
container.wire(modules=[__name__])

# Function automatically gets dependencies injected
await index_document(my_document, "doc_123")

Testing with Mocks

import unittest.mock as mock
from retail_shelf_monitoring.containers import ApplicationContainer

def test_document_indexing():
    container = ApplicationContainer()

    # Mock the repository
    mock_repository = mock.AsyncMock()
    mock_repository.index_document.return_value = IndexDocumentResponse(
        document_id="test", success=True, message="Success"
    )

    # Override dependency
    with container.document_repository.override(mock_repository):
        container.wire(modules=[__name__])

        # Test your code - mock will be injected automatically
        indexing_use_case = container.document_indexing_use_case()
        # ... test logic

Common Provider Types

1. Factory: Creates new instance each time
2. Singleton: Creates one instance, reuses it
3. Resource: Manages lifecycle of expensive objects
4. Configuration: Provides configuration values
5. Callable: Wraps functions/callables

The key insight is that Resource is specifically designed for objects that need both creation AND cleanup, making it perfect for database connections, file handles, network clients, etc.