This Symfony application enables natural language product search through AI-powered semantic understanding. It imports product data from XML, vectorizes product attributes using OpenAI embeddings, and stores them in Milvus vector database for efficient similarity search.

• Import electronic products from XML files
• Vectorize product properties with OpenAI Embeddings
• Store and search products in Milvus vector database
• Natural language product search via API
• Web interface with DeepChat for intuitive user interaction
• Flexible configuration for API keys and endpoints

• Symfony 6.4: Core application framework
• PHP 8.2+: Required runtime
• OpenAI API: For embeddings and chat completions
• Milvus: Vector database for similarity search
• DDEV: Local development environment
• Gitpod: Cloud development environment

The application follows a service-oriented architecture with key components organized into controllers, services, and entities.

1. Controllers:

   • ProductFinderController: Handles product search API endpoints
   • WebInterfaceController: Manages the web interface

2. Services:

   • XmlImportService: Parses XML files and extracts product data
   • OpenAIEmbeddingGenerator: Generates vector embeddings using OpenAI
   • MilvusVectorStoreService: Manages vector database interactions
   • OpenAISearchService: Generates natural language recommendations
   • PromptService: Manages prompts for OpenAI chat models

3. Entities:

   • Product: Represents products with properties, features, and specifications

1. User submits natural language query
2. Query is vectorized using OpenAI embeddings
3. Vector search finds similar products in Milvus
4. Results are filtered by relevance threshold (distance ≤ 0.5)
5. OpenAI generates natural language recommendations based on results
6. User receives product recommendations and matching products

For detailed architecture diagrams, see the Architecture Documentation.

• PHP 8.2 or higher
• Composer
• DDEV (recommended for local development)

1. Clone the repository:

   git clone git@github.com:iGore/symfony-product-finder.git
   cd symfony-product-finder
   

2. Install dependencies:

   ddev composer install
   

3. Configure environment variables in .env.local:

   OPENAI_API_KEY=your_openai_api_key
   OPENAI_EMBEDDING_MODEL=text-embedding-3-small
   OPENAI_CHAT_MODEL=gpt-3.5-turbo
   MILVUS_API_KEY=your_milvus_api_key
   MILVUS_HOST=your_milvus_endpoint
   MILVUS_PORT=443
   MILVUS_COLLECTION=products
   

4. Start the application:

   ddev start
   

For quick development without local setup, use Gitpod:

Gitpod provides a ready-to-use environment with DDEV pre-configured. The application is automatically started and accessible via the URL provided by Gitpod.

Import sample products from XML:

ddev php bin/console app:import-products src/DataFixtures/xml/sample_products.xml

The import process vectorizes both complete products and their individual features/specifications, enabling precise semantic matching.

Try the natural language search:

ddev php bin/console app:test-search "I need a waterproof smartphone with a good camera"

Access the chat interface at https://symfony-product-finder.ddev.site/ to search for products using natural language.

• Custom XML Format: Modify XmlImportService.php to support different XML structures
• Alternative Embedding Providers: Implement EmbeddingGeneratorInterface to use different vector providers
• Vector Database Configuration: Customize MilvusVectorStoreService.php for specific vector storage needs

Run the test suite:

ddev php bin/phpunit

• Controllers: src/Controller/ - API endpoints and web interface
• Services: src/Service/ - Business logic and integrations
• Entities: src/Entity/ - Data models
• DTOs: src/DTO/ - Data transfer objects for API requests/responses

This project is licensed under the MIT License.