The application utilizes large language models to generate descriptions from images, allowing users to search a database of Indian recipes. The application stores text embeddings in PostgreSQL using pgvector, enabling efficient similarity searches. The application supports multiple LLM providers and allows users to choose between OpenAI or local Ollama models. The pgvector extension provides vector similarity search functionality, making it possible for the application to execute queries on the recipe description embeddings. A distributed SQL database like YugabyteDB can further enhance scalability and resilience for AI applications using pgvector.