Embeddings & Semantic Search

Embeddings are powerful and dense numeric representations of data that capture the underlying meaning of words or concepts. In business applications, they enable more intelligent, context-aware search capabilities. By using Embedding Models, businesses can transform textual or other data into vector representations. These vectors are then stored in a vector database, which facilitates fast and efficient similarity searches using methods like cosine similarity. This allows for semantic search functionality, where results are based on meaning rather than exact keyword matches, improving search relevance, user experience, and overall operational efficiency.

Architecture

Solution Diagram Resources

You can download the Solution Diagram as a .drawio file for offline use. Alternatively, you may view and edit the Solution Diagram directly on draw.io.
Please note that any changes made online will need to be saved locally if you wish to keep them.

Download

Edit

In this architecture, the Cloud Application Programming (CAP) model serves as the central interface for managing application logic and executing searches. CAP natively supports embeddings as part of its data schema, allowing for seamless integration of vector representations. When user inputs are processed, they are converted into embeddings using Embedding Models via the Generative AI Hub. These vectors are then stored and indexed within SAP HANA Cloud's Vector Engine, enabling fast similarity searches through methods like cosine similarity. Upon initiating a search, the CAP model communicates with the Vector Engine to retrieve results based on the semantic meaning of the inputs, leading to contextually relevant responses. Additionally, various SDKs and plugins, such as the SAP Cloud SDK (for AI), CAP LLM Plugin and LangChain, enhance the embedding process and streamline integration with both, the Generative AI Hub and the Vector Engine.