Generating embeddings inside SurrealQL with a custom function

SurrealDB introduces an innovative approach to semantic search by integrating embedding generation directly within its SurrealQL language, eliminating the need for separate vectorization services or multiple network hops. By defining a custom function, fn::embed, users can embed text into vectors directly within the database, which streamlines the "text in, ranked results out" workflow by keeping the embedding logic centralized in SurrealQL. This method simplifies error handling, reduces latency, and enhances efficiency, as the embedding, vector indexing, and graph traversal occur server-side, allowing for seamless semantic search operations. The text emphasizes best practices such as managing API keys securely as database parameters and optimizing performance through capabilities like HNSW indexing and cosine distance metrics. SurrealDB's approach facilitates embedding at both write and query times, enabling enriched and ranked search results without requiring external microservices or additional coding layers, thus offering a robust solution for implementing efficient and scalable semantic search on user data.