Vector search

Search by meaning and context using vector search operations.

Vector search operations allow you to compare Embeddings to find content by similarity rather than by exact matches. E.g., to find text by meaning or image by context.

You can search over embeddings that were generated by RavenDB ongoing embeddings-generation tasks or by an external embeddings provider.
You can also generate the embeddings for your documents on-the-fly, while searching.
When you run a vector search, your search query is converted into an embedding as well, and compared against document embeddings using either a dynamic query for ad-hoc or infrequent searches, or a static index for optimized performance.
Vector search can be used by other RavenDB AI features. E.g., AI agents can use vector search to retrieve relevant data requested by the LLM.

Use cases

Vector search can help wherever you need to find similar items based on proximity rather than exact matches, e.g. -

Knowledge and document search
Find relevant documentation, policies, legal texts, or enterprise reports using natural language queries.
Product and content recommendations
Suggest similar products, articles, videos, or media based on descriptive queries and user preferences.
Customer support automation
Route questions to the best help articles, retrieve guides, and power chatbot responses with relevant information.
Business intelligence and analysis
Profile customers and uncover market trends by comparing behavioral and relationship-based similarities.
Media and content analysis
Discover similar images, moderate content, and monitor social media for brand mentions and sentiment.

Technical documentation

Learn about vector search operations, how they use embeddings to find content by meaning or context, their ability to generate embeddings on the fly during searches, and other key aspects of this feature.