AI Search: the search primitive for your agents

Engineers at Cloudflare introduced AI Search, a plug-and-play search primitive that enables hybrid search, combining semantic and keyword matching in a single query. This eliminates the need for separate indices and fusion logic, and allows for dynamically created instances with built-in storage and index. With AI Search, developers can attach metadata to documents and boost rankings at query time, as well as query across multiple instances in a single call. To demonstrate AI Search, a customer support agent was created that searches for relevant product documentation and the customer's past resolutions, retrieving relevant information and saving a searchable summary of the issue and resolution. The agent uses the AI Search namespace bindings and Workers to dynamically create instances, giving each customer their own searchable context and history. The resulting system enables efficient and accurate knowledge retrieval and resolution tracking, reducing the time and effort required for customer support.