Chat

August 26, 2026 ยท View on GitHub

Conversational search request handled by Manticore Buddy via CALL CHAT. When this object is set, the /search endpoint answers through an LLM using KNN-retrieved rows as context instead of returning regular search hits. Required fields are query, table, and model_name. Chat model management (CREATE CHAT MODEL, etc.) remains SQL-only. For more information see Conversational search

Properties

NameTypeDescriptionNotes
queryStringUser question to send to the chat model
tableStringVectorized table to retrieve context from
modelNameStringName of the chat model
conversationUuidStringExisting conversation id to continue the dialog, or an empty string to start a new conversation. If omitted, a new id is generated.[optional]
vectorFieldStringA specific vector field to search by. If omitted, Buddy uses the first `FLOAT_VECTOR` field from `SHOW CREATE TABLE`.[optional]
fieldsStringLegacy alias for `vector_field`. A request must not include both `vector_field` and `fields`.[optional]