The chunkingConfig property on datasources (of type UnstructuredChunkingConfig) has a new file_type property that can be either 'txt' or 'markdown'. This new property should only be present for connectors, file datasources don't need the file_type property (its optional). Also for connectors (not file uploads), the entire chunkingConfig is optional. If present, that means the embeddingField will be chunked and produce multiple rows, if it's not there it will just embed the embeddingField as-is.
see struct/datasource on branch 512-chunk-synced-rows of agentcloud to see the updated types