Warning of embedding `input must have less than 8192 tokens` with discourse ai

how to configure embedding model? in discourse ?or model provider?