This method starts a job to update the indices that are used for serving LLM queries. Once the created job is complete, the LLM query endpoint will use the blocking and dedup model as well as the data in the mastering project to find matches for incoming queries.
Typically, after you make changes to clusters, you publish them and then use LLM updateClusters. If there no changes (clusters have not been published since the last time you used LLM) -1 is returned.
This method returns the ID of an operation unless LLM is up-to-date with the latest dedup data. If LLM has not been initialized it returns the initialization operation. If LLM has been initialized, but data needs to be materialized, it will return the materialization operation. If LLM initialization is required, this method may generate both operations and return the initialization operation. If the data is up-to-date, -1 is returned.
During an LLM update, existing LLM data is used to process match requests. Once the update is complete the newer data is used for matching. LLM manages update transactions to ensure LLM data is always consistent.