A year in the making. Thanks to @twilio and @Atlassian for their feedback as we built this with them. This is our take on the elusive problem of routing between LLMs: A preference-aligned approach that offers a practical mechanism to capture subjective evaluation criteria.