icy-fish opened a new issue, #14029: URL: https://github.com/apache/apisix/issues/14029
### Description I'm using APISIX as my LLM gateway to proxy LLM request and collection agent trajectory data, send those data to Kafka, then processed by logstash. To ensure requests in single agent chat session being processed in the order of its original sending sequence, the best way may be enqueue requests of a single user into a specified Kafka partition, that means I need to ensure same kafka-logger "key" config for a specified user. Existing kafka-logger can only set a fixed string as "key" for single route, means all requests processed by this route will go to same kafka partition. This can partly solve my issue, but comes out a new problem: only 1 Kafka partition being used, then only 1 consumer(Logstsh) can actually work to ensure logs being processed orderly, I cannot scale-up my service to increase throughput. So the best solution may be set unique kakfa-logger "key" for a single user, which in my case, can be extracted from HTTP request header, like "x-auth-token". Hope you can considering this feature, thanks! -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
