In https://github.com/elastic/kibana/pull/168440, we are implementing token count tracking for streaming responses for OpenAI inside of the framework. This is a small iteration but ideally this should be generalised. Per @mikecote:
We should add a capability for connector types to report extra data within their code (instead of the framework)
Streaming responses should generally be handled, not just for the gen-ai connector
In https://github.com/elastic/kibana/pull/168440, we are implementing token count tracking for streaming responses for OpenAI inside of the framework. This is a small iteration but ideally this should be generalised. Per @mikecote:
gen-ai
connector