GitHub user xuang7 added a comment to the discussion: Ambient "operator 
recommender" — predictive next-operator suggestions on the canvas

The idea sounds very good! This would significantly improve the usability of 
workflow building, and the plan discussed above seems like a solid way to get 
there incrementally.

As discussed, version 1 can return hardcoded recommendations, so the 
`/recommend` endpoint does not need to call an LLM yet. This allows you to wire 
up and demo the full loop with zero API cost: the frontend listens for the 
"operator added" event, calls the endpoint, and renders the ghost suggestions. 
The `agent-service` should still run locally, since that is where the new 
endpoint should live, but for version 1, the default configuration should be 
sufficient because no LLM call is involved.

For version 2, which connects to a real LLM, you could follow the suggestions 
earlier in this thread: reuse the same gateway that the chatbot already uses, 
and start with a smaller, cheaper model. That part can come later after we get 
more input from the community.

GitHub link: 
https://github.com/apache/texera/discussions/5240#discussioncomment-17578163

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to