forked from elastic/kibana
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[Obs AI Assistant] Improve recall speed (elastic#176428)
Improves recall speed by outputting as CSV with zero-indexed document "ids". Previously, it was a JSON object, with the real document ids. This causes the LLM to "think" for longer, for whatever reason. I didn't actually see a difference in completion speed, but emitting the first value took significantly less time when using the CSV output. I also tried sending a single document per request using the old format, and while that certainly improves things, the slowest request becomes the bottleneck. These are results from about 10 tries per strategy (I'd love to see others reproduce at least the `batch` vs `csv` strategy results): `batch`: 24.7s `chunk`: 10s `csv`: 4.9s --------- Co-authored-by: Søren Louv-Jansen <sorenlouv@gmail.com> Co-authored-by: Kibana Machine <42973632+kibanamachine@users.noreply.github.com>
- Loading branch information
Showing
2 changed files
with
91 additions
and
77 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters