Skip to content

api: add Model to RequestMessage - #456

Closed
wseaton wants to merge 3 commits into
llm-d:mainfrom
wseaton:request-model-field
Closed

wseaton wants to merge 3 commits into
llm-d:mainfrom
wseaton:request-model-field

Conversation

@wseaton

@wseaton wseaton commented Sep 17, 2026 •

Copy link
Copy Markdown
Contributor

Producers set RequestMessage.Model, so a sampled span gets the model without a 2.8 ms payload scan at 128k-token prompts. Stacked on #455 (the last two commits are new).

span model lookup

Release note

Adds `RequestMessage.Model`. Producers copy it, and the worker labels trace spans from it instead of parsing the payload.

The worker forwards the queued bytes instead of re-encoding a map, and parses the payload for the span's model only when the span is sampled.

Signed-off-by: Will Eaton <weaton@redhat.com>
Producers that know the model set it, so the worker labels its span without parsing the payload. The payload is still parsed for a sampled span when Model is empty.

Signed-off-by: Will Eaton <weaton@redhat.com>
Signed-off-by: Will Eaton <weaton@redhat.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant