Skip to content

Fix the simulations CI job: lk agent simulate needs the text subcommand - #107

Merged
u9g merged 1 commit into
mainfrom
bcherry/fix-simulate-subcommand
Sep 25, 2026
Merged

u9g merged 1 commit into
mainfrom
bcherry/fix-simulate-subcommand

Conversation

@bcherry

@bcherry bcherry commented Sep 24, 2026

Copy link
Copy Markdown
Contributor

What

The simulations CI job runs lk agent simulate --scenarios scenarios.yaml. Since livekit-cli v2.18 the root simulate command only prints help and exits 0 — the mode is a required subcommand (text or audio). So the job has been green without running a single simulation. This switches the job, and the four places that show the command (README, AGENTS.md, the scenarios.yaml header, and the workflow comment), to:

lk agent simulate text --scenarios scenarios.yaml

Verify with lk agent simulate --help on a current CLI; the workflow installs the latest release, so it has been affected since that release.

Why text mode

It exercises the LLM, tools and conversation logic without the STT/TTS pipeline, it's deterministic enough for CI, and LiveKit schedules text-mode simulations so they don't compete with live sessions for inference. Audio runs execute in real time and bill the speech providers every turn — right for a release candidate, not a recurring job.

Note

The first green run after this merges will be the first time the scenarios have actually executed in a while, so a failure there is information, not necessarily a regression from this change.

Since livekit-cli v2.18, the root `lk agent simulate` command only prints help
and exits 0; the mode is a required subcommand (`text` or `audio`). The
simulations job has been running the old flat form, so it has been passing
without starting a single simulation. Switch the job and the four places that
show the command to `lk agent simulate text --scenarios scenarios.yaml`.

Text mode is the right mode for CI: it exercises the LLM, tools and
conversation logic without the audio pipeline, and is scheduled so it doesn't
compete with live sessions for inference.
@bcherry
bcherry requested a review from u9g September 24, 2026 16:49
@u9g
u9g merged commit e1854f7 into main Sep 25, 2026
5 checks passed
@u9g
u9g deleted the bcherry/fix-simulate-subcommand branch September 25, 2026 13:03
bcherry pushed a commit that referenced this pull request Sep 25, 2026
Keeps the rewritten Testing paragraph and carries over #107's
`lk agent simulate text` fix.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01U7yszNU7t5aCHJsKLmU4hv
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.

2 participants