chore(ci): run proxy benchmark manually

[elyasmnvidian]

on:
  pull_request:
  push:
    branches: [main]

Every pull request and main push starts the proxy benchmark. The latest post-merge run spent 51 minutes waiting for a runner and then occupied that runner for four minutes:

workflow created: 2026-08-19T22:34:23Z
job started:      2026-08-19T23:25:43Z
job completed:    2026-08-19T23:29:43Z

Run: https://github.com/NVIDIA-NeMo/Switchyard/actions/runs/32309422644

Fix

The workflow now runs only when someone selects Run workflow. This removes the proxy benchmark from automatic pull-request and main CI while keeping the same AIPerf inputs available for manual measurements.

The downside is that CI no longer runs this performance check automatically. A developer must start it when a routing or proxy change needs performance evidence.

Before and after

Before: pull_request, push to main, workflow_dispatch
After:  workflow_dispatch only

The required CI Success and DCO checks are unchanged.

Testing

git diff --check
ruby -e 'require "yaml"; YAML.load_file(".github/workflows/perf.yml")'

Both commands pass.

Summary by CodeRabbit

• Chores
  • Performance checks now run only when manually started.
  • Automatic runs for pull requests and code updates have been removed.
  • Superseded performance runs are no longer automatically cancelled.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 9ec6c85b-4c5f-42d7-a21c-51bb65ceb726

📥 Commits

Reviewing files that changed from the base of the PR and between d52722e and 4fb96de.

📒 Files selected for processing (1)

• .github/workflows/perf.yml

Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review.

---

Walkthrough

The performance workflow was renamed to indicate manual execution. Automatic pull request and main branch push triggers were removed. Manual dispatch remains available.

Changes

Performance workflow

Layer / File(s)	Summary
Manual benchmark trigger .github/workflows/perf.yml	The workflow name now includes manual. Automatic pull request and main branch push triggers were removed. Manual dispatch remains configured.

Estimated code review effort: 1 (Trivial) | ~3 minutes

Merge Risk: ⚪ Minimal · up to 4fb96

This PR limits the proxy benchmark to manual runs without introducing a concrete correctness or production risk. No actionable merge-blocking risk remains after normal checks and review.

Poem

A rabbit taps the workflow door,
“Run benchmarks when you choose once more!”
No push or pull request will start the race,
Manual hops now set the pace.
Carrot-certified and neat!

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: making the proxy benchmark run manually.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

---

_{Comment @coderabbitai help to get the list of available commands.}