[CI] Add /run-full-ci and /run-extra-ci slash commands (#38734)

.claude/skills/ci-workflow-guide/SKILL.md

@@ -388,6 +388,8 @@ group: pr-test-{event_name}-{branch}-{pr_sha}-{stage}
 | `/rerun-failed-ci` | Reruns failed jobs in the latest workflow run |
 | `/tag-and-rerun-ci` | Adds `run-ci` label + reruns failed |
 | `/tag-and-rerun-ci extra` | Adds both `run-ci` and `run-ci-extra` labels + reruns failed |
+| `/run-full-ci` | Short form of `/tag-and-rerun-ci extra` (baseline + extra). Alias: `/rerun-full-ci` |
+| `/run-extra-ci` | Adds both labels + reruns **only** `PR Test Extra`, leaving baseline runs alone. Alias: `/rerun-extra-ci` |
 | `/rerun-test <test-file> [<test-file> ...]` | Reruns specific test file(s) via `rerun-test.yml`. A file arg containing a glob metacharacter (`*`, `?`, `[...]`) expands against `test/registered/` and the multimodal test dir to every matching `test_*.py` (e.g. `/rerun-test test_*backend*.py` — wrap in backticks so GitHub doesn't italicize the `*`); matches are deduped, grouped by dispatch shape, and can't carry a `::test` selector. No match → single ⛔ reply, nothing dispatched. Each reply echoes its originating command (`Results for …`) so concurrent commands stay distinguishable |
 | `/rerun-group <group> [<group> ...]` | Expands registered test groups, then reuses `/rerun-test` |
 

.github/FOLDER_README.md

@@ -12,8 +12,8 @@ Recognized permission keys:
 
 | Key | Grants |
 | --- | --- |
-| `can_tag_run_ci_label` | `/tag-run-ci-label`, `/tag-and-rerun-ci` |
+| `can_tag_run_ci_label` | `/tag-run-ci-label`, `/tag-and-rerun-ci`, `/run-full-ci`, `/run-extra-ci` |
-| `can_rerun_failed_ci` | `/rerun-failed-ci`, `/tag-and-rerun-ci` |
+| `can_rerun_failed_ci` | `/rerun-failed-ci`, `/tag-and-rerun-ci`, `/run-full-ci`, `/run-extra-ci` |
 | `cooldown_interval_minutes` | rate limit in `pr-gate.yml`; `0` also grants `/rerun-test`, `/rerun-group` |
 
 `/rerun-test` and `/rerun-group` are gated on the commenter alone: either

.github/workflows/slash-command-handler.yml

@@ -19,6 +19,10 @@ jobs:
       (contains(github.event.comment.body, '/tag-run-ci-label') ||
        contains(github.event.comment.body, '/rerun-failed-ci') ||
        contains(github.event.comment.body, '/tag-and-rerun-ci') ||
+       contains(github.event.comment.body, '/run-full-ci') ||
+       contains(github.event.comment.body, '/run-extra-ci') ||
+       contains(github.event.comment.body, '/rerun-full-ci') ||
+       contains(github.event.comment.body, '/rerun-extra-ci') ||
        contains(github.event.comment.body, '/rerun-group') ||
        contains(github.event.comment.body, '/rerun-test'))
     runs-on: ubuntu-latest

docs/docs/developer_guide/contribution_guide.mdx

@@ -122,14 +122,17 @@ For CI to run on a pull request, it must have the "run-ci" label. Authorized use
 - `/tag-run-ci-label`: Adds the "run-ci" label. Only **future** commits trigger CI; the current commit is unaffected. Add the `extra` argument (`/tag-run-ci-label extra`) to additionally apply the "run-ci-extra" label, opting the PR into the extra test workflow (`pr-test-extra.yml`).
 - `/rerun-failed-ci`: Reruns workflows from the latest commit with conclusion **failed, skipped, cancelled, or timed out**.
 - `/tag-and-rerun-ci`: Runs both. Use this on a fresh PR to kick off CI on the current commit — `/tag-run-ci-label` alone won't. Accepts the same `extra` argument (`/tag-and-rerun-ci extra`).
+- `/run-full-ci`: Baseline **and** extra CI. Short form of `/tag-and-rerun-ci extra` — applies both labels, then reruns everything at the current commit that didn't succeed.
+- `/run-extra-ci`: Extra CI **only**. Applies both labels, then reruns just `PR Test Extra`, leaving baseline CI runs alone. Use it when baseline is already green and you only need the extra tier.
 - `/rerun-test <test-spec> [<test-spec> ...]`: Reruns one or more specific tests directly. A spec may select a file, class, or method using `<file>::<TestClass>[.<test_method>]`. Multiple specs and file globs are supported. `--changed` (short form `-c`) adds every test file the PR itself adds or modifies (under `test/registered/` or the multimodal test directory), so a PR that touches several tests can rerun them all without listing them. Examples: `/rerun-test test_srt_endpoint.py`, `/rerun-test registered/core/test_srt_endpoint.py::TestSRTEndpoint.test_simple_decode`, `/rerun-test test_a.py test_b.py`, `/rerun-test test_*backend*.py`, `/rerun-test --changed`, or `/rerun-test -c`.
-- `/rerun-group <group> [<group> ...]`: Expands one or more registered test groups (for example, `/rerun-group hicache`) and dispatches their tests through the same selective-rerun workflow.
+- `/rerun-group <group> [<group> ...]`: Expands one or more registered test groups (for example, `/rerun-group rust-server`) and dispatches their tests through the same selective-rerun workflow. Groups are defined in [`scripts/ci/rerun_test_groups.json`](https://github.com/sgl-project/sglang/blob/main/scripts/ci/rerun_test_groups.json).
 
 The rerun commands have the following permission rules:
 
 | Command | PR author | Other users | Additional rule for fork PRs |
 | --- | --- | --- | --- |
 | `/rerun-failed-ci` | Always allowed on their own PR | Must have `can_rerun_failed_ci` in `CI_PERMISSIONS.json` | None |
+| `/run-full-ci`, `/run-extra-ci` | Not auto-allowed -- these apply labels, which the PR-author allowance does not cover | Must have both `can_tag_run_ci_label` and `can_rerun_failed_ci` in `CI_PERMISSIONS.json` (every listed user has both) | None |
 | `/rerun-test`, `/rerun-group` | No special allowance; judged as any other commenter | Must have `cooldown_interval_minutes: 0` in `CI_PERMISSIONS.json`, or `write`/`admin` repository permission | None -- the same rule applies wherever the PR comes from |
 
 <Warning>

scripts/ci/utils/slash_command_handler.py

@@ -32,6 +32,8 @@ PRECISION_BASELINE_TEST = "registered/debug_utils/test_nightly_precision_regress
 PRECISION_BASELINE_REFRESH_FLAG = "--refresh-precision-baseline"
 CHANGED_TESTS_FLAG = "--changed"
 CHANGED_TESTS_SHORT_FLAG = "-c"
+# Workflow `name:` field, which is what the runs API reports (not the filename).
+EXTRA_WORKFLOW_NAME = "PR Test Extra"
 
 
 MAINTENANCE_ISSUE_NUMBER = 21065
@@ -495,6 +497,94 @@ def handle_rerun_failed_ci(gh_repo, pr, comment, user_perms, react_on_success=Tr
         return False
 
 
+def _run_extra_ci(gh_repo, pr):
+    """
+    Make `PR Test Extra` run for the current head SHA.
+
+    Assumes `run-ci` + `run-ci-extra` are already on the PR. pr-test-extra.yml
+    gates on them through call-gate -> pr-gate.yml, which live-fetches the
+    label set at run time, so re-running a run that was red at the gate picks
+    up labels added seconds earlier.
+
+    Deliberately rerun-only, with no workflow_dispatch fallback: the extra
+    workflow's dispatch inputs carry no `pr_head_sha`, so a fork PR could only
+    be dispatched against `main` — testing the wrong tree. A pull_request run
+    exists at essentially every head SHA anyway, so the missing-run case is
+    reported rather than papered over.
+
+    Returns (acted, message).
+    """
+    extra_runs = [
+        run
+        for run in gh_repo.get_workflow_runs(head_sha=pr.head.sha)
+        if run.name == EXTRA_WORKFLOW_NAME
+    ]
+    if not extra_runs:
+        return (
+            False,
+            f"No `{EXTRA_WORKFLOW_NAME}` run exists for this commit, so there is "
+            "nothing to re-run. Push a commit to create one.",
+        )
+
+    # Same newest-wins rule handle_rerun_failed_ci uses: older runs at this SHA
+    # are superseded and re-running them fights the live run for runners.
+    run = _latest_run_per_workflow(extra_runs)[0]
+
+    if run.status != "completed":
+        return (
+            True,
+            f"⏳ [`{EXTRA_WORKFLOW_NAME}`]({run.html_url}) is already "
+            f"{run.status} for this commit; left it alone.",
+        )
+
+    try:
+        if run.conclusion == "success":
+            # Nothing failed, so there are no jobs for rerun_failed_jobs() to
+            # target — the ask is an explicit re-run of green work.
+            print(f"  Full rerun of successful run {run.id}")
+            run.rerun()
+        else:
+            try:
+                print(f"  rerun_failed_jobs on {run.id} ({run.conclusion})")
+                run.rerun_failed_jobs()
+            except Exception as e:
+                print(f"  rerun_failed_jobs rejected ({e}) - full rerun")
+                run.rerun()
+    except Exception as e:
+        return False, f"Failed to re-run [`{EXTRA_WORKFLOW_NAME}`]({run.html_url}): {e}"
+
+    return True, f"🚀 Re-running [`{EXTRA_WORKFLOW_NAME}`]({run.html_url})."
+
+
+def handle_run_extra_ci(gh_repo, pr, comment, user_perms):
+    """
+    Handles /run-extra-ci: label the PR for extra CI and re-run only
+    `PR Test Extra`. Baseline CI runs are left untouched.
+
+    Gated on the same two permissions as the steps it performs — the label add
+    (can_tag_run_ci_label) and the re-run (can_rerun_failed_ci) — rather than a
+    new key, so every user who can already drive CI can drive extra CI.
+    Returns True if action was taken, False otherwise.
+    """
+    if not user_perms.get("can_rerun_failed_ci", False):
+        print("Permission denied: can_rerun_failed_ci is false.")
+        return False
+
+    tagged = handle_tag_run_ci(
+        gh_repo, pr, comment, user_perms, react_on_success=False, tag_extra=True
+    )
+    if not tagged:
+        return False
+
+    print("Waiting 5 seconds for labels to propagate...")
+    time.sleep(5)
+
+    acted, message = _run_extra_ci(gh_repo, pr)
+    pr.create_issue_comment(message if acted else f"⛔ {message}")
+    comment.create_reaction("+1" if acted else "confused")
+    return acted
+
+
 MULTIMODAL_TEST_DIR = "python/sglang/multimodal_gen/test"
 
 MULTIMODAL_PATH_TO_RUNNER = {
@@ -1575,15 +1665,28 @@ def main():
     tokens = first_line.split()
     tag_extra = len(tokens) > 1 and "extra" in tokens[1:]
 
+    # /run-full-ci is the short, spelled-out form of `/tag-and-rerun-ci extra`
+    # and shares its branch below. /rerun-* spellings are accepted silently as
+    # aliases: they sit one keystroke from the documented name and from the
+    # neighbouring /rerun-* commands, and failing them closed would just look
+    # like the bot ignoring the comment.
+    is_full_ci = first_line.startswith(("/run-full-ci", "/rerun-full-ci"))
+    is_extra_ci = first_line.startswith(("/run-extra-ci", "/rerun-extra-ci"))
+    if is_full_ci:
+        tag_extra = True
+
     if first_line.startswith("/tag-run-ci-label"):
         handle_tag_run_ci(repo, pr, comment, user_perms, tag_extra=tag_extra)
 
+    elif is_extra_ci:
+        handle_run_extra_ci(repo, pr, comment, user_perms)
+
     elif first_line.startswith("/rerun-failed-ci"):
         handle_rerun_failed_ci(repo, pr, comment, user_perms)
 
-    elif first_line.startswith("/tag-and-rerun-ci"):
+    elif first_line.startswith("/tag-and-rerun-ci") or is_full_ci:
         # Perform both actions, but suppress individual reactions
-        print(f"Processing combined command: /tag-and-rerun-ci (tag_extra={tag_extra})")
+        print(f"Processing combined command: {first_line} (tag_extra={tag_extra})")
 
         tagged = handle_tag_run_ci(
             repo, pr, comment, user_perms, react_on_success=False, tag_extra=tag_extra