Add docs on trigger ci (#13513)

Co-authored-by: sglang-bot <sglangbot@gmail.com>

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

@@ -22,8 +22,6 @@ jobs:
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
-        # We checkout the current context to get the script,
-        # but the script will fetch permissions from main as requested.
 
       - name: Set up Python
         uses: actions/setup-python@v5
@@ -32,7 +30,7 @@ jobs:
 
       - name: Install dependencies
         run: |
-          pip install requests PyGithub
+          pip install PyGithub
 
       - name: Handle Slash Command
         env:

docs/developer_guide/contribution_guide.md

@@ -68,19 +68,21 @@ You can follow the pull request merge process described in [MAINTAINER.md](https
 You will need to work with the Merge Oncall, Codeowner, and other reviewers to get their approvals.
 Then your PR can be merged.
 
-## How to trigger CI
+## How to Trigger CI Tests
 
-To trigger CI, the pull request must have the "run-ci" label.
+We have a lot of open PRs but limited CI machines, so only top and trusted contributors have permission to trigger CI tests.
 
-- The "run-ci" label is not added automatically to new pull requests.
-- Only collaborators with triage or higher permission can add the "run-ci" label.
-- If you have triage/write/maintain/admin access, you can manually add the label by clicking "Labels" on the right side of the pull request.
-- If you do not have triage or higher permission, please request a review and ask a collaborator to add the label for you.
+Users with permission are listed here:
+https://github.com/sgl-project/sglang/blob/main/.github/CI_PERMISSIONS.json
 
-After the "run-ci" label is added, the PR author can trigger CI by:
+For CI to run on a pull request, it must have the **run-ci** label. Authorized users can add the label or rerun failed tests by commenting on the PR with one of these commands:
 
-1. Pushing new commits, or
-2. Clicking "Re-run" / retrigger on the workflow page.
+- `/tag-run-ci-label`: Tag the "run-ci" label. Every future commits will trigger CI.
+- `/rerun-failed-ci`: Rerun the failed/flaky tests of the last commit.
+
+If you have permission, the [Slash Command Handler](https://github.com/sgl-project/sglang/actions/workflows/slash_command_handler.yml) will execute your command and react with a 👍 on your comment. It may take up to several minutes to react. Here is a usage [example](https://github.com/sgl-project/sglang/pull/13498#issuecomment-3547552157).
+
+If you don’t have permission, please ask maintainers to trigger CI for you.
 
 ## Code style guidance
 - Avoid code duplication. If the same code snippet (more than five lines) appears multiple times, extract it into a shared function.

scripts/ci/slash_command_handler.py

@@ -2,11 +2,10 @@ import json
 import os
 import sys
 
-import requests
-from github import Github
+from github import Auth, Github
 
 # Configuration
-PERMISSIONS_FILE_URL = "https://raw.githubusercontent.com/sgl-project/sglang/main/.github/CI_PERMISSIONS.json"
+PERMISSIONS_FILE_PATH = ".github/CI_PERMISSIONS.json"
 
 
 def get_env_var(name):
@@ -19,15 +18,18 @@ def get_env_var(name):
 
 def load_permissions(user_login):
     """
-    Downloads the permissions JSON from the main branch and returns
+    Reads the permissions JSON from the local file system and returns
     the permissions dict for the specific user.
     """
     try:
-        print(f"Fetching permissions from {PERMISSIONS_FILE_URL}...")
-        response = requests.get(PERMISSIONS_FILE_URL)
-        response.raise_for_status()
+        print(f"Loading permissions from {PERMISSIONS_FILE_PATH}...")
+        if not os.path.exists(PERMISSIONS_FILE_PATH):
+            print(f"Error: Permissions file not found at {PERMISSIONS_FILE_PATH}")
+            return None
+
+        with open(PERMISSIONS_FILE_PATH, "r") as f:
+            data = json.load(f)
 
-        data = response.json()
         user_perms = data.get(user_login)
 
         if not user_perms:
@@ -79,8 +81,13 @@ def handle_rerun_failed_ci(gh_repo, pr, comment, user_perms):
         # We only care about completed runs that failed
         if run.status == "completed" and run.conclusion == "failure":
             print(f"Rerunning workflow: {run.name} (ID: {run.id})")
-            run.rerun_failed()
-            rerun_count += 1
+            try:
+                # PyGithub uses rerun_failed_jobs() or rerun() depending on version/intent
+                # The traceback suggested rerun_failed_jobs
+                run.rerun_failed_jobs()
+                rerun_count += 1
+            except Exception as e:
+                print(f"Failed to rerun workflow {run.id}: {e}")
 
     if rerun_count > 0:
         comment.create_reaction("+1")
@@ -98,19 +105,21 @@ def main():
     comment_body = get_env_var("COMMENT_BODY").strip()
     user_login = get_env_var("USER_LOGIN")
 
-    # 2. Initialize GitHub API
-    g = Github(token)
-    repo = g.get_repo(repo_name)
-    pr = repo.get_pull(pr_number)
-    comment = repo.get_issue(pr_number).get_comment(comment_id)
-
-    # 3. Load Permissions (Remote Check)
+    # 2. Load Permissions (Local Check)
     user_perms = load_permissions(user_login)
 
     if not user_perms:
         print(f"User {user_login} does not have any configured permissions. Exiting.")
         return
 
+    # 3. Initialize GitHub API with Auth
+    auth = Auth.Token(token)
+    g = Github(auth=auth)
+
+    repo = g.get_repo(repo_name)
+    pr = repo.get_pull(pr_number)
+    comment = repo.get_issue(pr_number).get_comment(comment_id)
+
     # 4. Parse Command and Execute
     # split lines to handle cases where there might be text after the command
     first_line = comment_body.split("\n")[0].strip()