Files
awesome-kubernetes/src/pr_guardian.py
Nubenetes Bot 15ffc91ec2 fix(ci): PR Guardian must not auto-format the generated v2-docs/ tree
The Guardian's auto_format_file ("&"->"and" in headings, URL trailing-slash
normalization) ran on every changed .md including the generated v2-docs/, then
committed the result back to the PR branch. Because PRs land on develop while
release branches merge to master via plain git merge (no Guardian run), develop
ended up with the normalized form and master with the generator's raw form —
a perpetual cosmetic develop<->master drift that spawned recurring no-op
'Release: Agentic V2 Portal Update' PRs (e.g. #399).

v2_optimizer.py is the single source of truth for v2-docs/ formatting, so the
Guardian now skips that tree in the auto-format loop and drops it from the
auto-commit 'git add'. Hand-authored docs/ and README/inventory are unaffected.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-20 12:14:49 +02:00

162 lines
6.2 KiB
Python

import os
import re
import sys
import httpx
import asyncio
import subprocess
from github import Github
from src.config import GH_TOKEN, TARGET_REPO
from src.gemini_utils import call_gemini_with_retry
async def analyze_pr_diff(diff_text: str) -> str:
prompt = f"""
You are the Nubenetes PR Guardian. Analyze this PR diff against the Nubenetes GEMINI.md mandates.
Check for:
1. Are new links added using the high-density description standard? (No generic clickbait)
2. Are URLs normalized? (No tracking params)
3. Are there emojis in H2/H3 titles? (Should be rejected)
If the PR looks good, say '✅ PR Meets Nubenetes Standards.'
If it violates mandates, list the violations clearly so the contributor can fix them.
PR Diff:
{diff_text}
"""
try:
# Mandate 50: Use Flash/Lite for rapid auditing and specify text format to avoid JSON parsing loops
response = await call_gemini_with_retry(prompt, prefer_flash=True, response_format="text", role="PR-Guardian")
await asyncio.sleep(1.0) # Safety delay to respect global RPM quota
return str(response)
except Exception as e:
return f"⚠️ Guardian AI temporarily unavailable: {e}"
def auto_format_file(filepath: str):
if not os.path.exists(filepath):
return
with open(filepath, "r", encoding="utf-8") as f:
content = f.read()
original_content = content
# 1. Clean URLs
from src.gemini_utils import normalize_url
def repl_link(match):
text = match.group(1)
url = match.group(2)
if url.startswith(("http://", "https://")):
try:
cleaned = normalize_url(url)
return f"[{text}]({cleaned})"
except Exception as e:
print(f"[WARN] URL normalization failed for '{url[:60]}': {str(e)[:100]}")
return match.group(0)
return match.group(0)
content = re.sub(r'\[([^\]]+)\]\((https?://[^\)]+)\)', repl_link, content)
# 2. Clean headers (e.g., ampersands to "and", strip emojis)
lines = content.splitlines()
new_lines = []
for line in lines:
if line.strip().startswith("#"):
line = line.replace("&", "and")
line = re.sub(r'[\U00010000-\U0010ffff]', '', line)
for char in "🧠☁️🌟🛡️🧪💎🕵️✨⚠️🔴🟡✅📍🎥":
line = line.replace(char, "")
line = re.sub(r'\s+', ' ', line).strip()
match = re.match(r'^(#+)\s*(.*)', line)
if match:
level, title = match.groups()
line = f"{level} {title.strip()}"
new_lines.append(line)
content = "\n".join(new_lines) + ("\n" if original_content.endswith("\n") else "")
# 3. Clean <center> tags (Mandate 19)
content = re.sub(r'<center(?! markdown="1")>', '<center markdown="1">', content)
if content != original_content:
with open(filepath, "w", encoding="utf-8") as f:
f.write(content)
print(f"[*] Auto-formatted file: {filepath}")
def main():
pr_number = os.getenv("PR_NUMBER")
if not pr_number:
print("No PR_NUMBER provided.")
return
g = Github(GH_TOKEN)
repo = g.get_repo(TARGET_REPO)
pr = repo.get_pull(int(pr_number))
# Get changed files in PR and run auto-formatting
try:
files = pr.get_files()
modified_files = [f.filename for f in files if f.status in ["modified", "added"]]
for filepath in modified_files:
# The v2-docs/ tree is generated by src/v2_optimizer.py, which is its
# single source of truth for formatting. Auto-formatting it here
# ("&"->"and" in headings, URL trailing-slash stripping) only fights
# the generator: the rewrite lands on develop via the PR while master
# gets the generator's raw form via release merges, producing a
# perpetual cosmetic develop<->master drift. Skip the generated tree.
if filepath.endswith(".md") and not filepath.startswith("v2-docs/"):
auto_format_file(filepath)
except Exception as e:
print(f"Warning: Failed to fetch files for auto-formatting: {e}")
# Fetch base branch and get the updated diff (with auto-fixes applied)
diff_text = ""
try:
subprocess.run(["git", "fetch", "origin", pr.base.ref, "--depth=1"], capture_output=True, check=True)
diff_process = subprocess.run(["git", "diff", f"origin/{pr.base.ref}"], capture_output=True, text=True, check=True)
diff_text = diff_process.stdout
except Exception as e:
print(f"Warning: Failed to get git diff locally: {e}.")
if not diff_text.strip():
print("Diff is empty locally, falling back to GitHub PR diff.")
try:
# Mandate: URL Expansion & Normalization: Use follow_redirects=True for httpx
diff_text = httpx.get(pr.diff_url, follow_redirects=True).text
except Exception as e:
print(f"Failed to fetch GitHub PR diff: {e}")
# Skip if too huge or not markdown
if len(diff_text) > 100000:
print("Diff is too large for AI analysis (>100KB). Skipping detailed AI check, but auto-formatting was completed.")
return
analysis = asyncio.run(analyze_pr_diff(diff_text))
# Check if comment already exists to avoid duplication
bot_signature = "<!-- PR_GUARDIAN_SIGNATURE -->"
existing_comment = None
for comment in pr.get_issue_comments():
if bot_signature in comment.body:
existing_comment = comment
break
final_comment = f"{bot_signature}\n### 🛡️ Nubenetes PR Guardian\n\n{analysis}"
if existing_comment:
existing_comment.edit(final_comment)
print("Existing Guardian comment updated.")
else:
pr.create_issue_comment(final_comment)
print("Guardian comment posted.")
# Determine if validation passed
clean_analysis = analysis.strip().lower()
if "pr meets nubenetes standards" in clean_analysis:
print("✅ PR Meets Nubenetes Standards!")
sys.exit(0)
else:
print("❌ PR violates Nubenetes mandates. Exiting with 1.")
sys.exit(1)
if __name__ == "__main__":
main()