site-autofix

A Node.js CLI tool that detects broken links, auto-fixes href references, generates redirect configurations for major platforms, and monitors endpoints for 500-class server errors.

What it does

• Broken Link Detection: Crawls your site using Playwright, discovers all internal links, and checks each one for 404s, redirect chains, and server errors.
• Auto-Fix Suggestions: Computes fix suggestions for broken links using fuzzy matching and path similarity against known-good URLs. Applies fixes to source files when confidence is high enough.
• Redirect Generation: Produces ready-to-use redirect configurations for Next.js (next.config.js redirects array), Netlify (_redirects file), and nginx (map block).
• Health Monitoring: Performs periodic synthetic checks against your endpoints, detects 500-class errors, extracts stack traces, classifies the error pattern (timeout, OOM, database, upstream, config, unhandled exception), and recommends bounded remediation actions (restart, rollback, feature-flag-off, or open-issue).

Installation

npm install -g site-autofix

Or run directly from the repo:

git clone <repo-url>
cd site-autofix
npm install
npm run build

CLI Reference

site-autofix scan <url>

Crawl a site and report broken links.

site-autofix scan https://example.com

Option	Default	Description
--max-pages <n>	100	Maximum pages to crawl
--timeout <ms>	15000	Request timeout
--concurrency <n>	5	Concurrent link checks
--exclude <patterns...>	—	URL patterns to exclude (regex)
--output <format>	console	Output: json, console, or both
--output-file <path>	—	Write JSON report to file

Exit codes: 0 = all links healthy, 1 = broken links found, 2 = scan error.

site-autofix fix <url>

Scan for broken links and generate redirect configurations.

site-autofix fix https://example.com --target ./redirects --format netlify

Option	Default	Description
--target <dir>	.	Directory to write redirect configs
--format <fmt>	all	Redirect format: nextjs, netlify, nginx, or all
--min-confidence <n>	0.5	Minimum confidence threshold (0-1)
--max-pages <n>	100	Maximum pages to crawl
--timeout <ms>	15000	Request timeout
--concurrency <n>	5	Concurrent link checks
--output <format>	console	Output: json, console, or both
--output-file <path>	—	Write JSON report to file

Generated files:

• generated-redirects.js (Next.js)
• _redirects (Netlify)
• nginx-redirects.conf (nginx)

site-autofix monitor <url>

Periodically check endpoints for 500-class errors.

site-autofix monitor https://example.com --endpoints / /api/health /api/data --rounds 5

Option	Default	Description
--endpoints <paths...>	/	Endpoint paths to check
--interval <ms>	30000	Interval between rounds
--timeout <ms>	10000	Request timeout
--rounds <n>	1	Number of rounds (0 = indefinite)
--output <format>	console	Output: json, console, or both
--output-file <path>	—	Write JSON report to file

Error classifications:

• timeout — Endpoint timed out
• oom — Out of memory detected
• database-error — Database connection issue
• upstream-dependency — Upstream service failure (502/503/504)
• configuration-error — Missing config or env vars
• unhandled-exception — Uncaught error with stack trace
• unknown — Unclassifiable error

Recommended actions per classification:

• restart — Restart the service (timeout, OOM, database)
• rollback — Roll back to last known good deploy (config errors)
• feature-flag-off — Disable the affected feature (unhandled exceptions)
• open-issue — File an issue with repro artifacts (upstream, unknown)

Architecture

src/
  cli.ts                 CLI entry point (commander)
  types.ts               Shared TypeScript types
  link-scanner.ts        Crawl + check links (Playwright + fetch)
  link-fixer.ts          Fuzzy matching + fix computation
  redirect-generator.ts  Next.js / Netlify / nginx config output
  health-monitor.ts      Synthetic health checks + error classification
  reporter.ts            Console + JSON output formatting

How it works

1. Scanning uses Playwright to render pages (handling JS-rendered content), then extracts all <a href> elements. Each discovered internal link is checked via fetch with redirect tracking.

2. Fixing takes the broken links and compares them against known-good URLs using a weighted combination of path segment overlap (60%) and Levenshtein edit distance (40%). Links with redirect targets get a 95% confidence automatic fix.

3. Redirect generation converts the fix map into platform-specific config files, each with inline comments explaining how to integrate them.

4. Monitoring runs periodic fetch requests against configured endpoints, collects response headers and bodies for failing requests, extracts stack traces using regex patterns for Node.js/Python/Java/Ruby, classifies the error, and maps it to a bounded remediation action.

Requirements

• Node.js >= 18
• Playwright browsers (installed automatically on first run, or run npx playwright install chromium)

License

MIT