🔍 Backup Log Analyzer
Paste any backup or infrastructure log — Claude identifies errors, root causes, and fix commands in seconds
→ Try it live
Overview
Drop any log output into the textarea and hit Analyze. Claude Opus 4.7 streams back a structured report with a color-coded severity badge, numbered issues, root causes, and copy-pasteable remediation commands — all in a few seconds. No account required.
Diff Mode — paste two log captures (before / after a change or outage) and get a regression analysis: issues that appeared, issues that resolved, and patterns that changed severity. Switch modes with the toggle above the textarea.
Batch Mode — analyze up to 5 log files in one run (upload or paste). Each file gets its own section in a single consolidated report.
Offline error taxonomy — 12 known PPDM, NetWorker and Data Domain error patterns are classified instantly, with root cause and fix commands, without calling the API. Anything unrecognized falls through to Claude.
History — your last 20 analyses are kept in the browser (localStorage) and can be reloaded with one click.
Supports logs from:
| Source | Examples |
|---|
| Dell EMC PPDM | Protection job failures, Elasticsearch cluster health, asset errors |
| Dell EMC NetWorker | NSR peer mismatches, savefs connection errors, backup failures |
| Data Domain | DDBoost status, filesystem capacity alerts, replication errors |
| Veeam | Job failures, proxy errors, repository warnings |
| Linux syslog | journalctl, /var/log/messages, dmesg |
| Windows Event Log | Application, System, Security event exports |
Severity Levels
The analysis opens with a color-coded severity badge:
| Badge | Meaning |
|---|
| 🔴 CRITICAL | Service down or data at risk — act immediately |
| 🟠 HIGH | Significant failure that will impact backups |
| 🟡 MEDIUM | Warnings that need attention soon |
| 🟢 LOW | Minor issues with minimal impact |
| 🔵 INFO | No problems detected |
Claude Code Plugin
Install as a Claude Code plugin to get /analyze-log directly in your terminal:
From a terminal:
claude plugin marketplace add Moodswing9/claude-plugins
claude plugin install backup-log-analyzer@moodswing9
This installs /analyze-log and the skill.
Or inside Claude Code: /plugin marketplace add Moodswing9/claude-plugins, then /plugin install backup-log-analyzer@moodswing9.
Only need the knowledge skill, without the slash commands? npx skills add Moodswing9/backup-log-analyzer -g installs just the skill.
| Command | What it does |
|---|
/analyze-log /var/log/ppdmwatch/ppdmwatch.log | Analyse a single log file — severity, root cause, fix commands |
/analyze-log /var/log/before.log --diff /var/log/after.log | Diff mode — compare two captures to surface regressions |
The skill auto-activates in Claude Code when you ask about log severity triage, PII redaction patterns, diff mode regression analysis, or the Next.js API route architecture — no command needed. Requires ANTHROPIC_API_KEY and pip install anthropic.
Self-Hosting
The live version at backup-log-analyzer.vercel.app is open to use. To run your own instance:
Prerequisites
Setup
git clone https://github.com/Moodswing9/backup-log-analyzer.git
cd backup-log-analyzer
npm install
Create .env.local:
ANTHROPIC_API_KEY=sk-ant-...
Run
npm run dev
Open http://localhost:3000.
Deploy
npx vercel --prod
npx vercel env add ANTHROPIC_API_KEY production
Stack
| Layer | Technology |
|---|
| Framework | Next.js 16 (App Router) |
| Language | TypeScript |
| Styling | Tailwind CSS 4 |
| AI | Claude Opus 4.7 (claude-opus-4-7) via Anthropic SDK |
| Streaming | Web ReadableStream → response.body.getReader() |
| Deployment | Vercel |
Architecture
POST /api/analyze
↓ validates input (max 50k chars)
↓ streams claude-opus-4-7 response via ReadableStream
↓ client reads chunks with getReader(), appends to state
↓ react-markdown renders streamed output live
Security & Privacy
| Protection | How it works |
|---|
| PII redaction | Bearer/Basic tokens, credential JSON fields ("password": "..." etc.), email addresses, and IPv4 octets are stripped server-side before the log text is forwarded to Claude |
| Per-IP rate limiting | 10 requests per minute per IP (sliding window, in-memory). Exceeding the limit returns HTTP 429 |
| No persistence | Logs are never stored, logged, or retained. Each request is passed to the Anthropic API and discarded after the streaming response completes |
License
This project is licensed under the MIT License — see LICENSE for details.
