Analyze error logs with GPT-4o and create GitHub issues plus Slack alerts

Created by

Last update

Last update 7 hours ago

Who is this for

DevOps engineers, SREs, and development teams who want to automate error monitoring and reduce mean time to resolution (MTTR). Ideal for teams using GitHub for issue tracking and Slack for incident response.

What this workflow does

This workflow automates the entire error management lifecycle - from log ingestion to GitHub issue creation and Slack notification - using GPT-4o-powered root cause analysis. When an application error log is received, it parses the payload, checks for duplicates against existing GitHub issues, generates a structured root cause analysis with fix suggestions, creates a formatted GitHub Issue, and routes Slack notifications by severity. A 30-minute Wait node prevents notification flooding.

How to set up

Add your GitHub Personal Access Token (repo scope) credential
Add your OpenAI API credential to the AI analysis node
Add your Slack OAuth2 credential (chat:write scope) to all Slack nodes
Configure n8n Variables: GITHUB_OWNER and GITHUB_REPO
Update Slack channel names (#incident / #dev-alerts) to match your workspace
Activate the workflow and copy the webhook URL for your application logger

Requirements

GitHub repository with Personal Access Token (repo scope)
OpenAI API account with GPT-4o access
Slack workspace with OAuth2 app installed
Two Slack channels: one for critical incidents, one for general dev alerts

How to customize

Adjust the duplicate detection score threshold (default 60) in the Code node. Modify the GPT-4o prompt to focus on specific error categories. The Wait node duration (30 minutes) can be tuned to match your alerting policy.

Key features

Scoring-based duplicate detection (no extra API calls required)
GPT-4o structured JSON output with graceful fallback parser
Severity-based Slack routing (#incident vs #dev-alerts)
Dynamic GitHub labels: bug, auto-generated, environment, and critical
n8n Variables used for GitHub owner and repo - no hardcoded values

Node List

#	Node Name	Type	Purpose
1	Webhook for Error Logs	Webhook	Accepts error log payload via HTTP POST
2	Parse and Enrich Log	Code	Normalizes level, extracts error type, builds duplicate search keyword
3	Search GitHub Issues	GitHub	Fetches open issues labeled `bug,auto-generated` from the target repo
4	Score GitHub Issue Duplicates	Code	Scores each issue for similarity; flags duplicates at threshold ≥ 60
5	Check for Duplicates	If	Routes to skip path (true) or analysis path (false)
6	Notify Duplicate to Slack	Slack	Posts link to existing issue when duplicate detected
7	Respond with Duplicate Status	Respond to Webhook	Returns 200 OK JSON acknowledgment for duplicate path
8	OpenAI Error Analysis	OpenAI	GPT-4o analyzes root cause and returns structured JSON fix suggestions
9	Build GitHub Issue Body	Code	Parses AI JSON and builds Markdown issue body with tables and code blocks
10	Create GitHub Issue	GitHub	Creates GitHub issue with title, body, and dynamic labels
11	If Critical Error	If	Checks isCritical flag to route Slack notification channel
12	Post Critical Alert to Slack	Slack	Posts @here alert to #incident with full error details
13	Post Error Summary to Slack	Slack	Posts summary to #dev-alerts with GitHub issue link
14	Wait 30 Minutes	Wait	Enforces 30-minute cooldown to prevent notification flooding

Total: 14 nodes (+ 6 Sticky Notes)

Sticky Note Compliance

#	Sticky Note Title	Color	Role
1	Main Sticky Note (Overview)	Yellow	Workflow overview, How it works, Setup steps, Customization
2	Receive and parse log	White	Covers webhook reception and log parsing
3	Check for duplicate GitHub issues	White	Covers GitHub search and duplicate scoring
4	Handle duplicates and notify	White	Covers duplicate branch (notification + webhook response)
5	Analyze error and suggest fixes	White	Covers AI analysis and issue creation
6	Send alerts and summarize	White	Covers severity check, Slack notifications, and wait

All sticky notes use H2 headings (## ) and follow n8n public guidelines.

Webhook payload schema

{
  "service": "payment-api",
  "level": "CRITICAL",
  "message": "NullPointerException at PaymentProcessor.java:142",
  "stack_trace": "java.lang.NullPointerException...",
  "environment": "production",
  "timestamp": "2025-01-15T09:23:45Z",
  "trace_id": "abc-123-xyz",
  "endpoint": "/api/v2/payments",
  "http_method": "POST",
  "status_code": 500,
  "user_id": "user_98765"
}

Required fields: service, level, message
Optional fields: All others - missing values are handled gracefully with fallbacks.

How duplicate detection works

Match condition	Points
Service name found in issue title	+40
Error type found in issue title	+40
Keyword overlap (words > 4 chars)	+5 per word (max +20)

Threshold: Score ≥ 60 → duplicate detected → skip issue creation, notify Slack.