Hey everyone! 
Like many of you, I manage several GitHub repositories - some active, some dormant, and a few I honestly forgot existed until I stumbled upon them one random night. I tried a few analytics tools, but they either cost too much for a solo dev, required yet another dashboard I’d never check, or gave me raw numbers without any real insight.
What I really wanted was simple: Every Monday, get a short AI-written report in Telegram telling me what happened across my repos, what needs attention, and what I should prioritize next week. No logging into another site, no BI setup, no spreadsheets.
So I built it in n8n. Took about an afternoon.
What it does
The workflow runs on two triggers:
Weekly schedule: Every Monday at 9 AM, it automatically fetches all repos, aggregates activity, and sends a full digest to Telegram.
Telegram commands: Send /report,/issues,/prs, or/statusto your bot anytime for an instant update.
It uses Qwen via OpenRouter to turn raw GitHub event data into readable, actionable reports. The key design choice: stats are aggregated before hitting the LLM - only counts (commits, issues, PRs, releases) go to the model, not raw event payloads. This keeps token costs negligible even with many repos.
What’s inside the workflow
| Feature | Details |
|---|---|
| Full repo scan | Fetches all repos via GitHub API, auto-filters archived |
| Smart aggregation | Groups events by type, tracks contributors, flags stale repos |
| 4 report modes | /report (full digest), /issues, /prs, /status (health check) |
| AI fallback | If the model call fails, a Code node formats raw stats — you never get silence |
| Error alerts | If all GitHub APIs fail, Telegram gets an error alert |
| Message splitting | Long reports split into multiple messages to respect Telegram’s 4096-char limit |
Sample output (so you can see what I mean)
When I send /report, I get 4 consecutive messages in Telegram:
Message 1: Highlights & summary
📊 GitHub Weekly Digest
14 May – 21 May 2026
This week recorded 13 activities across 5 repositories:
11 issues opened, no commits, no pull requests.
🔥 HIGHLIGHTS
• dothanhvinh17 dominated 100% of activity via issue creation
• No commits or releases published this week
⚠️ NEEDS ATTENTION
• 4 repos are stale (0 events)
• triage-demo has 11 issues and 0 commits — imbalance
Message 2-3: Per-repo breakdown with STALE warnings
Message 4: Weekly insights & next-week priorities
📈 WEEKLY INSIGHTS
• Activity highly concentrated: 100% of actions in 1 repo
• Risk: 4/5 repos show no activity
🔮 NEXT WEEK PRIORITIES
1. Address backlog in triage-demo
2. Review stale repos for archiving
3. Target ≥1 commit or PR across any repo
The /status command is my favorite quick check:
🩺 Repository Health Check
🟢 triage-demo: Active (11 issues)
🟡 dothanhvinh17: Low activity (0 events)
🔴 rag-kb: Stale (0 events)
🔴 bepchu7-landing: Stale (0 events)
Get started
The workflow is published (verified, free) on the n8n.io:
Send Weekly GitHub Digests to Telegram with Qwen via OpenRouter - n8n.io
One-click import, just add your credentials (GitHub PAT, Telegram bot token, OpenRouter API key), and you’re good.
I also wrote a detailed setup guide with full sample outputs and customization tips on my site - it covers how to filter repos, swap the LLM, change the schedule, add new commands, or switch to Discord/Slack:
Full Guide + Sample Reports - vinhautomation.com
Why I’m sharing this
I know the n8n community is full of people who build their own solutions rather than pay for yet another SaaS subscription. This workflow is my contribution to that spirit - something that took me a few hours to build but saves me time every single week.
It’s intentionally simple. Simple means it runs cheap, breaks rarely, and is easy to debug. But there’s plenty of room to extend it (multi-channel delivery, Obsidian sync, social auto-posting - I listed some ideas in the guide).
Questions for the community:
- Anyone else doing GitHub monitoring via n8n? How do you handle repo discovery at scale?
- Has anyone tried replacing the LLM with a local model via Ollama? Curious about the latency trade-off.
- What other data sources would you add to a unified weekly engineering report?
Would love to hear your thoughts and ideas! 
NEEDS ATTENTION section actually show regressions instead of just static red dots.