alerting-stack

What problem does it solve?

This Skill eliminates “set up alerts but don’t get woken up or understand what happened” by providing a lightweight, single-channel alerting stack tailored for a self-hosted Walter-VM environment.

Core Features & Use Cases

• Telegram-first, single channel notification: routes multiple alert tiers into one pinned chat to keep alert handling cognitively cheap.
• Four-tier coverage for realistic failure modes: combines uptime checks (Uptime-Kuma), VM resource watchdogs (cron script with debouncing), LLM spend/budget triggers (LiteLLM → n8n → Telegram), and Hetzner cost trend monitoring (hcloud CLI → cron → Telegram).
• Debounced, transition-based alerts: sends messages on threshold transitions to prevent notification fatigue and alert flapping.
• Cost and budget awareness: alerts on spend thresholds, unexpected usage spikes, and projected month-end Hetzner cost drift.

Quick Start

Configure your Telegram bot token and chat ID in WALTER_TELEGRAM_BOT_TOKEN and WALTER_TELEGRAM_CHAT_ID, then add the Tier 1 Uptime-Kuma monitors and deploy the watchdog and Hetzner spend scripts using the provided locations and cron examples.