Show HN: Sift – only get pinged when something in your homelab matters (github.com)

Sift, a new AI-driven notification system for homelabs, aims to reduce the noise of unnecessary alerts, ensuring users only receive notifications that matter. By ingesting events from various servers, Sift employs a lightweight language model (LLM) to triage, categorize, and summarize these events, delivering urgent and actionable alerts directly to users' phones while routing routine issues into a daily digest. This approach significantly alleviates the common problem of alert fatigue faced by self-hosters, allowing critical notifications, such as server downtimes or backup failures, to stand out amidst the usual system chatter. Constructed with a focus on ease of use and low dependencies, Sift operates with just one small container and requires no extensive coding knowledge. It features a web console for remote diagnostics and can interact with servers through Telegram or several other platforms. The AI is optional, ensuring that even without a configured model, Sift can still manage notifications effectively using preset rules. This dual functionality makes it a versatile tool for homelab administrators who want to optimize their system monitoring while maintaining control over the type of alerts they receive.