Introducing nature_whistle: Telemetry-driven alerting with recovery notifications
Hi everyone,
I’d like to share my first open-source Elixir package: nature_whistle.
The goal is simple: turn Telemetry events into actionable alerts, including recovery notifications when things return to normal.
Most monitoring tools do a good job of telling you when something breaks. But they often leave out an equally important part: knowing when things are healthy again without manually checking dashboards.
nature_whistle tries to solve that gap.
What it does
-
Listens to Telemetry events from the BEAM, Phoenix, Ecto, Oban, or custom metrics
-
Triggers alerts when thresholds are crossed
-
Sends a “calm” notification when metrics return to normal
-
Supports Slack, Microsoft Teams, generic webhooks, and a console notifier
-
Uses retries with exponential backoff for HTTP-based notifications
-
Lightweight design built on ETS + Telemetry (no heavy runtime overhead)
Example usage
Here’s a simple example of how a monitor might look:
# 👉 PUT THIS CODE IN THE config.exs of your application after adding NatureWhistle.Application as one of your application's children
config :nature_whistle,
monitors: [
%{
event: [:vm, :memory],
threshold: 80,
notifier: NatureWhistle.Notifiers.Slack,
message: "Memory usage is high!"
}
]
You can replace this with your real-world config showing:
-
Phoenix request latency
-
Ecto query time
-
Oban job failures
-
or any custom telemetry event
Why I built it
I originally got the idea after watching a talk by an Elixir developer talk on Telemetry in Elixir. It made me think about how much valuable operational data is already available in Telemetry events — and how easy it could be to build a small layer on top to turn them into meaningful alerts.
That exploration led to nature_whistle.
Design goals
-
Minimal footprint (ETS + Telemetry only)
-
Easy integration into existing apps
-
Extensible notifier system
-
Works with both infrastructure and application-level metrics
Links
GitHub:
Hex:
Questions for the community
-
What metrics do you currently alert on in your Elixir systems?
-
Do you currently use recovery notifications, or only failure alerts?
-
What notifier integrations would be most useful to you?
-
Would this fit alongside your existing monitoring setup, or replace part of it?
Feedback, criticism, and suggestions are very welcome.
