Home

Argus wiki

Operational Prometheus / OpenTelemetry metrics for discord.py bots, in one line.

from argus import Argus
Argus(bot)

This wiki is the in-depth reference. The README is the quick start; everything below assumes you know discord.py and Prometheus and want the specifics.

Start here

New to Argus? Follow a tutorial end to end:

• Single bot - instrument one bot: metrics, dashboard, Grafana.
• Fleet at scale - the control plane across many processes and regions.
• OpenTelemetry (OTLP) - push metrics to an OTLP backend.
• Per-guild analytics - the ClickHouse analytical path.

Pages

• FAQ - quick answers, including how the dashboard auth token works.
• Architecture and Invariants - the three layers, the seven invariants, and why they exist.
• Configuration - every kwarg/env var, precedence, and resolution.
• Compatibility - discord.py 2.x support and the fork situation.
• Security - what is exposed, how to gate it, and safe deployment.
• Metrics Reference - every metric, its kind, labels, and source.
• Dashboard - the built-in SPA, the SSE snapshot protocol, auth.
• History and ClickHouse - the per-guild analytical path and the event sink.
• OTLP - the OpenTelemetry push adapter.
• Clustering - multi-process deployments and the cluster label.
• Fleet - the standalone control plane (Global/Fleet/Cluster view).
• Releasing - release-please, PyPI Trusted Publishing, the GHCR images.
• Contributing - dev setup, the checks, the DCO.

Mental model

core observes the bot and maintains a backend-neutral metric registry. adapters (Prometheus, OTLP) translate that registry into a backend. exposition serves /metrics and the dashboard on the bot's own aiohttp loop. history is a separate, optional path that drains per-guild events to an OLAP store. Data flows one way: bot -> core -> adapters/exposition; core never imports an adapter.