I Built a Self-Hosted AI Incident Diagnosis Tool That Only Returns a Root Cause When Multiple Diagnoses Agree

Short summary

Argus is an open-source, self-hosted AI incident diagnosis tool written in Go that runs five independent diagnoses against the same incident window and only returns a root cause when they converge. It ingests OTLP or Prometheus remote_write telemetry, runs as a single Docker binary, and uses an Anthropic API key for LLM reasoning. The author validated it on synthetic cases and real postmortems but seeks feedback on messy production telemetry.

• •Open-source Go tool runs five independent AI diagnoses and only reports root cause on consensus
• •Single binary ingests OTLP/Prometheus; telemetry never leaves the host
• •Validated on synthetic cases and reconstructed Cloudflare postmortems; seeking real-world testing

Generated with AI, which can make mistakes.