AI SRE Arena (Project Arena)
github.com/edgedelta/project-arena- Category
- Developer Tools
- Rank
- No. 1866Tools index
Previous survey · No. 1783 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Model & Agent Evaluation
- GitHub
- 13 stars
- Date
About
Project Arena is a vendor-neutral benchmark for testing AI investigation tools on Kubernetes incidents. It deploys a disposable Kubernetes fixture, injects faults, saves investigation records from any product, and scores completed investigations with a configurable AI judge across detection, diagnosis, and mitigation. This lets teams compare AI SRE products on the same scenarios.
At a glance
Project Arena creates disposable Kubernetes incidents and scores saved investigations. It is worth using if you need to compare how observability products or agents diagnose the same faults.
Good fit
SRE teams that can run controlled incident exercises and review a judge’s ratings against known facts.
Consider something else when
Use a production incident review if you need evidence about on-call behavior, real alert noise or operational impact.
Start here
Run one smoke scenario on a disposable cluster. Verify that the fault was injected, save each product’s investigation, and compare the judge’s ratings with the known incident facts.
Cost and requirements
MIT-licensed. The harness requires Python 3.10 or later and Kubernetes access. Local setup uses Docker and kind. Hosted clusters and a model judge can add costs.
Limits to know
Detection and diagnosis are separate outcomes. The published comparison uses different starting conditions for some products, so compare equivalent triggers and record which scenarios each product actually investigated.
Sources checked 2026-11-05. Source-based assessment.
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.