Vibeleaderboard
Index / tool

AI SRE Arena (Project Arena)

github.com/edgedelta/project-arena
Visit github.com
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.

Tags

benchmarkkubernetessreai-agentsevaluationincident-response

Tech Stack

Python

Comments (0)

No comments yet

Editorially curated, with community endorsements as a secondary signal. Corrections welcome.