Hugging Face's logo

Datasets:
openenvforge
/
3ambench

RL Environment
Tasks:
Other
Modalities:
Tabular
Text
Formats:
json
Languages:
English
Size:
< 1K
Tags:
harbor
openenv
agent
benchmark
dense-reward
prometheus
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
3ambench / NOTICE.md
openenvforge's picture
openenvforge
3amBench v0.1.0
f3b0a91 verified
preview code
|
Raw History Blame Contribute Delete
1.1 kB

NOTICE

3amBench / AlertForge is licensed under the Apache License 2.0 (see LICENSE).

Third-party material

  • Distractor alert rules (src/alertforge/distractors.py, and the rules/{node-exporter,kubernetes,blackbox, postgres,redis,kafka,nginx,jvm}.yml files in generated tasks) are written in the style of, and several are adapted from, samber/awesome-prometheus-alerts (CC BY 4.0) and kubernetes-monitoring/kubernetes-mixin (Apache-2.0). Changes: rule names kept, expressions simplified, for/thresholds adjusted, labels set to {severity, team}, runbook URLs rewritten to each task's fictional domain.
  • Tools used by the grader, shipped inside the task image (not in this repository): promtool from Prometheus v3.5.0 and amtool from Alertmanager v0.28.1 (Apache-2.0).
  • The multiwindow, multi-burn-rate recipe follows the Google SRE Workbook, chapter 5 "Alerting on SLOs".

All companies, services, teams and incidents in the tasks are fictional.