eloplus-agent 1.0.0

EloPlus on-premise agent — automatic AI workload demand response

EloPlus Agent

Automatically pause flexible AI training jobs during grid price spikes.
Connects your infrastructure to EloPlus in under 10 minutes.

Requirements

• Python 3.9+
• Linux or macOS
• Network access to your scheduler (kubectl, ray, scontrol, or shell)

Quick Install

curl -sSL https://eloplus.ai/install.sh | bash -s -- --token YOUR_TOKEN

Manual Install

pip install eloplus-agent
eloplus init --token YOUR_TOKEN
eloplus start

Non-interactive / scripted setup

pip install eloplus-agent
eloplus init \
  --token YOUR_TOKEN \
  --scheduler kubernetes \
  --mode automatic \
  --non-interactive
eloplus start

Commands

Command	Description
eloplus status	Current agent and grid state
eloplus test	Verify scheduler integration
eloplus test --dry-run	CLI smoke test (no API/scheduler)
eloplus logs	Recent activity log
eloplus upgrade	Update to latest PyPI release
eloplus stop	Stop the agent daemon

Scheduler Support

Scheduler	Integration
kubernetes	Pods labeled eloplus=flexible
ray	Jobs tagged eloplus:flexible
slurm	Jobs with flexible in the name
shell	Processes defined in config

Install optional deps when needed:

pip install "eloplus-agent[kubernetes]"
pip install "eloplus-agent[ray]"

Modes

• automatic — pauses/resumes without intervention
• semi-auto — asks for confirmation via Slack before acting
• manual — alerts only; you execute pause/resume

Version checks

On every eloplus start, the agent checks GET /v1/agent/version:

• Below minimum → refuses to start; run pip install --upgrade eloplus-agent or eloplus upgrade
• Below latest → prints an update notice (no auto-upgrade without your consent)

Support

[e-mail]