SoTA@Home Agent Skill

Submit a generation to the orchestrator and retrieve the winning train.py.

Updated Mar 14, 2026

Search Tags:#kubernetes #openai #orchestrator #submission #minio #autoresearch #train.py

Ready to run

npx skills add https://github.com/Sprit3Dan/sotaathome --skill sota-home-agent-skill

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: SoTA@Home Agent Skill
Source: https://github.com/Sprit3Dan/sotaathome/tree/main
Command: npx skills add https://github.com/Sprit3Dan/sotaathome --skill sota-home-agent-skill

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires bash, curl, jq, and includes scripts (resource) components.

What problem does it solve?

SoTA@Home Agent Skill enables external agents to submit autoresearch jobs to a central orchestrator and retrieve the resulting train.py from the best run.

Core Features & Use Cases

HTTP endpoints for job submission, cluster status, and per-task status, enabling automation and monitoring.
End-to-end multi-generation workflow: enqueue, orchestrate, evaluate, promote, and re-submit next generations.
Retrieval of final train.py for reproducibility and deployment in downstream pipelines; supports custom agent scripts uploaded to S3.

Quick Start

Submit a generation via the orchestrator and then retrieve the winning train.py upon completion.

Frequently Asked Questions about SoTA@Home Agent Skill

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I submit autoresearch jobs to a Kubernetes cluster orchestrator?▼

You can submit autoresearch jobs by sending HTTP requests to the FastAPI orchestrator's job submission endpoints. The workflow then orchestrates evaluation, promotes the best run, and re-submits subsequent generations automatically.

What infrastructure do I need to run an autoresearch workflow with MinIO and OpenAI?▼

Running an autoresearch workflow requires a Kubernetes cluster, a FastAPI orchestrator, MinIO or S3-compatible storage for custom agent scripts, and OpenAI API access for init-container spec generation.

How do I retrieve train.py from a completed evaluation run?▼

You retrieve train.py from the winning run by querying the orchestrator's per-task status endpoints after evaluation completes. This provides the final script for reproducibility and downstream deployment pipelines.

Can I monitor Kubernetes cluster status during an autoresearch generation?▼

Yes, you can monitor Kubernetes cluster status during an autoresearch generation by polling the orchestrator's dedicated HTTP endpoints. This enables continuous automation and tracking of the multi-generation evaluation workflow.

Do I need bash and curl to automate the autoresearch orchestration workflow?▼

Yes, automating the autoresearch orchestration workflow requires bash, curl, and jq dependencies to interact with the FastAPI HTTP endpoints and parse cluster status JSON responses effectively.