spark-vllm-docker AGENTS.md

spark-vllm-docker AGENTS.md is an instructions file for Codex, OpenCode from eugr/spark-vllm-docker. It costs 346 tokens per session, scanned A, original, MIT.

Repository instructions for scripts that prepare and run vLLM, a system for serving language models, on NVIDIA DGX Spark computers. They separate development tasks from actually launching services or downloading models.

In plain words
What is it for?
Use it when changing or testing the project's Bash, Python, Docker, recipe, or documentation files, or when preparing and launching vLLM services.
Why use it?
They help a coding agent choose the right guide and avoid changing computers, containers, or clusters without first checking what will happen.

Instructions file for CodexOpenCode

About the project

eugr/spark-vllm-docker is a Docker setup for running vLLM, an inference server for serving AI models, on DGX Spark computers. It supports single machines and multi-node clusters for model inference, with startup scripts and configuration for distributed hardware. The catalogue instruction provides an operational runbook for preparing hosts and running repository recipes.

eugr/spark-vllm-docker · 2,226 stars · on GitHub

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/eugr/spark-vllm-docker/agents-md
Clone the repo
git clone --depth 1 https://github.com/eugr/spark-vllm-docker

Made for: Codex, OpenCode.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for spark-vllm-docker AGENTS.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/eugr/spark-vllm-docker/agents-md.svg)](https://agentmods.dev/instructions/eugr/spark-vllm-docker/agents-md)
Your own site
<a href="https://agentmods.dev/instructions/eugr/spark-vllm-docker/agents-md"><img src="https://agentmods.dev/badge/instructions/eugr/spark-vllm-docker/agents-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 346 This file is loaded in full into every session.
When invoked 346 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00346 $0.00346
Opus 5 $0.00173 $0.00173
Sonnet 5 $0.00069 $0.00069
Haiku 4.5 $0.00035 $0.00035

Measured 5d ago against content hash 3722699d489b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

spark-vllm-docker AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

AGENTS.md · 36 lines

What it actually says

Agent Instructions

These instructions apply to the entire repository.

Repository

This project provides Bash and Python orchestration for running vLLM on one or more NVIDIA DGX Spark systems. Work from the repository root and read README.md for the public project overview.

Choose One Guide

  • Use or operate the repository: For host preparation, recipe selection, cluster discovery, image or model setup, recipe launches, and live-server verification, follow docs/AGENT_RUNBOOK.md.
  • Develop the repository: For inspection, fixes, features, reviews, tests, or changes to scripts, recipes, mods, Dockerfiles, and documentation, follow docs/AGENT_DEVELOPMENT.md.
  • Both: Follow the development guide first. Use the operational runbook afterward only when the user also requested a real build, download, or launch.

Read only the guide relevant to the task unless the work crosses that boundary. A recipe --dry-run used to validate generated commands is development. A non-dry recipe run, --setup, discovery, image preparation, model download, or container launch is operation.

Common Boundaries

  • Inspect before changing repository, host, container, or cluster state.
  • Preserve unrelated user changes and existing local configuration or artifacts.
  • Do not expose credentials or .env contents in chat, logs, diffs, or commands.
  • Operational tasks do not authorize source changes. Development tasks do not authorize real deployments. Perform both only when the user requests both.
  • Do not prune, overwrite, stop, remove, or force-refresh existing resources unless the requested task requires it.
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 5d ago First seen · 36 lines · 346 tokens per session scan A 3722699d489b

Subscribe to this mod's changes

spark-vllm-docker AGENTS.md is an instructions file published in the GitHub repository eugr/spark-vllm-docker (2,226 stars, last pushed today), licensed MIT. It adds 346 tokens to every session, about $0.0017 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.