← all publishers

helixml

@helixml source repo

5 published skills

  1. Helix Org CLI · helixml
    Use the Helix CLI to operate helix-org (bots, topics, processors, chat) and raw REST via helix api. Use when listing/managing org bots, chatting with chief-of-staff or other bots, or calling any authenticated Helix API path.
    0
    installs
  2. Deploy Ramjet · helixml bundle
    Deploy or upgrade ramjet and OpenAI-compatible GPU engines on a Linux node with Docker Compose. Use for first-time node setup, creating a Compose stack, validating GPU/container prerequisites, promoting an immutable image, performing an LB-only rollout, or planning a rollback.
    0
    installs
  3. Optimize Ramjet Node · helixml bundle
    Measure and tune a ramjet GPU inference node using controlled, reproducible experiments. Use for GPU/NUMA placement, vLLM or DSpark scheduler settings, routing weights, cache locality, throughput, latency, concurrency, context-size, or image A/B optimization.
    0
    installs
  4. Load Test Ramjet Node · helixml bundle
    Execute guarded, reproducible load tests against a ramjet GPU inference deployment, especially the node06 8x RTX PRO 6000 stack. Use when asked to benchmark TPS, TTFT, TPOT, concurrency, long context, cache locality, agent/tool correctness, full-box capacity, or an engine candidate under load, and when comparing two serving configurations without disrupting production.
    0
    installs
  5. Troubleshoot Ramjet Node · helixml bundle
    Diagnose a ramjet GPU inference node by checking NVIDIA GPUs, Docker Compose services, OpenAI endpoints, router health, Prometheus metrics, logs, networking, and basic synthetic requests. Use for outages, degraded health, low throughput, cache misses, stuck requests, GPU errors, or post-deploy verification.
    0
    installs