# Ops LLM

> Local LLM health checks and cache management. Probe Ollama/vLLM/SGLang endpoints, clean model caches.

- Skill: `majiayu000/ops-llm` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/ops-llm`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/ops-llm/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/ops-llm

---


# LLM Ops

Manage local LLM runtimes and caches.

## Commands

```bash
# Check all common LLM endpoints (Ollama, vLLM, SGLang)
./scripts/health.sh

# Check specific endpoint
./scripts/health.sh --target ollama:http://127.0.0.1:11434

# Continue even if some fail
./scripts/health.sh --warn-only

# Show cache sizes (dry-run)
./scripts/cache-clean.sh

# Actually clean caches
./scripts/cache-clean.sh --execute

# Clean additional path
./scripts/cache-clean.sh --path ~/.cache/torch --execute
```

## Default Endpoints Checked

- Ollama: `http://127.0.0.1:11434`
- vLLM: `http://127.0.0.1:8000`
- SGLang: `http://127.0.0.1:30000`

## Default Cache Directories

- `~/.cache/ollama`
- `~/.cache/huggingface`
- `~/.cache/vllm`

## Environment Variables

| Variable             | Default     | Description                  |
| -------------------- | ----------- | ---------------------------- |
| `LLM_HEALTH_TIMEOUT` | 2           | Seconds to wait per endpoint |
| `LLM_CACHE_DIRS`     | (see above) | Space-separated cache paths  |