Skip to main content

Example Workspace Cache Service

Service​

  • Name: Local Model Cache
  • Service ID: local-model-cache
  • Class: infrastructure
  • Readiness: documented

Purpose​

Provide a shared local cache for model files so multiple projects do not redownload the same artifacts.

Lifecycle​

This service is filesystem-backed and does not run a persistent network listener.

cache-tool status --json
cache-tool cleanup --dry-run
cache-tool cleanup --max-size 500GB

Health Output​

{
  "service_id": "local-model-cache",
  "status": "degraded",
  "dependencies": [],
  "ports": [],
  "storage": [
    { "path": "D:\\.hf", "status": "available", "used_gb": 460, "limit_gb": 500 }
  ],
  "warnings": ["cache usage is above 90 percent"],
  "errors": [],
  "next_safe_action": "run cleanup dry-run and review candidates"
}

Ports​

No ports.

Paths​

  • Durable index: D:\015-DPW\local-model-cache\index
  • Cache: D:\.hf
  • Logs: D:\015-DPW\local-model-cache\logs

Resource Bounds​

  • Cache warning threshold: 450 GB.
  • Cache hard limit: 500 GB unless operator expands it.
  • Cleanup must support dry-run.
  • Logs rotate at 5 MB per file and retain 20 files.

Agent Rules​

  • Agents may run status and cleanup dry-run commands.
  • Agents may not delete cache entries without operator approval unless a project-specific runbook grants permission.