{
  "observed_at": "2026-11-03T23:21:56.226349+00:00",
  "source": "Actual local runner checkpoint plus transient systemd unit properties; not controller token telemetry or independent evaluation",
  "model": "Qwen/Qwen2.5-0.5B-Instruct-GGUF",
  "model_revision": "9217f5db79a29953eb74d5343926648285ec7e67",
  "model_file": "qwen2.5-0.5b-instruct-q4_k_m.gguf",
  "model_sha256": "74a4da8c9fdbcd15bd1f6d01d621410d31c6fc00986f5eb687824e7b93d7a9db",
  "model_license": "Apache-2.0; upstream LICENSE retained privately with downloaded weights",
  "model_source": "https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF/tree/9217f5db79a29953eb74d5343926648285ec7e67",
  "server": "llama.cpp b12081 / 0.8.0-dev / 8fc037672",
  "server_archive_sha256": "2a773dcda8b301e3650e5832d3e7b014afcdffdb6364d4d4e5cf3a12bcbc2eeb",
  "server_archive_digest_verified": true,
  "server_license": "MIT; upstream notice retained privately with downloaded binary",
  "server_source": "https://github.com/ggml-org/llama.cpp/releases/tag/b12081",
  "context_tokens": 1024,
  "threads": 1,
  "parallel_requests": 1,
  "batch_size": 32,
  "gpu_layers": 0,
  "unit_before_stop": {
    "ActiveState": "active",
    "RuntimeMaxUSec": "15min",
    "Result": "success",
    "MemoryPeak": "576987136",
    "CPUUsageNSec": "27403851000",
    "CPUQuotaPerSecUSec": "600ms",
    "MemoryHigh": "576716800",
    "MemoryMax": "629145600"
  },
  "service_stopped_at": "2026-11-03T23:20:20Z",
  "service_scope": "Transient localhost-only model server, resource-limited independently of agent cycle; stopped after six requests. Not a public inference API. MemoryPeak is cgroup memory including cache, not process RSS.",
  "completed_requests": 6,
  "provider_reported_tokens": {
    "prompt_tokens": 867,
    "completion_tokens": 335,
    "total_tokens": 1202
  },
  "summed_request_duration_ms": 43058,
  "generation_settings": {
    "endpoint": "http://127.0.0.1:8094/v1/chat/completions",
    "model": "qwen2.5-0.5b-instruct-q4_k_m",
    "maxTokens": 160,
    "temperature": 0,
    "seed": 42,
    "timeoutSeconds": 120
  },
  "resumption": "Two requests, then four remaining requests. A final resume after stopping the server sent zero requests and left checkpoint bytes identical.",
  "costs": "No paid model API endpoint used. Existing host/subscription operating allocations are unchanged; incremental infrastructure cost is not measured. Token usage is local provider-reported, not a bill or an agent-token count."
}