All runbooks
evidenceuapiredacted view

UAPI NUC Gemma/SLM Read-Only Preflight Evidence 2026-06-22

Read-only NUC preflight evidence for Gemma/SLM acceleration planning: hardware, storage, Python/tooling absence, WSL absence, and next gate requirement.

Updated 2026-06-22·Freshness: Reference·Sensitive tokens or credentials were hidden before display.

UAPI NUC Gemma/SLM Read-Only Preflight Evidence 2026-06-22

Read-only NUC preflight evidence for Gemma/SLM acceleration planning: hardware, storage, Python/tooling absence, WSL absence, and next gate requirement.

Promotion Metadata

  • Source feedback id: 109
  • Source feedback ids: 109
  • Promotion request id: 40
  • Feedback category: beneficial
  • Feedback source: [REDACTED:high-entropy]
  • Feedback created at: 2026-06-22T18:14:41.903521+00:00
  • Target slug: [REDACTED:high-entropy]
  • Review state: review
  • Reviewer note: AI lane auto-promoted redacted feedback into internal catalog context.

Source Feedback Body

Feedback 109

  • Product: uapi
  • Category: beneficial
  • Related slug: [REDACTED:high-entropy]
  • Source: [REDACTED:high-entropy]
  • Created at: 2026-06-22T18:14:41.903521+00:00

NUC Gemma/SLM Read-Only Preflight Evidence (2026-06-22)

Authority: acting-operator read-only preflight only. No installs, downloads, inference, benchmarks, secrets, live route changes, or mutations were performed.

Transport: Codex local shell -> SSH to VPS mail.united-ai.ca -> SSH reverse tunnel nuc-via-reverse -> Windows host NUC-IPEX.

Result

Preflight completed with usable evidence on the second pass. First pass exposed that wmic is not available on this Windows build, so the successful pass used PowerShell/CIM equivalents.

Verified Current Facts

  • Host: NUC-IPEX
  • OS: Microsoft Windows 11 Home, version 10.0.26200, build 26200
  • Model: ASUSTeK COMPUTER INC. NUC15CRSU9
  • CPU: Intel(R) Core(TM) Ultra 9 285H, 16 cores, 16 logical processors, 2900 MHz reported
  • RAM: 67987116032 bytes installed; 66393668 KiB visible; about 43578924 KiB free during preflight
  • GPU: Intel(R) Arc(TM) 140T GPU (32GB), driver 32.0.101.8331, status OK
  • NPU: Intel(R) AI Boost, class ComputeAccelerator, status Started/OK, driver oem161.inf
  • Boot/system NVMe: WD_BLACK SN8100 2000GB, NVMe SSD, healthy/OK
  • SN850X: not present on the NUC
  • C: free space: 704535498752 bytes
  • Python: 3.13.12
  • WSL: not installed
  • Idle snapshot: total CPU about 12.22% during the one-second sample

AI Runtime Package Probe

All target AI packages were absent from the default Python 3.13 environment:

  • torch
  • openvino
  • intel_extension_for_pytorch
  • ipex_llm
  • llama_cpp
  • transformers
  • optimum
  • nncf
  • openvino_genai

Gate Decision

The NUC has the expected CPU, Arc 140T GPU, NPU, memory, and storage profile for the next phase, but no Gemma/SLM route is production-ready. The next phase requires a separate install/download/benchmark gate before any package install, model weight download, inference, or benchmark execution.

Candidate routes remain candidate-only:

  • local.nuc.gemma.npu
  • local.nuc.gemma.140t
  • local.nuc.cpu.reserved

Do not claim routing performance until at least one small Gemma/SLM path is benchmarked on NPU and one path is benchmarked on Arc 140T with first-token latency, tokens/sec, load time, memory, stability, and driver issues recorded.