noonghunna/club-3090

v0.8.1 — fix / docs-fidelity release. A focused patch stack shipped deliberately before any v0.8.x feature work: pull argparse usage errors now exit 64 not 2 (#370), the missing Gemma-4-31b vllm-pr41800 overlay vendored + registered (#153 / #155), the newer-driver 3090 MAX_MODEL_LEN=105000 HARDWARE anchor (#149), the EXAMPLES.md thinking-default correction (#372), plus the v0.8.0 docs-fidelity cleanup. No model-behavior change; all gated by the full pre-tag suite at the tag commit.

Post-tag follow-ups — NOT in the v0.8.1 tree (transparency note).
These landed on master after this tag and ship in the next release; they are listed here only so a v0.8.1-pinned reader isn't surprised. They are not part of the v0.8.1 release tree:

• #156 — report.sh lspci PCIe/P2P diagnostics (LnkSta / ACS / topology) for #137/#351 (closes #148)
• #157 — froggeric chat-template v19 adopted (closes #150): +10pp hermesagent-20 (50→60%), 7 packs flat, streaming/​soak clean, TPS-neutral

If you track master you already have both. If you pin the v0.8.1 tag or GHCR image, you get them at the next release. The CHANGELOG/Bug-fix list below is auto-generated and correctly reflects only the v0.8.1 tree.

v0.8.1 — 2026-05-17

🐛 Bug fixes

• fix(patch-attribution): register vendored gemma-4-31b pr41800 overlay (follow-up to #153/#154) (b4b20ff)
• fix(gemma-4-31b): vendor missing vllm-pr41800 overlay into the model tree (closes #153) (9c79192)
• fix(pull): argparse usage errors exit 64, not 2 — distinguishable from honest hard-stop (#370) (820eb38)

📝 Documentation

• docs(examples): correct "thinking on by default" — shipped composes set enable_thinking=false (#372) (46bb271)
• docs(hardware): newer-driver 3090 caps long-text.yml at MAX_MODEL_LEN=105000 (#149) (b0774f9)
• docs: fix v0.8.0 docs-fidelity gaps (trc-ack first-run heads-up, exit-code honesty, GGUF message claim) (78a7dee)
• docs: cross-link the v0.8.0 universal pull flow from the existing user guides (afe56f7)

[Pin: git checkout v0.8.1] · Full diff