noonghunna/club-3090

v0.8.2 — Universal pull: honest failure on-ramp + a wider model catalogue

What this gives you

• An honest "will it fit?" answer for any safetensors model — scripts/pull.sh <model> --profile-like <variant> --recommend returns a plain verdict (FITS / FITS-but-needs-acceptance / DOES-NOT-FIT) that tracks the real gate, carries the boot-fit≠runtime caveat, and never silently passes.
• Failed pulls become fixable signal, not dead-ends — every hard-block leaves a redacted, path-scrubbed diagnostic; one consented command (scripts/pull.sh --submit-last, works with or without gh) sends it back so the gap can be closed. No telemetry, no auto-send.
• ~12 more model architectures recognised out-of-the-box — Phi, Gemma/Gemma3, Starcoder2, Cohere/Command-R, Mixtral, Qwen2/3-MoE and more no longer need --experimental-arch; genuine native models reach a clean serve verdict instead of a blanket refusal (per-repo remote-code still fail-closed — zero false-pass).
• Optional non-NVIDIA hardware detection for the eval path (AMD ROCm / Apple), graceful-degrading — no change for NVIDIA users.
• Bundled: N-GPU NVLink auto-detection (multi-4 + Gemma-4-26B dual composes) and a documentation restructure (quick-start README, GETTING_STARTED.md, FAQ).

Know before you use it

• Generated composes are a minimal, known-safe starting point: capacity (context/KV/mem-util) is the reference profile's, not auto-tuned to your GPU — run --recommend (or tools/kv-calc.py --solve-max-ctx) for the real fit and tune ${MAX_MODEL_LEN}. An opt-in capacity optimiser is planned for a later release.
• Advanced features (MTP / TurboQuant / DFlash) stay in the pre-baked Genesis composes — generated composes are non-Genesis by design.
• GGUF / cross-engine generation is deferred to a separate §9 design-unlock — not in this release.

v0.8.2 — 2026-05-19

✨ Features

• feat(pull): v0.8.2 STEP V5 — recommend UX + report-a-failed-pull doc + §9-reconciliation (c5b5e9b)
• feat(nvlink): auto-detect NVLink on N-GPU topologies; add detection to multi4 + gemma-4-26b dual (8f8ec1c)
• feat(pull): v0.8.2 STEP V4 — optional whichllm hw-detect subprocess (CONTRACT-3, hw-detect-only) (3917728)
• feat(switch): v0.8.2 STEP V3 — switch.sh ↔ compose_registry parity (CONTRACT-2b-ii) (e6503bc)
• feat(pull): v0.8.2 STEP V3 — arch-registry expansion + chat-template attribution/drift_guard (999c93f)
• feat(pull): v0.8.2 STEP V2 — surface pointer + --submit-last/--submit (gh + gh-less, consented, F5 reuse) (e1cdcb5)
• feat(pull): v0.8.2 STEP V1 — capture-on-hard-block pt1-gate emitter + BaseCaptureBundle protocol lift (20f1557)
• feat(report): lspci PCIe/P2P diagnostics subsection (LnkSta/ACS/topology) (#148) (af2e45a)

🐛 Bug fixes

• fix(pull): v0.8.2 STEP V5 — recommend must not label a fits-clean model "DOES NOT FIT" (26949d7)
• fix(pull): v0.8.2 STEP V3 — deliver CONTRACT-2's engine-supported broadening (TRC two-class) (d78b9a9)
• fix(pull): v0.8.2 STEP V2 — gh-less issue body must not carry the absolute capture path (52451ca)
• fix(launch): force LC_NUMERIC=C so the VRAM-budget printf survives comma-decimal locales (#159) (186dc93)
• fix(deriver): correct stale "GGUF not supported until v0.8.1" message — now misleading post-v0.8.1-ship (344ab87)

📝 Documentation

• docs(architecture): bring current-state docs up to v0.8.2 (recommend / submit on-ramp / arch-registry / hwdetect) (c5c8f46)
• docs(generator): state plainly that generated-compose capacity is the reference profile's, NOT fit-adapted (247b1dc)
• docs(pull): v0.8.2 STEP V6 — correct §9/headline to the true bundled release scope (b791271)
• docs: fix duplicate MULTI_CARD.md entry in docs index (966a8d1)
• docs: reorder docsindex (GSD first), add FAQ TOC + promote troubleshooting ladder, add tool-calling example (a891b39)
• docs: add GETTING_STARTED.md, Gemma 4 model READMEs, restructure main README with quick start first (6368bae)
• docs: fix stale NVLINK_MODE comment, INTERNALS.md cliff status, and dead companion repo link (28bd0e8)
• docs(container-runtimes): Proxmox passthrough — NVLink is the fragile path, not Proxmox (#161) (3f066a0)
• docs(benchmarks): add @hlo-world dual-3090 PCIe x4 dual-dflash-noviz row (#158) (135f2c4)
• docs(upstream): froggeric v19 re-eval PASSED — ADOPTED (#150) (ec1fd65)

🧹 Maintenance

• chore(chat-template): re-vendor latest froggeric Qwen3.6 template for re-eval (#150) (8a9ea6c)

[Pin: git checkout v0.8.2] · Full diff