feat(frank-large): unsloth llama.cpp Qwen3.8-Flash-Next UD-IQ4_XS (200k/1 slot, MTP depth 2)

Replace the CIRU v4.4.1 Orca runtime with the unsloth llama.cpp prebuilt
(b11030-mix-5ff778e, ROCm gfx1151, bundles its own ROCm) serving
unsloth/Qwen3.8-Flash-Next-GGUF:UD-IQ4_XS (87.3 GiB, three HF split parts
concatenated) on the Strix Halo iGPU. One conversation, 200,000 ctx,
PARALLEL_SLOTS=1, MTP depth 2 via the explicit -md shared-Q8_0 head
(sidecar auto-discovery does not search the mtp/ subfolder), text-only.
Fleet contract unchanged: port 7770, alias frank-large.

- serve-frank-large.sh: unsloth binary, -m/-md, --spec-type draft-mtp
  --spec-draft-n-max 2, --jinja, preserve_thinking; gates on binary,
  GGUF, MTP head; conditional --mmproj on ENABLE_VISION=1
- download-frank-large-unsloth.sh: resumable fetch of 3 parts + MTP head
  + mmproj, each verified against its LFS sha256 fetched live from the
  Hub tree API; concatenates parts, records/verifies SHA256SUMS
- deploy-remote.sh: LLM_LARGE_MODEL/LLM_LARGE_DRAFT paths, pkill pattern
  on the model path (not the binary name), wait_port_free gate
- tests: seams for 200k/1-slot, explicit -md, text-only, handler gates,
  sweep safety

scripts/frank-large.env

@@ -1,44 +1,39 @@
-# frank-large — CIRU v4.4.1 (QSA window hotfix) serving
-# jcbtc/Qwen3.8-Flash-CIRU-STRIX-Orca on the Strix Halo iGPU (gfx1151).
+# frank-large — unsloth llama.cpp prebuilt (b11030-mix-5ff778e, ROCm
+# gfx1151) serving unsloth/Qwen3.8-Flash-Next-GGUF:UD-IQ4_XS on the Strix
+# Halo iGPU.
 #
-# These are the launcher tunables read by the runtime's
-# scripts/ciru/run-orca-server.sh (see the serve script). The 36-variable
-# kernel bundle and the PLE pager defaults live in the runtime's own
-# profiles/strix-halo-production.env, which the launcher sources AFTER this
-# file; every profile line is a ${VAR:-default}, so a value set here wins.
+# These are the tunables read by scripts/serve-frank-large.sh. The flag
+# list is owned by the serve script; this file only sets fleet values.
 #
-# DO NOT name a new tunable TEMP, HOST or GPU_MAX_HEAP_SIZE: the HIP runtime
-# reads those bare (the 2026-08-13 TEMP incident). HOST is the one exception
-# because the launcher itself consumes it for --host; its value doubles as a
-# JIT-cache directory name, which is why 0.0.0.0 is safe.
+# DO NOT name a new tunable TEMP, HOST or GPU_MAX_HEAP_SIZE: the HIP
+# runtime reads those bare (the 2026-08-13 TEMP incident). HOST is the one
+# exception because the serve script itself consumes it for --host.
 
-# The assembled payload (scripts/download-frank-large-orca.sh): the launcher
-# refuses to start without the main GGUF, the MTP head and the PLE sidecar.
-MODEL_DIR=/home/dank/models/Qwen3.8-Flash-CIRU-STRIX-Orca
+# The assembled payload (scripts/download-frank-large-unsloth.sh): the
+# serve script refuses to start without the main GGUF and the MTP head.
+MODEL_DIR=/home/dank/models/Qwen3.8-Flash-Next-UD-IQ4_XS
 
-# Fleet endpoint: the same port and alias the halogen predecessor served, so
+# The prebuilt runtime (see the serve script for the fetch line).
+RUNTIME_DIR=/home/dank/GitPackages/unsloth-llama-b11030
+
+# Fleet endpoint: the same port and alias the CIRU predecessor served, so
 # no client, probe or models.yml entry changes address.
 HOST=0.0.0.0
 PORT=7770
 
-# One stream, full context (operator call, 2026-09-19). MTP requires exactly
-# one slot in this runtime, so the drafter's constraint and the operator's
-# ask are the same number. 262144 is the model's trained window; the GGUF
-# carries no rope scaling.
-CONTEXT_SIZE=262144
+# One stream, 200k context (operator's call, 2026-09-21). The model's
+# trained window is 262144; 200000 is the requested serving shape.
+CONTEXT_SIZE=200000
 PARALLEL_SLOTS=1
 
-# MTP head: on. Depth 4 is the Orca variant's own default (mtp_depth_default
-# in the launcher); the sidecar is the Q8_0 MTP file from the model repo.
+# MTP shared-Q8_0 head: on. Draft depth 2 is the model card's default
+# ("higher drafts more but each guess is accepted less often"). The shared
+# head borrows the token embedding and output projection from the main
+# model, so the file is 2.6 GB rather than 4.1.
 ENABLE_MTP=1
+SPEC_DRAFT_N_MAX=2
 
-# Text-only for now: the F16 projector is on disk at
-# vision/mmproj-Qwen3.8-Flash-Orca-F16.mmproj; flip to 1 to enable it. The
-# fleet's image path is frank-small either way, so models.yml stays
-# input: [text] until this changes.
+# Text-only for now: the F16 projector is on disk at vision/mmproj-F16.gguf;
+# flip to 1 to enable it. The fleet's image path is frank-small either way,
+# so models.yml stays input: [text] until this changes.
 ENABLE_VISION=0
-
-# PLE pager: the 52.4 GiB payload is mmap'd and paged by the engine's own
-# workers; this is the hot window kept resident. The vendor's production
-# profile default is 4096.
-PLE_CACHE_MIB=4096