feat: add Ling-3.0-flash vLLM service for Dankclaw (frank-default-large, port 7770)

- serve-ling3-vllm.sh: wrapper around jcbtc validated run-256k.sh (TRITON_MLA,
  MTP K1, 256k, 6 slots, ROCr idle-fix LD_PRELOAD); sets HOST/PORT/
  SERVED_MODEL_NAME=frank-default-large
- frank-default-large.service: systemd unit, User=dank, port 7770

scripts/frank-default-large.service

@@ -0,0 +1,18 @@
+[Unit]
+Description=vLLM Ling-3.0-flash INT4 (CIRU gfx1151) -- frank-default-large
+After=network.target
+
+[Service]
+Type=simple
+User=dank
+Group=dank
+WorkingDirectory=/home/dank/ling3-ciru/Ling-3.0-Flash-CIRU-int4-Strix-native
+Environment="HOST=0.0.0.0"
+Environment="PORT=7770"
+Environment="SERVED_MODEL_NAME=frank-default-large"
+ExecStart=/home/dank/vllmScripts/scripts/serve-ling3-vllm.sh
+Restart=on-failure
+RestartSec=10
+
+[Install]
+WantedBy=multi-user.target

scripts/serve-ling3-vllm.sh

@@ -0,0 +1,27 @@
+#!/usr/bin/env bash
+# serve-ling3-vllm.sh — Launch Ling-3.0-flash CIRU INT4 vLLM (gfx1151) on Dankclaw
+#
+# This wrapper sources the jcbtc ling3-runtime.env (sets VENV and MODEL_PATH),
+# applies our production settings (host, port, served-model-name), and hands
+# off to the validated run-256k.sh launcher (TRITON_MLA, MTP K1, 256k, 6 slots,
+# ROCr idle-fix LD_PRELOAD). Do not hand-roll the vllm serve command; the
+# launcher carries required env vars and the ROCr override.
+set -Eeuo pipefail
+
+PKG="/home/dank/ling3-ciru/Ling-3.0-Flash-CIRU-int4-Strix-native"
+
+if [ ! -f "$PKG/ling3-runtime.env" ]; then
+  echo "CIRU package runtime environment not found: $PKG/ling3-runtime.env" >&2
+  exit 1
+fi
+
+set -a
+# shellcheck disable=SC1091
+source "$PKG/ling3-runtime.env"
+set +a
+
+export HOST="${HOST:-0.0.0.0}"
+export PORT="${PORT:-7770}"
+export SERVED_MODEL_NAME="${SERVED_MODEL_NAME:-frank-default-large}"
+
+exec bash "$PKG/scripts/run-256k.sh"