docs(openwebui): frank-dragoon enabled in the picker beside frank-zealot

Live DB trio applied on Dankclaw (enable idx 0, model row + global access_grant, restart); verified /api/models as admin and role=user, chat answers pong. Battery rewritten to gate on /api/models — /api/v1/models/list lists only workspace models on 0.6.x.
2026-10-09 20:43:28 -07:00
505 changed files with 32485 additions and 39854 deletions
@@ -0,0 +1,71 @@
+# Tunables for frank-default-small (Tiel-Coder-35B-A3B UD-Q6_K_XL,
+# qwen35moe hybrid MoE, Dankserver, RTX 3090).
+#
+# Runtime: upstream llama.cpp (latest master, CUDA build with
+# -DGGML_CUDA_FA_ALL_QUANTS=ON -DCMAKE_CUDA_ARCHITECTURES=86) on RTX 3090
+# 24 GB; Ryzen 7 7700X; 62 GB RAM.  Replaced the ik_llama.cpp Ornith
+# CRACK Q8_0 deployment on 2026-09-01.
+#
+# Model: peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF, UD-Q6_K_XL quant
+# (30 GB).  Architecture: qwen35moe — 40 layers, full_attention_interval=4
+# (10 attention + 30 SSM/Mamba), 256 experts / 8 used, 2 KV heads × 256
+# dim, embedding 2048, rope dimension_count 64, rope freq_base 10 000 000.
+# Text-only (no mmproj).  Built-in chat template in the GGUF.
+#
+# Expert offload: --cpu-moe keeps all 256 expert weights in CPU RAM.
+# Non-expert weights (~1.6 GB) fit entirely on the 24 GB card.
+#
+# KV cache: Q8_0 for both K and V.  With CTX_SIZE=200700 (66 900 per slot
+# × 3 parallel) and 10 attention layers, the KV cache is ~2 GB — the
+# requested 2 GB VRAM buffer.
+#
+# WARNING: do not name any variable TEMP, HOST, or GPU_MAX_HEAP_SIZE. The
+# ROCm HIP runtime reads those bare env vars (see the 2026-08-13 TEMP
+# incident). This box is NVIDIA/CUDA, but the rule keeps the env portable.
+
+# Model file (UD-Q6_K_XL, 30 GB, text-only MoE hybrid).
+MODEL_DIR=/home/dank/models/tiel-coder-35b
+MODEL_FILE=Tiel-Coder-35B-A3B-UD-Q6_K_XL.gguf
+
+# Upstream llama.cpp llama-server (CUDA build, latest master).
+# Built with -DGGML_CUDA_FA_ALL_QUANTS=ON -DCMAKE_CUDA_ARCHITECTURES=86
+# using the staged CUDA 13.2 toolkit at /home/dank/cuda13-stage.
+LLAMA_SERVER=/home/dank/llama.cpp/build/bin/llama-server
+
+ALIAS=frank-default-small
+BIND_ADDR=0.0.0.0
+LISTEN_PORT=7770
+
+# Total context budget across all parallel slots.  200700 / 3 = 66 900
+# per slot.  The model's context_length is 262 144, so 66 900 is well
+# within range.
+CTX_SIZE=200700
+PARALLEL=3
+
+# Expert offload: all MoE expert weights pinned to CPU RAM.
+CPU_MOE=1
+
+# All non-expert layers on GPU (auto-fit will determine the exact count;
+# with --cpu-moe the non-expert footprint is ~1.6 GB, leaving ample room).
+N_GPU_LAYERS=999
+
+# KV cache: Q8_0 for both K and V.  ~2 GB total for 200700 ctx × 3 slots
+# × 10 attention layers × (2 KV heads × 256 dim × 1 byte) × 2 (K+V).
+CACHE_TYPE_K=q8_0
+CACHE_TYPE_V=q8_0
+
+VERBOSITY=2
+
+# Prefill batch sizing.
+BATCH_SIZE=4096
+UBATCH_SIZE=1024
+
+# Sampling.
+TEMPERATURE=1.0
+TOP_K=20
+TOP_P=0.95
+MIN_P=0.0
+
+# Compute threads lowered so other processes keep a core.
+THREADS=7
+THREADS_HTTP=8
@@ -0,0 +1,16 @@
+[Unit]
+Description=llama.cpp Tiel-Coder-35B-A3B Q6_K_XL qwen35moe hybrid MoE (dankserver, frank-default-small)
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+User=dank
+Group=dank
+Environment="FRANK_SMALL_ENV_FILE=/home/dank/serverAppConfigs/scripts/frank-default-small.env"
+ExecStart=/home/dank/serverAppConfigs/scripts/serve-frank-small.sh
+Restart=on-failure
+RestartSec=10
+
+[Install]
+WantedBy=multi-user.target
@@ -0,0 +1,126 @@
+# Tunables for frank-large (kingjones777/Qwen3.8-Flash-Next-ROCmFP4-
+# STRIX-GGUF on the kingjones30/ROCmFPX llama.cpp runtime, Dankclaw,
+# Strix Halo gfx1151, unified memory).
+#
+# Runtime: kingjones30/ROCmFPX (fork of charlie12345/ROCmFPX with the
+# qwen4exp-on-rocmfpx-d3ca537 patch applied — the model card's pinned
+# runtime: ROCmFP4 tensor types AND the qwen4exp architecture in one
+# tree; stock llama.cpp and charlie's fork alone cannot load this GGUF).
+# Built per the card recipe:
+#   cmake -B build -DGGML_HIP=ON -DGPU_TARGETS=gfx1151 -DGGML_NATIVE=ON -DCMAKE_BUILD_TYPE=Release
+# at /home/dank/GitPackages/kingjones-ROCmFPX/build.
+# Replaced the agentionai ROCmFP4-FAST imatrix deployment (charlie's
+# build-qwen38 runtime) on 2026-08-30.
+#
+# Model: Qwen3.8-Flash-Next base (qwen4exp MoE: 512 experts / 10 used,
+# 48 layers = 12 attention + 36 SSM, QSA indexer, hyper-connections, PLE
+# at layer 1), Q4_0_ROCMFP4_STRIX (file_type 105), 113.47 GiB single
+# file. STRIX is the top quality tier: PLE table at Q8_0 (50.66 GiB),
+# token embeddings + lm head at Q6_K. The PLE table NEVER enters GPU
+# memory — it streams from the file through the OS page cache. Two hard
+# card constraints:
+#  1. mmap MUST stay enabled (default) — --no-mmap forces the 50.66 GiB
+#     table into anonymous memory and the cgroup killer takes the
+#     process with nothing in the server log (only dmesg).
+#  2. Do NOT pin the PLE tensor to CPU with -ot — measured gen fell
+#     ~23 -> 13.4 tok/s; the kernel streams it better.
+# NOTE: the card's own build_attn_qsa still calls ggml_fill on the
+# F16-cast kq mask (the path that aborted our charlie build with
+# --flash-attn on before scripts/patches/qwen4exp-qsa-f32-mask.patch).
+# If a clean kingjones-tree build aborts graph build with
+# GGML_ASSERT(a-type == GGML_TYPE_F32), port that fix (F32 base mask +
+# cast at the attention call) into this tree and rebuild.
+#
+# WARNING: do not name any variable TEMP, HOST, or GPU_MAX_HEAP_SIZE. The ROCm HIP runtime reads those bare env vars (see the 2026-08-13 TEMP
+# incident). llama.cpp reads --host/--port/--temp as CLI flags, never as
+# env vars.
+
+# kingjones30/ROCmFPX runtime binary (card recipe build).
+LLAMA_BIN=/home/dank/GitPackages/kingjones-ROCmFPX/build/bin/llama-server
+
+# Model package from
+# https://huggingface.co/kingjones777/Qwen3.8-Flash-Next-ROCmFP4-STRIX-GGUF
+# Layout: single unsharded GGUF, text-only (the repo ships no mmproj;
+# the vision tower below is borrowed from the agentionai FAST package —
+# same base model, f16 — kept only if it loads clean, drop
+# MMPROJ_PATH if it does not).
+MODEL_DIR=/home/dank/models/kingjones-Qwen3.8-Flash-Next-ROCmFP4-STRIX
+MODEL_FILE=Qwen3.8-Flash-Next-Q4_0-ROCmFP4-STRIX.gguf
+MMPROJ_PATH=/home/dank/models/agentionai-Qwen3.8-Flash-Next-ROCmFP4-FAST/mmproj/mmproj-Qwen3.8-Flash-Next-f16.gguf
+
+# Qwen-Sharp chat template (tracked in the repo, staged on the host). The
+# Sharp template drives thinking on by default (medium effort); the server
+# emits reasoning in DeepSeek format so clients receive reasoning_content
+# separately instead of literal thinking tags in content.
+CHAT_TEMPLATE_FILE=/home/dank/serverAppConfigs/scripts/chat-template.jinja
+CHAT_TEMPLATE_KWARGS='{"enable_thinking":true,"preserve_thinking":true,"preserve_reasoning":true}'
+
+# Alias + network.
+ALIAS=frank-large
+BIND_ADDR=0.0.0.0
+LISTEN_PORT=7770
+
+# Serving: 401408-token context, TWO slots (parallel 2), batch 4096 /
+# ubatch 512, 16+16 threads, f16 KV, 16 context checkpoints, 8 GiB
+# prompt cache. --fit off: auto-fit segfaults on this model's
+# hyper-connection/indexer config; -ngl 999 puts every layer on the
+# SVM-backed GPU pool. ubatch 512 matches the card's re-measurement
+# (-ub 1024/2048/4096 showed no effect; -t 32 vs 16 is noise).
+#
+# CTX_SIZE=401408 / PARALLEL=2: carried over from the FAST deployment
+# where it loaded clean (~74 s, health 200). The card measures the full
+# 262144 window at ~71.7 GiB GTT at load on the same 128 GB board; QSA
+# caps KV growth so the extra window is cheap. KV footprint scales
+# with the 12 attention layers only. Verify load on first STRIX boot.
+CTX_SIZE=401408
+PARALLEL=2
+
+BATCH_SIZE=4096
+UBATCH_SIZE=512
+NUM_THREADS=16
+THREAD_BATCH=16
+CACHE_TYPE_K=f16
+CACHE_TYPE_V=f16
+CTX_CHECKPOINTS=16
+CACHE_RAM=8192
+SLOT_SAVE_PATH=/home/dank/serverAppConfigs/scripts/slot-state
+SPLIT_MODE=none
+FIT=off
+NGL=999
+FLASH_ATTN=on
+
+# Sampling (card instruct table, non-thinking).
+TEMPERATURE=0.7
+TOP_K=20
+TOP_P=0.8
+MIN_P=0.0
+PRESENCE_PENALTY=1.5
+
+# Reasoning on (Sharp template drives it; preserve_thinking + preserve_reasoning
+# keep the reasoning trail in the conversation). DeepSeek format projects the
+# reasoning into the API's reasoning_content field instead of literal thinking
+# tags.
+REASONING=on
+REASONING_FORMAT=deepseek
+
+# Strix Halo unified memory + native gfx1151.
+HSA_OVERRIDE_GFX_VERSION=11.5.1
+
+VERBOSITY=4
+
+
+# Speculative decoding: disabled. The card's Usage block ships no spec
+# flags. The Flash-Next MTP heads do not load on this loader (requires
+# per-layer blk.N.hc_attn_norm; every head carries only blk.48 —
+# verified 3x on build-qwen38). DRAFT_MODEL_* point at the retired
+# agentionai FAST head and are unused while SPEC_TYPE=none.
+DRAFT_MODEL_DIR=/home/dank/models/mtp-Qwen3.8-Flash-Next-FAST
+DRAFT_MODEL_FILE=Qwen3.8-Flash-Next-MTP-ROCmFP4-FAST.gguf
+SPEC_TYPE=none
+SPEC_DRAFT_N_MAX=3
+SPEC_DRAFT_N_MIN=0
+SPEC_DRAFT_P_MIN=0.75
+SPEC_DRAFT_NGL=999
+SPEC_DRAFT_DEVICE=ROCm0
+SPEC_DRAFT_TYPE_K=f16
+SPEC_DRAFT_TYPE_V=f16
@@ -0,0 +1,16 @@
+[Unit]
+Description=ROCmFPX (kingjones30 fork) llama.cpp Qwen3.8-Flash-Next ROCmFP4 STRIX 177B MoE (dankclaw, frank-large)
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+User=dank
+Group=dank
+Environment="FRANK_LARGE_ENV_FILE=/home/dank/serverAppConfigs/scripts/frank-large.env"
+ExecStart=/home/dank/serverAppConfigs/scripts/serve-frank-large.sh
+Restart=on-failure
+RestartSec=10
+
+[Install]
+WantedBy=multi-user.target
@@ -0,0 +1,23 @@
+# Tunables for frank-npu (FastFlowLM NPU endpoint, Dankclaw)
+# Runtime: FLM 1.0.1 (/usr/bin/flm) on the XDNA2 NPU (50 TOPS). Boot tag is
+# nuextract3:4b (NuExtract3-4B-NPU2, Qwen3.5-4B skeleton, structured-extraction
+# model); serve routes every request by the exact FLM registry tag in the
+# request's model field, so any registered tag (gemma4-it:e4b, qwythos:9b, ...)
+# is reachable on this port regardless of the boot tag. The NPU is a
+# single-lane device: generations serialize FIFO, and the first request after a
+# model swap pays the ~6-11 s load cost. The NPU draws from the unified RAM
+# pool shared with frank-large (iGPU).
+FLM_MODEL=nuextract3:4b
+FLM_HOST=0.0.0.0
+FLM_PORT=7771
+# turbo = max NPU power mode (xrt-smi configure --pmode turbo at serve start).
+# No other tenant uses the NPU on this host, so the top mode costs nothing to
+# neighbours; drop to `performance` if iGPU power-budget contention ever shows.
+FLM_PMODE=turbo
+# FLM_ASR=1 additionally loads Whisper-V3-Turbo-NPU2 (+0.62 GB) and exposes
+# /v1/audio/transcriptions. NuExtract3 answers text-only; audio stays off.
+FLM_ASR=0
+# -1 = no --ctx-len flag: every pack loads with its own trained context
+# (nuextract3:4b -> 32768, gemma4-it:e4b -> 131072). A global override would
+# over-allocate KV for the small packs.
+FLM_CTX_LEN=-1
@@ -0,0 +1,17 @@
+[Unit]
+Description=FastFlowLM NPU sight+sound inference (frank-npu) on Dankclaw
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+User=dank
+Group=dank
+Environment="FRANK_NPU_ENV_FILE=/home/dank/serverAppConfigs/scripts/frank-npu.env"
+ExecStart=/home/dank/serverAppConfigs/scripts/serve-frank-npu.sh
+LimitMEMLOCK=infinity
+Restart=on-failure
+RestartSec=10
+
+[Install]
+WantedBy=multi-user.target
@@ -0,0 +1,180 @@
+#!/usr/bin/env bash
+# Serve kingjones777/Qwen3.8-Flash-Next-ROCmFP4-STRIX-GGUF as
+# frank-large via the kingjones30/ROCmFPX llama.cpp runtime on
+# Dankclaw (Strix Halo gfx1151, unified memory).
+#
+# Runtime: kingjones30/ROCmFPX (fork of charlie12345/ROCmFPX with the
+# model card's qwen4exp-on-rocmfpx patch) built at
+# /home/dank/GitPackages/kingjones-ROCmFPX/build (HIP, gfx1151,
+# ROCm 7.2.4, card recipe). The ROCmFP4 quant types plus the qwen4exp
+# architecture must be in one tree; stock llama.cpp cannot load this
+# GGUF.
+#
+# Model: Qwen3.8-Flash-Next base (qwen4exp MoE: 512 experts / 10 used,
+# 48 layers = 12 attention + 36 SSM, QSA indexer, hyper-connections, PLE
+# at layer 1) quantized Q4_0_ROCMFP4_STRIX, 113.47 GiB single file. The
+# 50.66 GiB PLE n-gram table stays OUT of GPU memory and streams via
+# mmap from the OS page cache — never add --no-mmap or -ot for it.
+#
+# Profile: card sampling table (temp 0.7, top_p 0.80, top_k 20, min_p 0.0,
+# presence 1.5 — non-thinking instruct) mixed with the validated serving
+# profile: ctx 401408 x 2 slots, b 4096 / ub 512, t 16 / tb 16, f16 KV,
+# fa on (if the graph build aborts on the QSA F16 mask, see the note in
+# the env file: port scripts/patches/qwen4exp-qsa-f32-mask.patch into
+# this tree and rebuild), cont-batching, prompt cache + 8 GiB cache-ram
+# + idle-slot cache, 16 ctx checkpoints, --fit off (auto-fit segfaults
+# on qwen4exp hyper-connection config). Speculative decoding off: the
+# Flash-Next MTP heads do not load on this loader; the card ships no
+# spec flags.
+set -euo pipefail
+
+SCRIPT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
+ENV_FILE="${FRANK_LARGE_ENV_FILE:-$SCRIPT_DIR/frank-large.env}"
+if [[ -f "$ENV_FILE" ]]; then
+    set -a
+    # shellcheck disable=SC1090
+    source "$ENV_FILE"
+    set +a
+fi
+
+# --- defaults (used when the env file does not set them) -------------------
+: "${LLAMA_BIN:=/home/dank/GitPackages/kingjones-ROCmFPX/build/bin/llama-server}"
+: "${MODEL_DIR:=/home/dank/models/kingjones-Qwen3.8-Flash-Next-ROCmFP4-STRIX}"
+: "${MODEL_FILE:=Qwen3.8-Flash-Next-Q4_0-ROCmFP4-STRIX.gguf}"
+: "${MMPROJ_PATH:=/home/dank/models/agentionai-Qwen3.8-Flash-Next-ROCmFP4-FAST/mmproj/mmproj-Qwen3.8-Flash-Next-f16.gguf}"
+: "${CHAT_TEMPLATE_FILE:=$SCRIPT_DIR/chat-template.jinja}"
+: "${CHAT_TEMPLATE_KWARGS:={}}"
+: "${ALIAS:=frank-large}"
+: "${BIND_ADDR:=0.0.0.0}"
+: "${LISTEN_PORT:=7770}"
+: "${CTX_SIZE:=401408}"
+: "${PARALLEL:=2}"
+: "${BATCH_SIZE:=4096}"
+: "${UBATCH_SIZE:=512}"
+: "${NUM_THREADS:=16}"
+: "${THREAD_BATCH:=16}"
+: "${CACHE_TYPE_K:=f16}"
+: "${CACHE_TYPE_V:=f16}"
+: "${CTX_CHECKPOINTS:=16}"
+: "${CACHE_RAM:=8192}"
+: "${SLOT_SAVE_PATH:=$SCRIPT_DIR/slot-state}"
+: "${SPLIT_MODE:=none}"
+: "${FIT:=off}"
+: "${NGL:=999}"
+: "${FLASH_ATTN:=on}"
+: "${TEMPERATURE:=0.7}"
+: "${TOP_K:=20}"
+: "${TOP_P:=0.8}"
+: "${MIN_P:=0.0}"
+: "${PRESENCE_PENALTY:=1.5}"
+: "${REASONING:=on}"
+: "${REASONING_FORMAT:=deepseek}"
+: "${HSA_OVERRIDE_GFX_VERSION:=11.5.1}"
+: "${SPEC_TYPE:=none}"
+: "${DRAFT_MODEL_DIR:=/home/dank/models/mtp-Qwen3.8-Flash-Next-FAST}"
+: "${DRAFT_MODEL_FILE:=Qwen3.8-Flash-Next-MTP-ROCmFP4-FAST.gguf}"
+: "${SPEC_DRAFT_N_MAX:=3}"
+: "${SPEC_DRAFT_N_MIN:=0}"
+: "${SPEC_DRAFT_P_MIN:=0.75}"
+: "${SPEC_DRAFT_NGL:=999}"
+: "${SPEC_DRAFT_DEVICE:=ROCm0}"
+: "${SPEC_DRAFT_TYPE_K:=f16}"
+: "${SPEC_DRAFT_TYPE_V:=f16}"
+
+MODEL="$MODEL_DIR/$MODEL_FILE"
+for f in "$LLAMA_BIN" "$MODEL" "$CHAT_TEMPLATE_FILE"; do
+    if [[ ! -r "$f" ]]; then
+        echo "missing or unreadable required artifact: $f" >&2
+        exit 1
+    fi
+done
+if [[ ! -x "$LLAMA_BIN" ]]; then
+    echo "missing executable runtime: $LLAMA_BIN" >&2
+    exit 1
+fi
+
+# --- speculative decoding (MTP draft head) ---------------------------------
+# The MTP draft head is a separate GGUF sidecar loaded via --spec-draft-model
+# (alias --model-draft).  ngram-only specs do not need a draft file.
+# Set SPEC_TYPE=none in the env to disable speculative decoding entirely.
+SPEC_ARGS=()
+if [[ -n "$SPEC_TYPE" && "$SPEC_TYPE" != "none" ]]; then
+    SPEC_ARGS=(
+        --spec-type "$SPEC_TYPE"
+        --spec-draft-n-max "$SPEC_DRAFT_N_MAX"
+        --spec-draft-n-min "$SPEC_DRAFT_N_MIN"
+        --spec-draft-p-min "$SPEC_DRAFT_P_MIN"
+    )
+    if [[ "$SPEC_TYPE" == *draft-mtp* ]]; then
+        DRAFT_MODEL="$DRAFT_MODEL_DIR/$DRAFT_MODEL_FILE"
+        if [[ ! -r "$DRAFT_MODEL" ]]; then
+            echo "missing or unreadable draft model: $DRAFT_MODEL" >&2
+            exit 1
+        fi
+        SPEC_ARGS+=(
+            --spec-draft-model "$DRAFT_MODEL"
+            --spec-draft-ngl "$SPEC_DRAFT_NGL"
+            --spec-draft-device "$SPEC_DRAFT_DEVICE"
+            --spec-draft-type-k "$SPEC_DRAFT_TYPE_K"
+            --spec-draft-type-v "$SPEC_DRAFT_TYPE_V"
+        )
+    fi
+    if [[ -n "${SPEC_DRAFT_P_SPLIT:-}" ]]; then
+        SPEC_ARGS+=(--spec-draft-p-split "$SPEC_DRAFT_P_SPLIT")
+    fi
+fi
+
+# The STRIX repo ships no mmproj; this is the agentionai FAST package's
+# f16 vision tower for the same base model. Passed only when readable.
+MMPROJ_ARGS=()
+if [[ -n "${MMPROJ_PATH:-}" && -r "${MMPROJ_PATH:-}" ]]; then
+    MMPROJ_ARGS=(--mmproj "$MMPROJ_PATH")
+fi
+
+SERVER_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
+mkdir -p "$SLOT_SAVE_PATH"
+ROCM_PATH="${ROCM_PATH:-/opt/rocm}"
+LIBRARY_PATH="$SERVER_DIR:$ROCM_PATH/lib:$ROCM_PATH/lib/rocm_sysdeps/lib:$ROCM_PATH/lib/llvm/lib${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
+
+exec /usr/bin/env \
+    ROCM_PATH="$ROCM_PATH" \
+    HIP_VISIBLE_DEVICES=0 \
+    LD_LIBRARY_PATH="$LIBRARY_PATH" \
+    HSA_OVERRIDE_GFX_VERSION="$HSA_OVERRIDE_GFX_VERSION" \
+    "$LLAMA_BIN" \
+    --model "$MODEL" \
+    --alias "$ALIAS" \
+    --chat-template-file "$CHAT_TEMPLATE_FILE" \
+    --chat-template-kwargs "$CHAT_TEMPLATE_KWARGS" \
+    --host "$BIND_ADDR" \
+    --port "$LISTEN_PORT" \
+    --jinja \
+    --device ROCm0 \
+    --gpu-layers "$NGL" \
+    --split-mode "$SPLIT_MODE" \
+    --parallel "$PARALLEL" \
+    --ctx-size "$CTX_SIZE" \
+    --batch-size "$BATCH_SIZE" \
+    --ubatch-size "$UBATCH_SIZE" \
+    --fit "$FIT" \
+    --flash-attn "$FLASH_ATTN" \
+    --cache-type-k "$CACHE_TYPE_K" \
+    --cache-type-v "$CACHE_TYPE_V" \
+    --threads "$NUM_THREADS" \
+    --threads-batch "$THREAD_BATCH" \
+    --ctx-checkpoints "$CTX_CHECKPOINTS" \
+    --cache-ram "$CACHE_RAM" \
+    --slot-save-path "$SLOT_SAVE_PATH" \
+    --cache-prompt \
+    --cache-idle-slots \
+    --metrics \
+    --temp "$TEMPERATURE" \
+    --top-p "$TOP_P" \
+    --top-k "$TOP_K" \
+    --min-p "$MIN_P" \
+    --presence-penalty "$PRESENCE_PENALTY" \
+    --reasoning "$REASONING" \
+    --reasoning-format "$REASONING_FORMAT" \
+    --reasoning-budget -1 \
+    "${MMPROJ_ARGS[@]}" \
+    "${SPEC_ARGS[@]}"
@@ -0,0 +1,62 @@
+#!/usr/bin/env bash
+# Serve the FastFlowLM (FLM) sight+sound endpoint — frank-npu.
+#
+# Runs `flm serve <tag>` on the XDNA2 NPU (50 TOPS, Dankclaw). The NPU is a
+# single-lane device: generations serialize in FIFO order, so this is a
+# background/triage endpoint, not an interactive multi-user chat. The NPU
+# draws from the unified RAM pool shared with frank-large (the iGPU
+# llama.cpp service); see serverAppConfigs-docs/LLM/frank-npu.md
+# for the memory budget and the model-selection tradeoff.
+#
+# NOTE: FLM requires unlimited memlock. The unit sets LimitMEMLOCK=infinity.
+# NOTE: an unknown model tag does NOT 404 — /v1/chat/completions answers HTTP 200 with a
+# flat {"error":"Failed to load <tag> model!"} body and no choices; /v1/completions returns a
+# nested OpenAI-shaped server_error (code 500) also under HTTP 200. Client model IDs must be the exact FLM registry
+# tag (e.g. gemma4-it:e4b), never a service alias. Custom tags only resolve
+# when FLM_CONFIG_PATH points at the user registry that carries them and their
+# kernels are symlinked into /home/dank/.config/flm/xclbins/<PackName>.
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+ENV_FILE="${FRANK_NPU_ENV_FILE:-$SCRIPT_DIR/frank-npu.env}"
+if [[ -f "$ENV_FILE" ]]; then
+    set -a
+    # shellcheck disable=SC1090
+    source "$ENV_FILE"
+    set +a
+fi
+
+# --- defaults (used when the env file does not set them) -------------------
+: "${FLM_BIN:=/usr/bin/flm}"
+: "${FLM_MODEL:=gemma4-it:e4b}"
+: "${FLM_HOST:=0.0.0.0}"
+: "${FLM_PORT:=7771}"
+: "${FLM_PMODE:=turbo}"
+: "${FLM_ASR:=0}"
+: "${FLM_CTX_LEN:=131072}"
+
+# The user registry is a superset of the system registry (it carries the
+# custom tags too), and its model_path resolves relative to the registry file,
+# so exports keep a tag swap in the env file from needing script edits.
+export FLM_CONFIG_PATH="${FLM_CONFIG_PATH:-/home/dank/.config/flm/model_list.json}"
+export FLM_XCLBIN_PATH="${FLM_XCLBIN_PATH:-/home/dank/.config/flm/xclbins}"
+
+if [[ ! -x "$FLM_BIN" ]]; then
+    echo "missing flm binary: $FLM_BIN" >&2
+    exit 1
+fi
+
+# --- optional flags ----------------------------------------------------------
+OPT_ARGS=()
+if [[ "${FLM_ASR}" != "0" ]]; then
+    OPT_ARGS+=(--asr "${FLM_ASR}")
+fi
+if [[ "${FLM_CTX_LEN}" != "-1" ]]; then
+    OPT_ARGS+=(--ctx-len "${FLM_CTX_LEN}")
+fi
+
+exec "$FLM_BIN" serve "$FLM_MODEL" \
+    --host "$FLM_HOST" \
+    -p "$FLM_PORT" \
+    --pmode "$FLM_PMODE" \
+    "${OPT_ARGS[@]}"
@@ -0,0 +1,87 @@
+#!/usr/bin/env bash
+# Serve peculiar-ragdoll/Tiel-Coder-35B-A3B (UD-Q6_K_XL) as
+# frank-default-small via stock llama.cpp on Dankserver (RTX 3090).
+#
+# Runtime: upstream llama.cpp (latest master, CUDA build with
+# -DGGML_CUDA_FA_ALL_QUANTS=ON -DCMAKE_CUDA_ARCHITECTURES=86).
+# The model is a 35B-total / 8-active MoE hybrid (qwen35moe arch) with
+# SSM/Mamba layers (full_attention_interval=4: 10 attention + 30 SSM
+# layers out of 40).  256 experts, 8 used per token.
+#
+# Expert offload: --cpu-moe keeps all 256 expert weights in CPU RAM.
+# Non-expert weights (attention, SSM, shared expert, embeddings, output)
+# are ~1.6 GB and fit entirely on the 24 GB RTX 3090.
+#
+# KV cache: Q8_0 for both K and V.  With CTX_SIZE=200700 (66 900 per
+# slot × 3 parallel) and 10 attention layers (2 KV heads × 256 dim),
+# the KV cache is ~2 GB — the requested 2 GB VRAM buffer.
+#
+# Model is loaded with mmap (default).  The 30 GB Q6_K_XL file is
+# memory-mapped; only accessed pages enter RAM.
+#
+# Tunables come from frank-default-small.env unless FRANK_SMALL_ENV_FILE
+# overrides that path.
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+ENV_FILE="${FRANK_SMALL_ENV_FILE:-$SCRIPT_DIR/frank-default-small.env}"
+if [[ -f "$ENV_FILE" ]]; then
+    set -a
+    # shellcheck disable=SC1090
+    source "$ENV_FILE"
+    set +a
+fi
+
+# --- defaults (used when the env file does not set them) -----------------
+: "${MODEL_DIR:=/home/dank/models/tiel-coder-35b}"
+: "${MODEL_FILE:=Tiel-Coder-35B-A3B-UD-Q6_K_XL.gguf}"
+: "${LLAMA_SERVER:=/home/dank/llama.cpp/build/bin/llama-server}"
+: "${ALIAS:=frank-default-small}"
+: "${BIND_ADDR:=0.0.0.0}"
+: "${LISTEN_PORT:=7770}"
+: "${CTX_SIZE:=200700}"
+: "${PARALLEL:=3}"
+: "${CACHE_TYPE_K:=q8_0}"
+: "${CACHE_TYPE_V:=q8_0}"
+: "${N_GPU_LAYERS:=999}"
+: "${BATCH_SIZE:=4096}"
+: "${UBATCH_SIZE:=1024}"
+: "${VERBOSITY:=2}"
+: "${CPU_MOE:=1}"
+: "${TEMPERATURE:=1.0}"
+: "${TOP_K:=20}"
+: "${TOP_P:=0.95}"
+: "${MIN_P:=0.0}"
+: "${THREADS:=7}"
+: "${THREADS_HTTP:=8}"
+
+MODEL="$MODEL_DIR/$MODEL_FILE"
+
+if [[ ! -f "$MODEL" ]]; then
+    echo "missing required file: $MODEL" >&2
+    exit 1
+fi
+if [[ ! -x "$LLAMA_SERVER" ]]; then
+    echo "missing llama-server binary: $LLAMA_SERVER" >&2
+    exit 1
+fi
+
+exec "$LLAMA_SERVER" \
+    --model "$MODEL" \
+    $([ "$CPU_MOE" = "1" ] && printf -- "--cpu-moe") \
+    --cache-type-k "$CACHE_TYPE_K" \
+    --cache-type-v "$CACHE_TYPE_V" \
+    --ctx-size "$CTX_SIZE" \
+    --parallel "$PARALLEL" \
+    --gpu-layers "$N_GPU_LAYERS" \
+    --batch-size "$BATCH_SIZE" \
+    --ubatch-size "$UBATCH_SIZE" \
+    --temp "$TEMPERATURE" \
+    --top-k "$TOP_K" \
+    --top-p "$TOP_P" \
+    --min-p "$MIN_P" \
+    --threads "$THREADS" \
+    --threads-http "$THREADS_HTTP" \
+    --alias "$ALIAS" \
+    --host "$BIND_ADDR" \
+    --port "$LISTEN_PORT"
@@ -1,16 +1,5 @@
 # Server App Configs

-### AI Disclosures:
-
- Slop rating: 2/5 - Actively using, actively de-slopping, **repository mostly human owned**, lacks formal process.
- Completeness: Unfinished, planning planned, no plan yet
- Datacenter model/spend: Less than $20 of Deepseek 4.1 Flash and GLM-5.3-Flash
- Local model: Qwen 3.8 Flash Next
-
-### slop rating of 2 or lower indicates human reading is a worthwhile exercise, and that Frank has read the code and knows the architecture.
-### slop rating of 2 indicates that while this project is human owned, no human contribution process exists and/or release is agentically driven.
-
-
 The single repository for the two homelab servers: the Podman container
 stacks **and** the LLM inference fleet. One repo, one checkout per host,
 one deploy script.
@@ -24,7 +13,7 @@ one deploy script.
 | Host | User | IP | Podman stack | LLM service | LLM port |
 |------|------|----|--------------|-------------|----------|
 | dankserver | dank | 192.168.0.235 | media and library | frank-medium (ROCmFPX llama.cpp, Qwen3.8-27B AGENT) | 7770 |
-| dankclaw | dank | 192.168.0.143 | AI, search, development, observability | frank-dragoon (gufo, Qwen3.8-27B UD-Q8_K_XL) | 7770 |
+| dankclaw | dank | 192.168.0.143 | AI, search, development, observability | frank-large (gufo, Qwen3.8-27B UD-Q8_K_XL) | 7770 |

 ## Active Services

@@ -41,7 +30,7 @@ one deploy script.
 | Gitea | dankclaw | gitea.dankclaw.local | 3001 |
 | Grafana | dankclaw | grafana.dankclaw.local | 3002 |
 | Prometheus | dankclaw | prometheus.dankclaw.local | 9090 |
-| frank-dragoon | dankclaw | — (bare endpoint) | 7770 |
+| frank-large | dankclaw | — (bare endpoint) | 7770 |

 ## Structure

@@ -1,26 +1,26 @@
 #!/usr/bin/env bash
-# autoresearch.sh — canonical frank-archon benchmark entry point (dev box).
+# autoresearch.sh — canonical frank-primary benchmark entry point (dev box).
 #
 # The service runs on dankserver (192.168.0.235:7770), the two R9700s live
 # there, and the benchmark must own those cards, so the whole stop -> boot
 # scratch -> bench -> restore sequence runs ON dankserver (see
-# scripts/autoresearch-frank-archon.sh). This wrapper:
-#   1. copies the candidate config (the repo's scripts/frank-archon.env, the
+# scripts/autoresearch-frank-primary.sh). This wrapper:
+#   1. copies the candidate config (the repo's scripts/frank-primary.env, the
 #      file the autoresearch loop edits) and the bench script to dankserver,
 #   2. runs the bench over ssh, relaying its stdout (the METRIC lines) and
 #      exit code back here verbatim.
 # It deliberately does NOT push to origin: the scp dirties dankserver's
 # working tree, and the loop's uncommitted candidate config must not be
-# pushed. Only this file and scripts/autoresearch-frank-archon.sh are
+# pushed. Only this file and scripts/autoresearch-frank-primary.sh are
 # committed; the candidate env rides over ssh each run.
 set -uo pipefail
 DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 REMOTE=dankserver
 REMOTE_REPO=/home/dank/serverAppConfigs

-scp -q "$DIR/scripts/frank-archon.env" \
-    "$REMOTE:$REMOTE_REPO/scripts/frank-archon.env" || { echo "scp env failed" >&2; exit 1; }
-scp -q "$DIR/scripts/autoresearch-frank-archon.sh" \
-    "$REMOTE:$REMOTE_REPO/scripts/autoresearch-frank-archon.sh" || { echo "scp script failed" >&2; exit 1; }
+scp -q "$DIR/scripts/frank-primary.env" \
+    "$REMOTE:$REMOTE_REPO/scripts/frank-primary.env" || { echo "scp env failed" >&2; exit 1; }
+scp -q "$DIR/scripts/autoresearch-frank-primary.sh" \
+    "$REMOTE:$REMOTE_REPO/scripts/autoresearch-frank-primary.sh" || { echo "scp script failed" >&2; exit 1; }

-ssh "$REMOTE" "bash $REMOTE_REPO/scripts/autoresearch-frank-archon.sh $REMOTE_REPO/scripts/frank-archon.env"
+ssh "$REMOTE" "bash $REMOTE_REPO/scripts/autoresearch-frank-primary.sh $REMOTE_REPO/scripts/frank-primary.env"
@@ -79,23 +79,3 @@ MARRETA_APP_KEY=changeme
 # account on first boot from these values.
 MARRETA_ADMIN_EMAIL=admin@franksdank.cloud
 MARRETA_ADMIN_PASSWORD=changeme
-
-# open-notebook (podcast generator; internal-only, no vhost).
-# deploy-remote.sh seeds /home/dank/Containers/open-notebook/env from these
-# values (only if that file does not already exist) and refuses to start the
-# units while any of them is blank: a blank OPEN_NOTEBOOK_PASSWORD means
-# authentication OFF, and a blank encryption key means no provider credential
-# can be saved. The seed writes SURREAL_PASS too — the spelling the `surreal`
-# CLI reads — from SURREAL_PASSWORD, so the app and the database cannot be
-# given different values.
-# OPEN_NOTEBOOK_ENCRYPTION_KEY: encrypts the stored AI provider keys at rest.
-#   Generate: openssl rand -hex 32
-# Changing it after credentials are saved makes them undecryptable.
-OPEN_NOTEBOOK_ENCRYPTION_KEY=changeme
-# UI/API password. Every API route — including the episode-audio download the
-# app serves without a session — is open to the LAN while this is unset.
-OPEN_NOTEBOOK_PASSWORD=changeme
-# SurrealDB root password (the root user NAME comes from
-# scripts/open-notebook.env, so this file holds only the secret).
-#   Generate: openssl rand -base64 32
-SURREAL_PASSWORD=changeme
@@ -162,25 +162,6 @@ marreta.franksdank.cloud {
 	}
 }

-# --- Open Notebook (guarded: Anubis → Authentik; notebook-proxy provider, internal_host caddy:8017) ---
-# The podcast generator. ONE upstream carries the whole app: the image's own
-# Next.js layer rewrites /api/* to the FastAPI beside it (INTERNAL_API_URL,
-# default http://localhost:5055 inside the container), so a browser on this
-# vhost never needs the API port. That is why scripts/open-notebook.env sets
-# API_URL to exactly this origin — left unset, the app auto-detects
-# `<host-you-opened>:5055`, and 5055 on this host belongs to seerr — and why
-# the published API port sits on loopback. App-level login
-# (OPEN_NOTEBOOK_PASSWORD) stays on behind the chain: two owners, edge identity
-# and the app's own session.
-# Dialled by published LAN port, not container name: open-notebook is a
-# rootless quadlet and this Caddy lives in root's storage, so root's DNS does
-# not see the container (the same reason marreta and SearXNG are dialled by IP).
-notebook.franksdank.cloud {
-	reverse_proxy anubis:8923 {
-		header_up X-Real-Ip {remote_host}
-	}
-}
-
 # --- Pi-hole (quadlet on slirp4netns; published 8083) ---
 pihole.franksdank.cloud {
 	basicauth {
@@ -439,16 +420,6 @@ http://:8017 {
 	handle @marreta {
 		reverse_proxy 192.168.0.143:8082
 	}
-
-	@notebook host notebook.franksdank.cloud
-	handle @notebook {
-		# Bare proxy, like ai.franksdank.cloud: the app's SSE answers
-		# (/api/search/ask, chat sessions) stream for minutes through this
-		# same listener with no flush or timeout tuning, so nothing here
-		# needs it — a read/write timeout on the route would be the thing
-		# that cuts them.
-		reverse_proxy 192.168.0.143:8502
-	}
 }

 # --- Optional: Jackett ---
@@ -23,7 +23,7 @@ Network=dankclaw_default
 # Image=ghcr.io/goauthentik/ldap:2026.8.3 (amd64 digest; podman 5.7.0 refuses
 # the tag's nested manifest list, so pin the single-platform digest).
 EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
-Image=ghcr.io/goauthentik/ldap:2026.8.5@sha256:4c75bd27918418d2d6fbbfe4581bbd5a199fe38942d24559c4483edd64a1ea87
+Image=ghcr.io/goauthentik/ldap@sha256:2e7b259eb8c37faac916946ddf48ec681ee9d3a9f724fe8521bf161973f13cf9
 Environment=AUTHENTIK_HOST=https://auth.franksdank.cloud
 Environment=AUTHENTIK_INSECURE=false
 Environment=AUTHENTIK_TOKEN=${AUTHENTIK_LDAP_OUTPOST_TOKEN}
@@ -20,7 +20,7 @@ Wants=authentik-server.service

 [Container]
 ContainerName=authentik-outpost
-Image=ghcr.io/goauthentik/server:2026.8.5@sha256:602227a86455d6772e3d16b7d852713c4f864f40df5df894abd1285bb53979a5
+Image=ghcr.io/goauthentik/server:2026.8.3
 Exec=worker
 Network=dankclaw_default
 EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
@@ -14,7 +14,7 @@ After=network.target

 [Container]
 ContainerName=authentik-postgres
-Image=docker.io/library/postgres:16-alpine@sha256:0cefaabd882e42bcb2a2f3dc56d3ce4770560ba62fd7f197f1a2e008fd5731ab
+Image=docker.io/library/postgres:16-alpine
 Network=dankclaw_default
 EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
 Environment=POSTGRES_DB=authentik
@@ -19,7 +19,7 @@ Wants=authentik-postgres.service

 [Container]
 ContainerName=authentik-server
-Image=ghcr.io/goauthentik/server:2026.8.5@sha256:602227a86455d6772e3d16b7d852713c4f864f40df5df894abd1285bb53979a5
+Image=ghcr.io/goauthentik/server:2026.8.3
 Exec=server
 Network=dankclaw_default
 EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
@@ -16,7 +16,7 @@ Wants=authentik-server.service

 [Container]
 ContainerName=authentik-worker
-Image=ghcr.io/goauthentik/server:2026.8.5@sha256:602227a86455d6772e3d16b7d852713c4f864f40df5df894abd1285bb53979a5
+Image=ghcr.io/goauthentik/server:2026.8.3
 Exec=worker
 Network=dankclaw_default
 EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
@@ -15,7 +15,7 @@ After=network.target
 [Container]
 ContainerName=gitea-postgres
 Network=gitea.network
-Image=docker.io/library/postgres:16-alpine@sha256:0cefaabd882e42bcb2a2f3dc56d3ce4770560ba62fd7f197f1a2e008fd5731ab
+Image=docker.io/library/postgres:16-alpine
 EnvironmentFile=/home/dank/Containers/gitea/env
 Environment=POSTGRES_USER=gitea
 Environment=POSTGRES_PASSWORD=${GITEA_DB_PASSWORD}
@@ -19,7 +19,7 @@ After=gitea-postgres.service network.target
 [Container]
 ContainerName=gitea-server
 Network=gitea.network
-Image=docker.io/gitea/gitea:latest@sha256:30777c91280d973fa02be48b1ade735bd6859e8f6f7af3b4444d519437852504
+Image=docker.io/gitea/gitea:latest
 EnvironmentFile=/home/dank/Containers/gitea/env
 Environment=DATABASE_TYPE=postgres
 Environment=DATABASE_HOST=gitea-postgres
@@ -29,13 +29,6 @@ Environment=DATABASE_PASSWD=${GITEA_DB_PASSWORD}
 Environment=DATABASE_PORT=5432
 Environment=USER_UID=1000
 Environment=USER_GID=1000
-# Prometheus scrapes /metrics over the published LAN port (job "gitea" in
-# shared/prometheus/prometheus.yml). Off, /metrics is a 404 and the target is
-# permanently red (the audit's "disabled metrics" class). 0.0.0.0 binds inside
-# the container; the only route in is this host's published 3000, and the
-# exposed surface is counters only — same posture as the LLM /metrics.
-Environment=GITEA__metrics__ENABLED=true
-Environment=GITEA__metrics__ENABLED_HOST=0.0.0.0
 PublishPort=0.0.0.0:3000:3000
 PublishPort=2222:22
 Volume=/home/dank/Containers/gitea/data:/data
@@ -25,7 +25,7 @@ Description=Chrome DevTools Protocol daemon (headless-shell) for hermes
 [Container]
 ContainerName=chrome-devtools-protocol
 Network=host
-Image=docker.io/chromedp/headless-shell@sha256:7bee201478b2f6266aae01243740ee85c16281e32f42d3aed9fdf04f4eace7a1
+Image=docker.io/chromedp/headless-shell@sha256:2d349b544a1ea6b5b5fd7c0fe99215ff662339c57407ee2e8c0a11af93516b04
 Entrypoint=/headless-shell/headless-shell
 Exec=--no-sandbox --disable-gpu --use-gl=angle --use-angle=swiftshader --remote-debugging-address=127.0.0.1 --remote-debugging-port=9222
 ShmSize=1073741824
@@ -1,43 +0,0 @@
-# If you change this file or any /etc/default/grub.d/*.cfg file,
-# run 'update-grub' afterwards to update /boot/grub/grub.cfg.
-# For full documentation of the options in these files, see:
-#   info -f grub -n 'Simple configuration'
-
-GRUB_DEFAULT=0
-GRUB_TIMEOUT_STYLE=hidden
-GRUB_TIMEOUT=0
-GRUB_DISTRIBUTOR=`( . /etc/os-release && echo ${NAME} )`
-# Identity (passthrough) DMA since 2026-10-18: the IOMMU stays enabled, so the
-# XDNA2 NPU (/dev/accel/accel0, PASID/SVA) is usable again, while the iGPU's DMA
-# runs untranslated. Decided in serverAppConfigs-docs/LLM/
-# dankclaw-iommu-translation-off.md; install with scripts/set-dankclaw-bootline.sh.
-GRUB_CMDLINE_LINUX_DEFAULT="quiet splash iommu=pt"
-
-# If your computer has multiple operating systems installed, then you
-# probably want to run os-prober. However, if your computer is a host
-# for guest OSes installed via LVM or raw disk devices, running
-# os-prober can cause damage to those guest OSes as it mounts
-# filesystems to look for things.
-#GRUB_DISABLE_OS_PROBER=false
-
-# Uncomment to enable BadRAM filtering, modify to suit your needs
-# This works with Linux (no patch required) and with any kernel that obtains
-# the memory map information from GRUB (GNU Mach, kernel of FreeBSD ...)
-#GRUB_BADRAM="0x01234567,0xfefefefe,0x89abcdef,0xefefefef"
-
-# Uncomment to disable graphical terminal
-#GRUB_TERMINAL=console
-
-# The resolution used on graphical terminal
-# note that you can use only modes which your graphic card supports via VBE/GOP/UGA
-# you can see them in real GRUB with the command `videoinfo'
-#GRUB_GFXMODE=640x480
-
-# Uncomment if you don't want GRUB to pass "root=UUID=xxx" parameter to Linux
-#GRUB_DISABLE_LINUX_UUID=true
-
-# Uncomment to disable generation of recovery mode menu entries
-#GRUB_DISABLE_RECOVERY="true"
-
-# Uncomment to get a beep at grub start
-#GRUB_INIT_TUNE="480 440 1"
@@ -1,65 +0,0 @@
-#!/bin/sh
-# halogen-fabric-clock hold | release | status: the AMD GPU's fabric clock (FCLK), held at its top level.
-#
-# On Strix Halo the GPU and the NPU working at the same time can hang the machine, and corrupt the NPU's results,
-# while the GPU's fabric clock changes speed; held at its top level, the two run together cleanly. halogen-flash-server
-# with HALOGEN_NPU_MODELS refuses to start until the clock is held. This holds it: the performance level `manual` and
-# only the top pp_dpm_fclk level, on every amdgpu device that has the control (the GPU and SoC clocks stay automatic).
-# Cost: about 1 W at idle. Run as root; halogen-fabric-clock.service runs `hold` at every boot and after a resume.
-#
-#   hold     hold it (waits up to 60 s for the GPU driver to come up)
-#   release  give it back to the driver (the performance level `auto`)
-#   status   print each device's state; exit 0 when every device is held
-set -u
-SYS="${HALOGEN_SYSFS:-/sys}"
-
-devices() {
-  for d in "$SYS"/class/drm/card*/device; do
-    [ -w "$d/pp_dpm_fclk" ] || [ -r "$d/pp_dpm_fclk" ] || continue
-    [ -e "$d/power_dpm_force_performance_level" ] || continue
-    echo "$d"
-  done | sort -u
-}
-
-held() {   # held DEVICE_DIR: the performance level high, or manual with only the top fclk level selected
-  perf=$(cat "$1/power_dpm_force_performance_level" 2>/dev/null)
-  [ "$perf" = high ] && return 0
-  [ "$perf" = manual ] || return 1
-  top=$(grep . "$1/pp_dpm_fclk" | tail -n 1)
-  [ "$(grep -c '\*' "$1/pp_dpm_fclk")" = 1 ] && [ "${top%\*}" != "$top" ]
-}
-
-cmd="${1:-status}"
-case "$cmd" in
-  hold)
-    n=0
-    while [ -z "$(devices)" ] && [ "$n" -lt 60 ]; do sleep 1; n=$((n + 1)); done
-    [ -n "$(devices)" ] || { echo "halogen-fabric-clock: no amdgpu device with a fabric clock control (pp_dpm_fclk) under $SYS/class/drm" >&2; exit 1; }
-    rc=0
-    for d in $(devices); do
-      lvl=$(grep . "$d/pp_dpm_fclk" | tail -n 1 | cut -d: -f1)
-      # the driver applies the level several seconds after the write, so the read-back waits up to 30 s
-      ok=0
-      if echo manual > "$d/power_dpm_force_performance_level" && echo "$lvl" > "$d/pp_dpm_fclk"; then
-        t=0; until held "$d" || [ "$t" -ge 300 ]; do sleep 0.1; t=$((t + 1)); done
-        held "$d" && ok=1
-      fi
-      if [ "$ok" = 1 ]; then
-        echo "halogen-fabric-clock: $d: held at level $lvl ($(grep . "$d/pp_dpm_fclk" | tail -n 1 | awk '{print $2}'))"
-      else
-        echo "halogen-fabric-clock: $d: could not hold it (level $lvl)" >&2; rc=1
-      fi
-    done
-    exit "$rc" ;;
-  release)
-    for d in $(devices); do echo auto > "$d/power_dpm_force_performance_level" && echo "halogen-fabric-clock: $d: released (auto)"; done ;;
-  status)
-    [ -n "$(devices)" ] || { echo "halogen-fabric-clock: no amdgpu device with a fabric clock control"; exit 1; }
-    rc=0
-    for d in $(devices); do
-      if held "$d"; then echo "$d: held ($(cat "$d/power_dpm_force_performance_level"))"
-      else echo "$d: NOT held ($(cat "$d/power_dpm_force_performance_level"))"; rc=1; fi
-    done
-    exit "$rc" ;;
-  *) echo "usage: halogen-fabric-clock hold | release | status" >&2; exit 2 ;;
-esac
@@ -1,19 +0,0 @@
-# Holds the AMD GPU's fabric clock at its top level at boot and after every resume, for halogen-flash-server with
-# the NPU (HALOGEN_NPU_MODELS): on Strix Halo the GPU and the NPU at once are safe only with that clock held.
-# Install (as root), from this directory:
-#   install -m 755 halogen-fabric-clock /usr/local/sbin/
-#   install -m 644 halogen-fabric-clock.service /etc/systemd/system/
-#   systemctl daemon-reload && systemctl enable --now halogen-fabric-clock.service
-# Check: halogen-fabric-clock status. Undo: systemctl disable halogen-fabric-clock.service, then
-#   halogen-fabric-clock release (or reboot). The unit runs once at boot and once after each resume (a resume can hand
-#   the clock back to the driver); it holds nothing open, so it shows inactive between runs.
-[Unit]
-Description=Hold the AMD GPU fabric clock at its top level (GPU and NPU together)
-After=systemd-modules-load.service suspend.target hibernate.target hybrid-sleep.target suspend-then-hibernate.target
-
-[Service]
-Type=oneshot
-ExecStart=/usr/local/sbin/halogen-fabric-clock hold
-
-[Install]
-WantedBy=multi-user.target suspend.target hibernate.target hybrid-sleep.target suspend-then-hibernate.target
@@ -23,7 +23,7 @@ After=gluetun.service network.target
 [Container]
 ContainerName=marreta_browser
 Network=gluetun.container
-Image=docker.io/lightpanda/browser:nightly@sha256:e72730e364d1fbf352c1149f29ffa4b278f3b1a55ff0dec3262298ebe624ec7a
+Image=docker.io/lightpanda/browser:nightly

 [Service]
 TimeoutStartSec=300
@@ -32,7 +32,7 @@ After=gluetun.service marreta-browser.service network.target
 [Container]
 ContainerName=marreta
 Network=gluetun.container
-Image=ghcr.io/manualdousuario/marreta:latest@sha256:30f2a34567b79b66da794b4b9423751bbbcd07ceff600d9f6a2d219e7e439362
+Image=ghcr.io/manualdousuario/marreta:latest
 # Relocate off SearXNG's 8080 inside the shared netns.
 Environment=NGINX_HTTP_PORT=8085
 Environment=APP_NAME=Marreta
@@ -33,7 +33,7 @@ Wants=network-online.target

 [Container]
 ContainerName=alertmanager
-Image=docker.io/prom/alertmanager:latest@sha256:39d8bb3966c694a124428a3a88fbcffd40bd069129ecd47c059f8a3f17c55f3b
+Image=docker.io/prom/alertmanager:latest
 Network=dankclaw_default
 User=0:0
 Exec=--config.file=/etc/alertmanager/alertmanager.yml --storage.path=/alertmanager
@@ -1,51 +1,55 @@
-# Quadlet unit: alloy (ROOT system service, boot-persistent).
+# Quadlet unit: alloy (rootless, boot-persistent).
 #
-# Install path on dankclaw: /etc/containers/systemd/alloy.container
-# Unit name: alloy.service (container name: alloy) — SYSTEM manager, root.
+# Install path on dankclaw: ~/.config/containers/systemd/alloy.container
+# Unit name: alloy.service (container name: alloy)
 #
-# WHY ROOT (not rootless, unlike its observability siblings). Alloy reads the
-# host journal. The journal files are root:systemd-journal 0640 (other::---);
-# a rootless container carries no supplementary groups and cannot open them —
-# a rootless read of system.journal returns DENIED (see the Alloy page in the
-# vault). Running as root, one loki.source.journal reads system.journal AND
-# user-1000.journal, so it carries both host logs and the rootless Quadlet
-# containers' logs. Validated 2026-10-20: a root probe read 31,825 journal
-# lines and pushed 26,966 into Loki in 20 s, zero drops.
+# Log shipper. This replaces the compose service of the same name in
+# dankclaw/compose.yaml ("Observability" block).
 #
-# NETWORK=host. Loki publishes 127.0.0.1:3100 (loopback only) on the host, and
-# no bridge container — rootless or root — can reach the host's loopback, so
-# this unit runs on the host network and pushes to 127.0.0.1:3100. The side
-# effect: the UI must bind loopback too, so Exec sets the listen-addr to
-# 127.0.0.1 (the old rootless unit used 0.0.0.0 only because container loopback
-# was not host loopback; under Network=host they are the same address, so the
-# unauthenticated UI stays loopback-only — nothing off-host reaches it).
+# IMAGE. franksdank/alloy is built locally by deploy-remote.sh
+# (build_compose_image) from dankclaw/alloy/Containerfile: upstream
+# grafana/alloy plus a static busybox for the healthcheck. Pull=never keeps a
+# boot from reaching for that local tag at a registry.
 #
-# THE /etc/machine-id MOUNT IS LOAD-BEARING. sd_journal locates the journal
-# directory by the local machine id; without it the container opens nothing and
-# the journal source reads exactly 0 lines — silently, no error, health green.
-# This mount is the entire reason the pipeline carries data. Never remove it.
+# NETWORK. It joins dankclaw_default so its config's loki.write endpoint
+# resolves the container name loki:3100.
 #
-# IMAGE. Built rootless by build_compose_image alloy, then mirrored into the
-# root image store by the deployer (a root build runs in the build container's
-# own resolver and fails on DNS here). Pull=never: boot must not reach for the
-# local tag at a registry.
+# THE UI PORT EXISTS ON PURPOSE. Alloy's UI listens on the container's own
+# 127.0.0.1 by default and cannot be moved from the config file (alloy 1.x
+# refuses the ui block), so the compose service had no reachable health
+# endpoint at all: the homepage status card for alloy reported "down"
+# continuously and there was nothing for a deploy to wait on. The command
+# below therefore moves the UI to the container's wildcard address and the
+# port is published on host loopback only, which puts alloy in the same class
+# as the other three: reachable from the host, by Caddy over the private
+# network, and by nothing off-host. Alloy has no authentication, which is
+# exactly why it is not published wider.
+#
+# SOURCES. The host journald socket and the host log/container-storage trees
+# are bind-mounted read-only; Alloy tails them and pushes to Loki. The config
+# lives at /opt/alloy/config/config.alloy, installed from
+# shared/alloy/config.alloy on every deploy.
+#
+# ORDERING. Wants/After loki.service: without a reachable Loki, Alloy's write
+# client fails and its log pipeline backs up.

 [Unit]
 Description=Alloy log shipper (Dankclaw)
-After=network-online.target
+After=network-online.target loki.service
+Wants=network-online.target loki.service

 [Container]
 ContainerName=alloy
 Image=localhost/franksdank/alloy:latest
 Pull=never
-Network=host
-Exec=run --server.http.listen-addr=127.0.0.1:12345 /etc/alloy/config.alloy
-Volume=/opt/alloy/config:/etc/alloy:ro
-Volume=/etc/machine-id:/etc/machine-id:ro
+Network=dankclaw_default
+Exec=run --server.http.listen-addr=0.0.0.0:12345 /etc/alloy/config.alloy
+PublishPort=127.0.0.1:12345:12345
+Volume=/opt/alloy/config:/etc/alloy
 Volume=/run/systemd/journal/socket:/run/systemd/journal/socket:ro
-Volume=/var/log/journal:/var/log/journal:ro
-Volume=/run/log/journal:/run/log/journal:ro
-HealthCmd=/usr/local/bin/busybox wget --spider -q http://127.0.0.1:12345/-/healthy
+Volume=/var/log:/var/log:ro
+Volume=/var/lib/containers:/var/lib/containers:ro
+HealthCmd=/usr/local/bin/busybox wget --spider -q http://localhost:12345/-/healthy
 HealthInterval=30s
 HealthTimeout=10s
 HealthRetries=3
@@ -56,4 +60,4 @@ Restart=always
 RestartSec=10

 [Install]
-WantedBy=multi-user.target
+WantedBy=default.target
@@ -1,56 +0,0 @@
-# Quadlet unit: blackbox exporter (rootless, host network, boot-persistent).
-#
-# Install path on dankclaw:
-#   ~/.config/containers/systemd/blackbox-exporter.container
-#
-# WHY Network=host. The probes it runs are (a) the host's own LAN IP
-# (192.168.0.143:8081 SearXNG, :3005 OpenWebUI — host-process/portmap binds a
-# bridge container cannot reach; netavark DNATs published ports only, see
-# Container Host Reachability in the vault) and (b) ComfyUI's published port
-# via 127.0.0.1 (rootless --network=host shares the real host netns, so the
-# 0.0.0.0:8188 publish answers on loopback). A bridge-networked prober could
-# reach neither class reliably.
-#
-# WHY the listener binds 127.0.0.1 + LAN. A bridge container CANNOT dial
-# its own host's LAN IP (netavark hairpin gap, measured 2026-10-21), but a
-# bridge container dialing the podman gateway 169.254.1.2 lands on the host
-# LOOPBACK — that is the scrape path, and it needs this exporter to be
-# LISTENING ON LOOPBACK (proved: 200 from the prometheus container to
-# http://169.254.1.2:9115/-/healthy with a loopback bind). The gateway
-# address itself is NOT bindable from the host netns — it exists only
-# inside each rootless netns; binding it dies "bind: cannot assign
-# requested address" (live 2026-10-21). The LAN bind stays for diagnostics.
-# No public DNS name, no guard chain: internal net only, per the exposure
-# policy. The listen flag is --web.listen-address and it is REPEATABLE, one
-# flag per bind; plain --listen-address does not exist and the exporter
-# exits in 100 ms with "unknown long flag" (live 2026-10-21).
-
-[Unit]
-Description=Blackbox exporter (HTTP probe prober)
-After=network-online.target
-
-[Container]
-Image=docker.io/prom/blackbox-exporter@sha256:ea1c7aa3ef3f307cd03e85f34abb604478ccb90b567fc690da0cd9776da76b13
-ContainerName=blackbox-exporter
-Network=host
-# The image's ENTRYPOINT is /bin/blackbox_exporter itself (proved live
-# 2026-10-21: `podman run IMAGE sh` dies "unexpected sh"), so Exec= carries
-# FLAGS ONLY — repeating the binary makes it a positional arg and kingpin
-# exits "unexpected /bin/blackbox_exporter".
-Exec=--config.file=/etc/blackbox_exporter/config.yml --web.listen-address=127.0.0.1:9115 --web.listen-address=192.168.0.143:9115
-Volume=/opt/blackbox/blackbox.yml:/etc/blackbox_exporter/config.yml:ro
-# Plain exec form like every sibling unit. Shell form (`/bin/sh -c "…"`)
-# is passed as one quoted token and quadlet then folds the following
-# --health-interval/--health-retries (and the image name) into the command
-# string — unit died twice that way, live 2026-10-21. wget only; this
-# busybox variant has no curl, and host-network wget answers on loopback.
-HealthCmd=wget -q -O /dev/null http://127.0.0.1:9115/-/healthy
-HealthInterval=30s
-HealthRetries=3
-
-[Service]
-Restart=always
-RestartSec=5
-
-[Install]
-WantedBy=default.target
@@ -1,56 +0,0 @@
-# Quadlet unit: cadvisor (rootless, boot-persistent) — DANKCLAW.
-#
-# Install path on dankclaw: ~/.config/containers/systemd/cadvisor.container
-# Unit name: cadvisor.service (container name: cadvisor)
-#
-# WHY THIS EXISTS. The app stack is ~34 rootless Podman containers and until
-# now nothing reported per-container CPU/memory — Grafana could only see the
-# host (node_exporter) and the LLM engines. cadvisor reads the host cgroup
-# tree and exports container_cpu_* / container_memory_* per container.
-#
-# THE LOAD-BEARING FLAG IS --cgroupns=host. Podman starts every container in a
-# PRIVATE cgroup namespace, so an exporter container sees only its OWN cgroup
-# subtree: cadvisor exported a single root series until this flag was added,
-# then it walked the whole tree (996 per-container series measured 2026-10-20).
-# The quadlet unit name is carried in the cgroup path — app.slice/<name>.service/
-# libpod-payload-<hex> — so Prometheus relabels that path into a `container`
-# label (shared/prometheus/prometheus.yml, cadvisor-dankclaw job) and no
-# libpod/podman socket is needed to name containers.
-#
-# BINDING. cadvisor has no per-interface listen flag (unlike node_exporter's
-# --web.listen-address), so it binds 0.0.0.0 INSIDE its own network namespace.
-# The bridge PublishPort is what keeps the unauthenticated endpoint off the
-# host's other interfaces: it publishes exactly loopback (host-local probes and
-# the deployer's health gate) and this host's LAN address (Prometheus, a bridge
-# container, reaches a host-published port at its LAN address — the same netavark
-# rule node-exporter relies on, only here the port really is published).
-#
-# It is rootless on purpose: dank owns the app containers' cgroups, so a
-# rootless cadvisor reads every one of them; root would buy nothing here.
-
-[Unit]
-Description=cAdvisor container metrics (Dankclaw)
-After=network-online.target
-Wants=network-online.target
-
-[Container]
-ContainerName=cadvisor
-Image=gcr.io/cadvisor/cadvisor:v0.49.1@sha256:3cde6faf0791ebf7b41d6f8ae7145466fed712ea6f252c935294d2608b1af388
-PodmanArgs=--pid=host --cgroupns=host
-PublishPort=127.0.0.1:8080:8080
-PublishPort=192.168.0.143:8080:8080
-Volume=/sys:/sys:ro
-Volume=/:/rootfs:ro
-Exec=--housekeeping_interval=30s
-HealthCmd=wget --spider -q http://localhost:8080/healthz
-HealthInterval=30s
-HealthTimeout=10s
-HealthRetries=3
-HealthStartPeriod=40s
-
-[Service]
-Restart=always
-RestartSec=10
-
-[Install]
-WantedBy=default.target
@@ -34,7 +34,7 @@ Wants=network-online.target prometheus.service

 [Container]
 ContainerName=grafana
-Image=docker.io/grafana/grafana:latest@sha256:488984b716c0862639f91a698adbe2862677a5a07bad42da5f03f3030883b205
+Image=docker.io/grafana/grafana:latest
 Network=dankclaw_default
 User=0:0
 PublishPort=127.0.0.1:3002:3000
@@ -47,24 +47,9 @@ Environment=GF_AUTH_GENERIC_OAUTH_CLIENT_SECRET=${GRAFANA_OAUTH_CLIENT_SECRET}
 # client_id query parameter — /application/o/<slug>/authorize/ does not exist.
 Environment=GF_AUTH_GENERIC_OAUTH_AUTH_URL=https://auth.franksdank.cloud/application/o/authorize/
 Environment=GF_AUTH_GENERIC_OAUTH_TOKEN_URL=https://auth.franksdank.cloud/application/o/token/
-# Standard OIDC userinfo. Grafana v13 fetches <api_url>/emails for the address
-# ONLY when the token carries no `email` claim (authentik has no /emails path and
-# 404s it, failing the login) — so the claim must actually be there: the fix is
-# the grafana-email-scope mapping (see scripts/authentik-grafana-sso-fix.sh),
-# not a Grafana flag. Do NOT append ?api=1 to this URL: Grafana then GETs
-# <...>/userinfo/?api=1/emails, a 403.
-Environment=GF_AUTH_GENERIC_OAUTH_API_URL=https://auth.franksdank.cloud/application/o/userinfo/
-# authentik's scopes are openid/email/profile. `user:email` (a GitLab-ism) is
-# not a known scope here; requesting it makes the authorize view crash.
-# Quoted: quadlet word-splits Environment=, so an unquoted spaced list
-# truncates SCOPES to its first token.
-Environment="GF_AUTH_GENERIC_OAUTH_SCOPES=openid email profile"
-Environment=GF_AUTH_GENERIC_OAUTH_EMAIL_ATTRIBUTE_PATH=email
-Environment=GF_AUTH_GENERIC_OAUTH_LOGIN_ATTRIBUTE_PATH=sub
-Environment=GF_AUTH_GENERIC_OAUTH_NAME_ATTRIBUTE_PATH=preferred_username
-# First SSO login auto-provisions the user. Safe: unauthenticated traffic never
-# reaches this callback — the Anubis -> Authentik guard fronts every request.
-Environment=GF_AUTH_GENERIC_OAUTH_ALLOW_SIGN_UP=true
+# ?api=1 = introspection-style userinfo: returns custom claims (groups) too.
+Environment=GF_AUTH_GENERIC_OAUTH_API_URL=https://auth.franksdank.cloud/application/o/userinfo/?api=1
+Environment=GF_AUTH_GENERIC_OAUTH_ALLOW_SIGN_UP=false
 Environment=GF_AUTH_GENERIC_OAUTH_AUTO_UPSERT=true
 Environment=GF_AUTH_GENERIC_OAUTH_ALLOW_ASSIGN_ORG=true
 Environment=GF_AUTH_GENERIC_OAUTH_ORG_ID=1
@@ -33,7 +33,7 @@ Wants=network-online.target

 [Container]
 ContainerName=node-exporter
-Image=docker.io/prom/node-exporter:latest@sha256:1b4e4438faca4dd7e001dd445d161a4a2091b0fededa84093b3a8dfeae1f1be0
+Image=docker.io/prom/node-exporter:latest
 Network=host
 PodmanArgs=--pid=host
 Exec=--path.procfs=/host/proc --path.sysfs=/host/sys --path.rootfs=/rootfs --web.listen-address=127.0.0.1:9100 --web.listen-address=192.168.0.143:9100
@@ -43,16 +43,8 @@ Wants=network-online.target

 [Container]
 ContainerName=prometheus
-Image=docker.io/prom/prometheus:latest@sha256:fa3302279710be2f21a8301ec17a8b4f2301fd829bd113839b3b6b02b6f191a2
+Image=docker.io/prom/prometheus:latest
 Network=dankclaw_default
-# Same-host scrape targets are dialed at host.containers.internal (the podman
-# gateway): a bridge container cannot dial its own host's LAN IP (netavark
-# hairpin gap, measured 2026-10-21), and traffic to the gateway lands on host
-# loopback / published ports (cadvisor/gpu jobs and the blackbox scrape all
-# ride it). The name resolves natively — NO HostAddr key: quadlet 5.7.0 calls
-# it unsupported and the WHOLE user quadlet batch then fails to generate,
-# silently orphaning every other unit (live 2026-10-21: prometheus.service
-# vanished while its container kept running as an orphan).
 User=0:0
 Exec=--config.file=/etc/prometheus/prometheus.yml --storage.tsdb.path=/prometheus --storage.tsdb.retention.time=30d --web.enable-lifecycle
 PublishPort=127.0.0.1:9090:9090
@@ -1,93 +0,0 @@
-# Quadlet unit: open-notebook's SurrealDB (rootless, boot-persistent).
-#
-# Install path on dankclaw: ~/.config/containers/systemd/open-notebook-surrealdb.container
-# Unit name: open-notebook-surrealdb.service (container name: open-notebook-surrealdb)
-#
-# WHAT IT IS. open-notebook's only datastore: notebooks, notes, sources,
-# provider credentials and podcast episodes all live here. The app talks to it
-# over the WebSocket RPC on the private bridge (SURREAL_URL in
-# scripts/open-notebook.env), which is why the container NAME is part of the
-# contract — renaming it breaks the app's database URL, not just this file.
-#
-# IMAGE. `surrealdb/surrealdb:v2` pinned by manifest-list digest, resolved from
-# the registry on 2026-10-18 (that digest is SurrealDB 2.7.0, image built
-# 2026-09-23; `v2` itself is a MOVING tag, so the digest — not the tag — is what
-# makes this reproducible). Upstream open-notebook's own compose file pins the
-# same `v2` tag, and its Dockerfile copies /surreal out of the same image for
-# the retired single-container variant, so v2 is the version this app is built
-# against. Pull=never + the documented pull step (docs: [[Open Notebook]]):
-#
-#   podman pull docker.io/surrealdb/surrealdb:v2@sha256:e5d7387e3727bbe2f7cd6d869e48cd34777272258572ead37c9656b25f745723
-#
-# NO PUBLISHED PORT. 8000 is deliberately not published: the app reaches it by
-# name on systemd-open-notebook, and a published 8000 would put the RPC with
-# the root credentials on every interface (upstream publishes 127.0.0.1:8000
-# "purely for local debugging" — a `podman exec` probe is enough here). The
-# binary binds 0.0.0.0:8000 INSIDE its own namespace via the image's
-# SURREAL_BIND env; without that image default it would bind loopback only and
-# the app could never connect, so no --bind flag belongs here.
-#
-# CREDENTIALS, AND WHY THEY ARE NOT IN THE COMMAND LINE. The `surreal start`
-# CLI reads its options from the environment (`--username` is [env:
-# SURREAL_USER=], `--password` is [env: SURREAL_PASS=] — verified against the
-# pinned image's own `start --help`), so the credentials arrive through the two
-# EnvironmentFile lines and the argv stays clean. Passing `--pass` instead —
-# what upstream's compose does — would put the database password in
-# /proc/<pid>/cmdline, readable by any `ps` on the host. Note the two
-# spellings: the DATABASE reads SURREAL_PASS, the APP reads SURREAL_PASSWORD,
-# and scripts/open-notebook.env declares both because they must hold the same
-# value; the deployer writes that pair when it seeds the host secret file.
-#
-# NO USER= OVERRIDE. The image runs as uid 65532 (Chainguard glibc-dynamic
-# base). With a NAMED volume that works unmodified: podman creates the volume's
-# backing directory owned by the container's user (measured with this digest —
-# the host directory lands on the mapped subuid and the store starts clean),
-# which is exactly why upstream's `user: root` line is not repeated here: that
-# line is commented in upstream as being "required for bind mounts", and this
-# unit has none. The consequence is that the DB files are NOT readable by dank
-# without `podman unshare` — accepted for a datastore nobody edits by hand;
-# episode AUDIO is a host directory for the opposite reason (see the app unit).
-#
-# HEALTH. The base image is distroless — measured against this digest, not even
-# `ls` exists (`exec: "ls": executable file not found in $PATH`), so there is no
-# `/bin/sh`, no curl and no wget. SurrealDB ships its own probe instead:
-# `surreal is-ready` connects to ws://localhost:8000 by default, needs no
-# credentials, prints `OK` and exits 0 while the server is up.
-#
-# It must be handed over as a QUADLET-SIDE JSON ARRAY, not a bare command.
-# podman wraps a bare `HealthCmd` string as `CMD-SHELL` (measured on this host's
-# podman 5.7: `--health-cmd "/surreal is-ready"` stores
-# `{"Test":["CMD-SHELL","/surreal is-ready"]}`), and with no `/bin/sh` in the
-# image that probe fails in ~30 ms with empty output on every interval while the
-# database answers perfectly — which is exactly how this service's first deploy
-# died on 2026-10-19 (`did not become healthy in 90s`). The quoted array survives
-# the generator (`--health-cmd "[\"/surreal\",\"is-ready\"]"`) and stores
-# `["CMD","/surreal","is-ready"]`, which execs directly and goes healthy. Do not
-# "simplify" it back to a bare command.
-
-[Unit]
-Description=SurrealDB (open-notebook datastore)
-After=network-online.target
-Wants=network-online.target
-
-[Container]
-ContainerName=open-notebook-surrealdb
-Image=docker.io/surrealdb/surrealdb:v2@sha256:e5d7387e3727bbe2f7cd6d869e48cd34777272258572ead37c9656b25f745723
-Pull=never
-Network=open-notebook.network
-EnvironmentFile=/home/dank/serverAppConfigs/scripts/open-notebook.env
-EnvironmentFile=/home/dank/Containers/open-notebook/env
-Exec=start --log info rocksdb:/mydata/mydatabase.db
-Volume=open-notebook-db:/mydata
-HealthCmd=["/surreal","is-ready"]
-HealthInterval=30s
-HealthTimeout=10s
-HealthRetries=3
-HealthStartPeriod=20s
-
-[Service]
-Restart=always
-RestartSec=5
-
-[Install]
-WantedBy=default.target
@@ -1,105 +0,0 @@
-# Quadlet unit: open-notebook (rootless, boot-persistent).
-#
-# Install path on dankclaw: ~/.config/containers/systemd/open-notebook.container
-# Unit name: open-notebook.service (container name: open-notebook)
-#
-# WHAT IT IS. The fleet's podcast generator: notes and sources go in, a
-# two-voice (up to four) episode MP3 comes out. One image runs THREE processes
-# under supervisord — the FastAPI API on 5055, the background worker that
-# writes and speaks the episode, and the Next.js UI on 8502 — so this one unit
-# owns the whole app, and the frontend refuses to start until the API answers
-# (the image's own wait-for-api.sh). Runbook: [[Open Notebook]].
-#
-# IMAGE. `lfnovo/open_notebook:v1-latest` pinned by manifest-list digest,
-# resolved from the registry on 2026-10-18; it is the tag upstream's own
-# docker-compose.yml pins for the multi-container shape. Do NOT reach for the
-# `-single` tags: that variant bundles SurrealDB inside the app image, is
-# deprecated upstream, and would give this host two databases. Pull=never, so
-# the pin is fetched by the documented one-time step (docs: [[Open Notebook]]),
-# which the deployer also runs when the image is absent:
-#
-#   podman pull docker.io/lfnovo/open_notebook:v1-latest@sha256:54b2025a4a4853c188f74b9b8246eca8c5480e5dd1babfa92db1038ff2f30212
-#
-# NETWORKS, in this order on purpose. The generated run line keeps file order,
-# so the DEFAULT ROUTE rides dankclaw_default: that is the network where this
-# host has measured that a bridge container can reach a host-namespace
-# listener through the gateway name (the Prometheus scrape of frank-zealot is
-# the proof — see [[Container Network Namespaces]]), and frank-tts lives there
-# too, so the TTS credential can be `http://frank-tts:7772/v1`. The private
-# systemd-open-notebook bridge is second and carries only the database; its
-# traffic is on-link, so losing the default route costs it nothing.
-#
-# PUBLISHED PORTS, and the 5055 collision. Upstream publishes 8502 and 5055,
-# and 5055 is exactly the port seerr already holds on this host
-# (`PublishPort=127.0.0.1:5055:5055`; `ss -ltn` on Dankclaw 2026-10-18 shows
-# 127.0.0.1:5055 taken). So the API leaves the container on host port 5056.
-# Only 8502 is published on all interfaces, because it is the single upstream
-# the guard chain dials: notebook.franksdank.cloud → anubis → authentik →
-# http://:8017 → 192.168.0.143:8502 (dankclaw/Caddyfile), and this Caddy lives
-# in root's storage, so a published port is the only way it can reach a rootless
-# quadlet. The API needs no outward face at all — the image's own Next.js layer
-# rewrites /api/* to the FastAPI beside it, so one origin serves UI and API —
-# and the fleet's rule is that loopback-only is for services whose user path is
-# the guard chain, which this became on 2026-10-19; 5056 stays published on
-# loopback because the deployer's health gate dials it there. That is also why
-# API_URL in scripts/open-notebook.env names the vhost rather than a port: left
-# to itself the frontend auto-detects `<host-you-opened>:5055`, and the UI loads
-# while every call fails. OPEN_NOTEBOOK_PASSWORD is not made redundant by the
-# chain: it is what keeps the API honest for everything else on this host, and
-# with it unset the API answers anyone, including the unauthenticated
-# episode-audio download.
-#
-# VOLUMES. `/app/data` is a named volume: uploads, the LangGraph SQLite
-# checkpoints and the extraction caches the image keeps under it
-# (UV_CACHE_DIR, HF_HOME) are app state nobody browses. The ONE exception is
-# finished audio, which is bind-mounted out of the container at
-# /home/dank/media/podcasts. That mount is interim and reversible: episode
-# destination is an open human-only decision (Gitea issue #7, label
-# `human-only` — agents must not work it), and until it lands, episodes are
-# ordinary files on the host. DATA_FOLDER is hard-coded `./data` in the app
-# (WORKDIR /app) and the DB stores audio paths RELATIVE to data/podcasts, so
-# covering that one subdirectory is enough and removing the line changes
-# nothing else. The image declares no USER, so the app runs as container root,
-# which in this host's rootless mapping (0 -> 1000, measured on this host —
-# see [[Seerr]]) writes those episode files as dank.
-#
-# HEALTH. `GET /health` on the API returns {"status":"healthy"} and is in the
-# password middleware's excluded-path list, so the probe stays honest after the
-# password is set (both read from api/main.py upstream). The image installs
-# curl and ffmpeg in the runtime layer, so the probe needs no shell — kept
-# argument-only for the generator's space-splitting. Note what this probe does
-# NOT prove: the route returns a static body, so it says the API serves, not
-# that the database answers. The DB has its own probe in its unit, and the
-# worker's failure to reach the database shows up as failed jobs, not as an
-# unhealthy container.
-
-[Unit]
-Description=open-notebook (podcast generator: API, worker, web UI)
-After=network-online.target open-notebook-surrealdb.service
-Wants=network-online.target open-notebook-surrealdb.service
-
-[Container]
-ContainerName=open-notebook
-Image=docker.io/lfnovo/open_notebook:v1-latest@sha256:54b2025a4a4853c188f74b9b8246eca8c5480e5dd1babfa92db1038ff2f30212
-Pull=never
-Network=dankclaw_default
-Network=open-notebook.network
-PublishPort=0.0.0.0:8502:8502
-PublishPort=127.0.0.1:5056:5055
-EnvironmentFile=/home/dank/serverAppConfigs/scripts/open-notebook.env
-EnvironmentFile=/home/dank/Containers/open-notebook/env
-Volume=open-notebook-data:/app/data
-Volume=/home/dank/media/podcasts:/app/data/podcasts
-HealthCmd=curl -fsS http://127.0.0.1:5055/health
-HealthInterval=30s
-HealthTimeout=10s
-HealthRetries=3
-HealthStartPeriod=60s
-
-[Service]
-Restart=always
-RestartSec=10
-TimeoutStartSec=900
-
-[Install]
-WantedBy=default.target
@@ -1,34 +0,0 @@
-# Quadlet network: private bridge for the open-notebook pair (rootless).
-#
-# Install path on dankclaw: ~/.config/containers/systemd/open-notebook.network
-# Unit name: open-notebook-network.service (podman network name:
-# systemd-open-notebook — quadlet prefixes "systemd-", so probe the runtime
-# name, never the filename; see [[Container Network Namespaces]]).
-#
-# WHY A PRIVATE BRIDGE. The app reaches its database by container name
-# (SURREAL_URL=ws://open-notebook-surrealdb:8000/rpc), and container DNS names
-# resolve only on the network that owns them. Putting SurrealDB here instead of
-# on dankclaw_default keeps the datastore reachable ONLY to the container that
-# owns it: 8000 is not published (an unpublished port is refused outright from
-# every other namespace anyway), and nothing else on the fleet-wide network can
-# dial it with the root credentials. Same reasoning as the gitea and authentik
-# database pairs.
-#
-# Kernel bridge, not slirp4netns: the 2026-08-22 open-webui incident (a dead
-# per-container userspace helper left a container that accepted TCP and reset
-# every request for two days, logging nothing). A bridge has no such helper.
-#
-# Subnet registry on dankclaw (rootless):
-#   10.88.0.0/16  default "podman" bridge
-#   10.89.0.0/24  gitea (systemd-gitea)
-#   10.89.1.0/24  compose stack (dankclaw_default)
-#   10.89.2.0/24  searxng (systemd-searxng)
-#   10.89.3.0/24  open-webui (systemd-open-webui)
-#   10.89.4.0/24  open-notebook (systemd-open-notebook)  <- this file
-# Pinned so the container IPs stay deterministic across reboots, like the
-# other pinned nets; 10.89.4.0/24 was the next unclaimed /24 in this registry.
-
-[Network]
-Driver=bridge
-Subnet=10.89.4.0/24
-Gateway=10.89.4.1
@@ -27,7 +27,7 @@ After=network.target
 [Container]
 ContainerName=open-webui
 Network=open-webui.network
-Image=ghcr.io/open-webui/open-webui:main@sha256:0b399a8ba34558e181d316750455f2a95376a38cd60a61b9c942e138898944bd
+Image=ghcr.io/open-webui/open-webui:main
 PublishPort=3005:8080
 Environment=OPENAI_API_BASE_URL=http://192.168.0.143:7770
 Environment=ANONYMIZED_TELEMETRY=False
@@ -45,4 +45,3 @@
 192.168.0.143 pihole.franksdank.cloud
 192.168.0.143 seerr.franksdank.cloud
 192.168.0.143 marreta.franksdank.cloud
-192.168.0.143 notebook.franksdank.cloud
@@ -24,7 +24,7 @@ Wants=network-online.target

 [Container]
 ContainerName=pihole
-Image=docker.io/pihole/pihole:latest@sha256:d34134716fc32038b052a8674c7c6879e0ff21f0e6d8af345166fffa2c27c113
+Image=docker.io/pihole/pihole:latest
 Network=slirp4netns
 PublishPort=0.0.0.0:53:53/udp
 PublishPort=0.0.0.0:53:53/tcp
@@ -16,7 +16,7 @@ After=searxng-network.service network.target
 [Container]
 ContainerName=gluetun
 Network=searxng.network
-Image=docker.io/qmcgaw/gluetun:latest@sha256:4af0c804003481e100180e1f8d4d5e9c48163fedc0f88d061881b615aef9d183
+Image=docker.io/qmcgaw/gluetun:latest
 AddCapability=NET_ADMIN
 AddDevice=/dev/net/tun:/dev/net/tun
 Environment=VPN_SERVICE_PROVIDER=${VPN_SERVICE_PROVIDER}
@@ -16,7 +16,7 @@ After=searxng-network.service network.target
 ContainerName=searxng-valkey
 Network=searxng.network
 NetworkAlias=valkey
-Image=docker.io/valkey/valkey:9-alpine@sha256:bf09d58e96d6b4524a15396b65b242bbe18beebbcc15a04d8ee75ddac62fae0b
+Image=docker.io/valkey/valkey:9-alpine
 Volume=/opt/searxng/valkey:/data

 [Service]
@@ -16,7 +16,7 @@ After=gluetun.service searxng-valkey.service network.target
 [Container]
 ContainerName=searxng
 Network=gluetun.container
-Image=ghcr.io/searxng/searxng:latest@sha256:f4aa660c54d062dedb268be84e821ea24e39f48c7a496af8dd02cc9eedb93847
+Image=ghcr.io/searxng/searxng:latest
 Environment=SEARXNG_SECRET=${SEARXNG_SECRET}
 Environment=SEARXNG_BASE_URL=${SEARXNG_BASE_URL}
 Volume=/opt/searxng/config:/etc/searxng
@@ -29,7 +29,7 @@ Wants=network-online.target

 [Container]
 ContainerName=seerr
-Image=ghcr.io/seerr-team/seerr:v3.5.1@sha256:d8fab675b55ad293ded98fd71efcfd46054cbad9746534ae1cc4794a88c0b3be
+Image=ghcr.io/seerr-team/seerr:v3.4.1
 Network=dankclaw_default
 PublishPort=127.0.0.1:5055:5055
 Volume=/home/dank/Containers/seerr/data:/app/config
@@ -1,26 +0,0 @@
-# /etc/udev/rules.d/98-halogen-npu-access.rules
-#
-# Why the NPU node is world-writable: `frank-dragoon` runs rootless podman under a
-# unit with `User=dank`, and the container's user namespace maps exactly one uid and
-# one gid (dank). `/dev/accel/accel0` is `root:render 0660`, so inside the container
-# the node reads `nobody:nogroup` and the engine dies with its own diagnosis:
-#
-#   halogen npu: /dev/accel/accel0 is here, but this process cannot open it
-#     (crw-rw---- nobody:nogroup).
-#
-# `--group-add keep-groups` is the vendor's advice and it does NOT work in this
-# shape: measured 2026-10-19 from the host as a `dank` shell that was already in
-# `render`, with and without the flag, the node still read `nobody:nogroup` and the
-# open was refused — an unmapped gid cannot be granted to a process inside a
-# namespace that does not map it. Same conclusion `99-comfyui-gpu-access.rules`
-# reached for `/dev/kfd` from the systemd-user-spawn direction: the mode is the only
-# spawn-context-independent door short of running the stack rootful.
-#
-# Blast radius: any process in this host's single uid domain (every rootless
-# container here shares it) can submit to the XDNA2 NPU. The NPU holds no user data
-# and no privileged context — the same trade already accepted for the GPU compute
-# nodes beside it. Installed by `deploy-remote.sh` when the NPU lane is armed and left
-# in place afterwards — the mode costs nothing while the lane is idle. `serve-frank-dragoon.sh`
-# refuses to arm the lane on a node that is not other-accessible, so a host that lost this
-# file fails loudly at boot instead of crash-looping.
-KERNEL=="accel[0-9]*", SUBSYSTEM=="accel", MODE="0666"
@@ -1,43 +0,0 @@
-# frank-tts — the cloned-voice speech façade (dankclaw/voiceservice/app.py).
-#
-# Standard library only, so there is nothing to pip-install and nothing to
-# compile: the image is a python runtime plus ffmpeg (the one binary the app
-# shells out to, for response_format=mp3) plus one COPY of app.py. Rebuild to
-# ship a new app.py.
-#
-# Build on dankclaw (deploy-remote.sh build_compose_image frank-tts does this,
-# including --format docker, which the fleet's other local builds also pass):
-#   podman build --format docker -t franksdank/frank-tts:latest \
-#       -f dankclaw/voiceservice/Containerfile .
-#
-# BASE PINNED BY DIGEST, like every other image this fleet runs from a
-# registry: a tag moves under a rebuild and a voice endpoint that changes
-# python or ffmpeg mid-week is a debug session nobody asked for. The digest is
-# the multi-arch INDEX of python:3.12-alpine as published 2026-10-18 (amd64 +
-# arm64 on linux); Dankclaw resolves the amd64 entry.
-#
-# WHY 3.12: the fleet's ceiling for this app is python <= 3.12 (single file,
-# no build step); the alpine flavour also keeps the image in the tens of MB
-# and ships busybox `wget`, which is what the quadlet unit's HealthCmd runs —
-# so no HEALTHCHECK belongs here (the probe lives in the unit, as in
-# alertmanager/prometheus; only distroless bases need it baked in).
-FROM docker.io/library/python:3.12-alpine@sha256:1b668429b3511ab407d8e00648891631b0b1a4d7e15e3ca70f38ab5b91ad4ab4
-
-# ffmpeg: mp3 output. The app always asks the upstream for wav and re-encodes
-# on the way out, so ffmpeg is the only external program in the picture.
-RUN apk add --no-cache ffmpeg
-
-# Unbuffered stdout: the unit's journal must show the "listening on" line (and
-# any backend error) the moment it happens, not when the buffer fills.
-ENV PYTHONUNBUFFERED=1
-
-COPY dankclaw/voiceservice/app.py /app/app.py
-
-# Non-root. The voice packs are bind-mounted read-only, so this user only ever
-# reads them — they must be world-readable (the pack builder writes 644/755).
-# The one thing it writes is its per-request temp dir under /tmp.
-RUN adduser -D -u 10000 -h /app -s /sbin/nologin shim
-USER shim
-
-EXPOSE 7772
-CMD ["python3", "/app/app.py"]
@@ -1,178 +0,0 @@
-# frank-tts-engine — the arm A TTS engine: qwentts.cpp's resident `tts-server`
-# (the CPU binary that frank-tts on :7772 forwards to). The runbook is
-# [[Frank TTS Engine]]; the fleet-facing endpoint is [[Frank TTS]].
-#
-# WHAT THIS IS. The upstream source is NOT vendored here: the image clones
-# ServeurpersoCom/qwentts.cpp at one pinned commit and builds it. Rebuilding is
-# therefore a network operation (github.com plus objects.githubusercontent.com
-# for the `ggml` submodule), and build_compose_image in scripts/deploy-remote.sh
-# skips it whenever this Containerfile's sha256 label already matches the built
-# image — the same offline-safe idiom as loki/alloy/comfyui, so a re-deploy of
-# an unchanged engine needs no network.
-#
-# THE PINNED COMMIT IS THE CONTRACT. 5916bf7c79dbb4e66fc059cdc507a79a23a28bb4
-# is the commit the 2026-10-18 CPU trial on Dankclaw measured (server-side TTFA
-# 94-112 ms, warm RTF 0.55-0.67, run spread +-0.01 — the numbers that won the
-# engine decision). Moving it means re-measuring, not just editing this line.
-# `ggml` is a submodule of the repo (the only one; vendor/cpp-httplib and
-# vendor/yyjson are ordinary directories in the tree), and the submodule commit
-# is recorded in the parent tree, so cloning the parent commit recursively is
-# reproducible without a second pin.
-#
-# THE BUILD LINE IS THE ONE PROVEN ON DANKCLAW, verbatim — with the Vulkan door
-# open since 2026-10-19. That configure was run in this exact base image and
-# answers `Found Vulkan: /usr/lib/x86_64-linux-gnu/libvulkan.so (found version
-# "1.3.275") found components: glslc` then `Including Vulkan backend`:
-#
-#   cmake -B build -DCMAKE_BUILD_TYPE=Release -DGGML_CCACHE=OFF \
-#       -DGGML_BACKEND_DL=OFF -DGGML_CPU_ALL_VARIANTS=OFF -DGGML_OPENMP=ON \
-#       -DGGML_BLAS=OFF -DGGML_CUDA=OFF -DGGML_HIP=OFF -DGGML_VULKAN=ON \
-#       -DGGML_METAL=OFF -DGGML_SYCL=OFF -DGGML_RPC=OFF \
-#   && cmake --build build -j 8
-#
-# Every flag is load-bearing, so do not "tidy" it:
-#   GGML_BLAS=OFF is FORCED, not a preference — Dankclaw has libopenblas.so.0
-#     but no BLAS dev headers, so the upstream default (BLAS on) fails to
-#     configure. Upstream's own Dockerfile and buildcpu.sh both pass BLAS on,
-#     which is why this file does not copy them.
-#   GGML_CCACHE=OFF because a container build layer has no persistent ccache
-#     directory to hit; on, it only writes a useless one.
-#   GGML_OPENMP=ON is what makes the runtime need libgomp1 (installed below).
-#   GGML_BACKEND_DL=OFF / GGML_CPU_ALL_VARIANTS=OFF: upstream's images build
-#     every x86 ISA as a loadable backend so one image runs anywhere. This one
-#     is built on the machine that runs it, so the single native CPU backend is
-#     what was measured. That is also why no -DGGML_NATIVE is passed: ggml
-#     defaults it ON outside wasm/cross-compiles, i.e. -march=native for THIS
-#     host — correct while build host == run host, wrong if anyone ever builds
-#     this image elsewhere, which is why the fleet builds it on Dankclaw.
-#   CUDA/HIP/METAL/SYCL/RPC stay OFF, VULKAN is ON: the endpoint computes on the
-#     gfx1151 iGPU (decision and co-tenancy gates:
-#     report/tts-npu-spike-verdict-2026-10-19.md, shape in [[Frank TTS Engine]]).
-#     Nothing else is available to this image: no ROCm userland, no NVIDIA
-#     userland, no macOS, no oneAPI, and RPC would add a second process to
-#     supervise on a box that already has one engine per card.
-#
-# BUILD SHARED LIBS, RUNTIME SMALL. BUILD_SHARED_LIBS defaults ON on Linux, so
-# tts-server links libggml*.so rather than swallowing them; the runtime stage
-# copies the binary and its shared objects out of the build tree and points
-# LD_LIBRARY_PATH at them, exactly as upstream's own Dockerfile does (the RPATH
-# baked at link time names the build directory, which does not exist here).
-#
-# BASE PINNED BY DIGEST, like every other image this fleet builds: the digest
-# is the multi-arch INDEX of ubuntu:24.04 as resolved from Docker Hub on
-# 2026-10-18; Dankclaw takes the amd64 entry. One base for both stages so there
-# is exactly one thing to re-resolve.
-#
-# NO HEALTHCHECK IN THIS FILE. The probe lives in the Quadlet unit
-# (HealthCmd=), which is the fleet pattern for anything with a shell — see the
-# alertmanager/prometheus units, and the opposite case in dankclaw/loki, whose
-# distroless image has no /bin/sh for the generator's CMD-SHELL wrapping.
-# `curl` is therefore installed in the runtime stage, for that probe only.
-
-ARG BASE_IMAGE=docker.io/library/ubuntu:24.04@sha256:68ee22004a9109b44973313b02ff7b1de250f1de0f5ecb628d73f579aaff33ac
-
-# ------------------------------------------------------------------ build
-FROM ${BASE_IMAGE} AS build
-
-# The three Vulkan build deps are not optional and not interchangeable:
-#   libvulkan-dev    vulkan/vulkan.h + the VulkanHeaders/VulkanLoader CMake
-#                    configs, i.e. what find_package(Vulkan) resolves to.
-#   glslc            ggml compiles its shaders with glslc, not glslangValidator.
-#                    Ubuntu's glslang-tools ships glslangValidator and NOT glslc,
-#                    so without this package `find_package(Vulkan COMPONENTS
-#                    glslc REQUIRED)` fails and there is no Vulkan backend.
-#                    The archive's glslc is old (2023.8): it compiles
-#                    GL_KHR_cooperative_matrix — the path ggml's quantized
-#                    matmul needs — but reports cooperative_matrix2,
-#                    integer_dot_product, bfloat16 and the micro-float
-#                    extensions as unsupported, so those shader families are
-#                    compiled out. If a Vulkan prefill ever reads as
-#                    inexplicably slow, suspect this compiler before the device;
-#                    the fix is a pinned glslang build passed as
-#                    -DVulkan_GLSLC_EXECUTABLE=<path>, not a different runtime.
-#   spirv-headers    find_package(SPIRV-Headers CONFIG REQUIRED); the distro
-#                    package ships /usr/share/cmake/SPIRV-Headers/.
-RUN apt-get update -qq \
-    && apt-get install -y -qq --no-install-recommends \
-        ca-certificates cmake g++ git make \
-        libvulkan-dev glslc spirv-headers \
-    && rm -rf /var/lib/apt/lists/*
-
-# Checked out detached at a full SHA, then the recorded submodule. No branch
-# name anywhere: a branch moves, this commit does not.
-ARG QWENTTS_COMMIT=5916bf7c79dbb4e66fc059cdc507a79a23a28bb4
-RUN git clone --no-checkout https://github.com/ServeurpersoCom/qwentts.cpp /src \
-    && git -C /src checkout --detach "${QWENTTS_COMMIT}" \
-    && git -C /src submodule update --init --depth 1 \
-    && git -C /src rev-parse HEAD \
-    && git -C /src submodule status
-
-WORKDIR /src
-RUN cmake -B build -DCMAKE_BUILD_TYPE=Release \
-        -DGGML_CCACHE=OFF \
-        -DGGML_BACKEND_DL=OFF \
-        -DGGML_CPU_ALL_VARIANTS=OFF \
-        -DGGML_OPENMP=ON \
-        -DGGML_BLAS=OFF \
-        -DGGML_CUDA=OFF \
-        -DGGML_HIP=OFF \
-        -DGGML_VULKAN=ON \
-        -DGGML_METAL=OFF \
-        -DGGML_SYCL=OFF \
-        -DGGML_RPC=OFF \
-    && cmake --build build -j 8
-
-# ---------------------------------------------------------------- runtime
-FROM ${BASE_IMAGE}
-
-# libgomp1: the OpenMP runtime the build above links against. curl: the unit's
-# HealthCmd. ca-certificates: so that curl can speak https if anyone ever
-# points a probe at a URL with a scheme this host terminates elsewhere.
-#
-# libvulkan1 + mesa-vulkan-drivers are what make `--device Vulkan0` mean
-# anything at runtime: libvulkan1 is the loader libggml-vulkan.so dlopens, and
-# mesa-vulkan-drivers is the RADV ICD that claims the gfx1151. Loader without an
-# ICD enumerates ZERO devices, and tts-server then dies in device lookup AFTER
-# both GGUFs have been read — the journal reads as a model fault, not a missing
-# driver. Enumerated from this base on 2026-10-19 with `--device /dev/dri
-# --group-add keep-groups` plus the host's ComfyUI udev rule: "Radeon 8060S
-# Graphics (RADV GFX1151)", API 1.4.318 (the heap figure is owned by
-# [[Frank TTS Engine]]; it is the GTT-backed aperture, not the ~2 GiB DRM
-# carve-out). No NVIDIA ICD is installed on purpose: one Vulkan device in the
-# enumeration is what makes the NAME Vulkan0 safe to write down.
-RUN apt-get update -qq \
-    && apt-get install -y -qq --no-install-recommends \
-        ca-certificates curl libgomp1 libvulkan1 mesa-vulkan-drivers \
-    && rm -rf /var/lib/apt/lists/*
-
-# Same unprivileged uid as the frank-tts façade (10000): the voice packs and the
-# model directory are host state mounted read-only, and one uid across the pair
-# means one thing to remember about what is world-readable.
-#
-# `useradd`, NOT `adduser`: this trimmed ubuntu:24.04 base ships no `adduser` at
-# all (verified 2026-10-18 against this digest — `adduser` is absent from the
-# base and is not pulled in by ca-certificates/curl/libgomp1 either), so an
-# `adduser` line dies at image-build time with crun's "executable file `adduser`
-# not found in $PATH". `useradd --system --uid 10000` succeeds and prints one
-# cosmetic warning (`uid 10000 is greater than SYS_UID_MAX 999`); the uid is the
-# point, and it is what matches the façade. The alpine façade Containerfile keeps
-# its own busybox `adduser -D -u 10000` — different base, different tool, same uid.
-RUN useradd --system --uid 10000 --home-dir /app --shell /usr/sbin/nologin tts
-
-WORKDIR /app
-COPY --from=build /src/build/tts-server /src/build/*.so* /app/
-COPY dankclaw/voiceservice/tts-engine-start.sh /app/start.sh
-# The boot warmer lives INSIDE the image because the engine publishes no port:
-# the unit reaches it with `podman exec`, and the script needs the same curl,
-# base64 and /voices the engine's own namespace has. It is a sibling of
-# start.sh, not a second entrypoint — CMD stays the server.
-COPY dankclaw/voiceservice/tts-engine-warm.sh /app/warm-voices.sh
-RUN chmod 755 /app/start.sh /app/warm-voices.sh
-
-# The binary's RPATH points inside the build tree; this is what makes it find
-# the libggml*.so copied beside it.
-ENV LD_LIBRARY_PATH=/app
-
-USER tts
-EXPOSE 8880
-CMD ["/app/start.sh"]
@@ -1,424 +0,0 @@
-#!/usr/bin/env python3
-"""frank-tts — the fleet's OpenAI-compatible speech façade (component: the shim).
-
-One local endpoint that answers ``voice=<name>`` with audio in a cloned voice.
-Consumers (the omp harness, OpenWebUI, open-notebook) speak the OpenAI
-``POST /v1/audio/speech`` shape and never learn a file path: a voice is a
-directory under ``$VOICES_DIR`` and ``reference.wav`` inside it is what makes
-the name real. A name that has no reference is not a voice — the shim answers
-400 with the list that is, instead of silently narrating in a fallback voice
-(silent fallback is how a wrong-voice endpoint stays wrong for weeks). A request
-that names NO voice at all takes ``$DEFAULT_VOICE``: that is a spoken choice in
-the env file rather than a surprise in the code, and the journal marks it
-``(default)`` so a name nobody asked for is never mistaken for one that was.
-
-Two backends sit behind the one surface, chosen by ``TTS_BACKEND``:
-
-``omni-http``
-    Forward to an OpenAI-shaped upstream TTS server (``$TTS_UPSTREAM``) and
-    carry the cloned voice with the request as ``reference_audio`` (base64 of
-    ``reference.wav``). The upstream does the inference; this process only
-    resolves the voice and, when asked, re-encodes to mp3 with ``ffmpeg``.
-
-``llama-tts-cli``
-    Run a local binary per request (``$LLAMA_TTS_BIN``), hand it the
-    reference with ``--tts-speaker-file``, read the WAV it writes. ``nice -n
-    19`` because this box also serves text and a voice request must never
-    starve it.
-
-Everything comes from the environment, read at REQUEST time (no
-import-time-only globals): the same module can be imported with an injected
-``Config`` or run with ``python3 app.py`` and the unit's env file.
-
-Runtime law: single process, standard library only, no build step. Audio
-bytes never reach the log — request lines and error text only.
-"""
-
-from __future__ import annotations
-
-import argparse
-import base64
-import json
-import os
-import re
-import subprocess
-import sys
-import tempfile
-import urllib.error
-import urllib.parse
-import urllib.request
-from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
-from pathlib import Path
-
-WAV_CONTENT_TYPE = "audio/wav"
-MP3_CONTENT_TYPE = "audio/mpeg"
-
-# A voice name is a plain directory name. Slashes and `..` are refused here so
-# a client-supplied name can never escape $VOICES_DIR.
-VOICE_NAME_RE = re.compile(r"[A-Za-z0-9][A-Za-z0-9._-]{0,63}")
-
-
-class ShimError(Exception):
-    """A failure with the HTTP status the client should see."""
-
-    def __init__(self, status, message):
-        super().__init__(message)
-        self.status = status
-        self.message = message
-
-
-class Config:
-    """Every knob of the shim. Constructed from the environment per request by
-    default, or injected into ``create_server`` for tests and embedding."""
-
-    def __init__(self, voices_dir, backend, upstream, upstream_model, llama_bin,
-                 llama_model, llama_mmproj, llama_lang, llama_threads, port,
-                 max_chars, timeout, default_voice=""):
-        self.voices_dir = Path(voices_dir)
-        self.backend = backend
-        self.upstream = upstream
-        self.upstream_model = upstream_model
-        self.llama_bin = llama_bin
-        self.llama_model = llama_model
-        self.llama_mmproj = llama_mmproj
-        self.llama_lang = llama_lang
-        self.llama_threads = llama_threads
-        self.port = port
-        self.max_chars = max_chars
-        self.timeout = timeout
-        self.default_voice = default_voice
-
-    @classmethod
-    def from_env(cls, environ=None):
-        env = os.environ if environ is None else environ
-        return cls(
-            voices_dir=env.get("VOICES_DIR", "/voices"),
-            backend=(env.get("TTS_BACKEND") or "omni-http").strip().lower(),
-            upstream=(env.get("TTS_UPSTREAM") or "http://host.containers.internal:8880").rstrip("/"),
-            upstream_model=env.get("TTS_UPSTREAM_MODEL") or "voxcpm2",
-            llama_bin=env.get("LLAMA_TTS_BIN") or "",
-            llama_model=env.get("LLAMA_TTS_MODEL") or "",
-            llama_mmproj=env.get("LLAMA_TTS_MMPROJ") or "",
-            llama_lang=env.get("LLAMA_TTS_LANG") or "english",
-            llama_threads=_int(env, "LLAMA_TTS_THREADS", 8),
-            port=_int(env, "SHIM_PORT", 7772),
-            max_chars=_int(env, "SHIM_MAX_CHARS", 15000),
-            timeout=float(_int(env, "SHIM_TIMEOUT", 600)),
-            default_voice=(env.get("DEFAULT_VOICE") or "").strip(),
-        )
-
-
-def _int(env, name, default):
-    try:
-        return int(str(env.get(name, "")).strip())
-    except (TypeError, ValueError):
-        return default
-
-
-def list_voices(voices_dir):
-    """Voice names in ``voices_dir``: directories that hold a ``reference.wav``.
-
-    A missing or unreadable directory is an empty fleet, not an error — the
-    health endpoint must still answer so the unit reports its state.
-    """
-    try:
-        entries = sorted(p.name for p in voices_dir.iterdir()
-                         if p.is_dir() and (p / "reference.wav").is_file())
-    except OSError:
-        return []
-    return entries
-
-
-def resolve_voice(voices_dir, voice):
-    """``$VOICES_DIR/<voice>/reference.wav`` for a valid name, else ``None``."""
-    if not isinstance(voice, str) or not VOICE_NAME_RE.fullmatch(voice):
-        return None
-    reference = voices_dir / voice / "reference.wav"
-    return reference if reference.is_file() else None
-
-
-def to_mp3(wav_bytes, timeout):
-    """Re-encode WAV bytes to mp3 with the image's ``ffmpeg`` (pipes, no temp
-    file: the audio never lands on disk to be left behind)."""
-    argv = ["ffmpeg", "-hide_banner", "-loglevel", "error", "-i", "pipe:0",
-            "-f", "mp3", "pipe:1"]
-    try:
-        proc = subprocess.run(argv, input=wav_bytes, capture_output=True,
-                              timeout=timeout)
-    except FileNotFoundError:
-        raise ShimError(500, "response_format=mp3 needs the ffmpeg binary, which is not on PATH")
-    except subprocess.TimeoutExpired:
-        raise ShimError(502, "ffmpeg mp3 conversion timed out after %gs" % timeout)
-    if proc.returncode != 0 or not proc.stdout:
-        raise ShimError(502, "ffmpeg mp3 conversion failed (exit %s): %s"
-                        % (proc.returncode, _tail(proc.stderr)))
-    return proc.stdout
-
-
-def _tail(raw):
-    """Last line of a subprocess' stderr, for an error message. Text only —
-    audio bytes are never echoed, and never logged."""
-    text = (raw or b"").decode("utf-8", "replace").strip()
-    return text.splitlines()[-1][:300] if text else "no output"
-
-
-def _error(message, voices=None):
-    body = {"error": {"message": message, "type": "invalid_request_error"}}
-    if voices is not None:
-        body["error"]["available_voices"] = voices
-    return body
-
-
-class SpeechHandler(BaseHTTPRequestHandler):
-    protocol_version = "HTTP/1.1"
-    server_version = "frank-tts"
-    sys_version = ""
-
-    # -- plumbing ----------------------------------------------------------
-    def config(self):
-        return getattr(self.server, "config", None) or Config.from_env()
-
-    def log_message(self, fmt, *args):
-        # One prefix for the journal. Request lines and outcomes only: a
-        # synthesized text, a base64 reference voice and the audio bytes
-        # itself never reach stderr, in success or in failure.
-        sys.stderr.write("frank-tts: %s - %s\n"
-                         % (self.address_string(), fmt % args))
-
-    def log_error(self, fmt, *args):
-        # The base class routes refused requests through a separate sink; keep
-        # them on the same line format so the journal reads uniformly.
-        self.log_message(fmt, *args)
-
-    def _json(self, status, obj, head=False):
-        self._bytes(status, json.dumps(obj).encode("utf-8"), "application/json", head)
-
-    def _bytes(self, status, data, content_type, head=False):
-        self.send_response(status)
-        self.send_header("Content-Type", content_type)
-        self.send_header("Content-Length", str(len(data)))
-        self.end_headers()
-        if not head:
-            self.wfile.write(data)
-
-    def _drain(self):
-        """Read and discard a body this route will not answer. HTTP/1.1
-        keep-alive makes an unread body the NEXT request line, so a client
-        that reuses the connection would see a parse error it never sent."""
-        if getattr(self, "_body_read", False):
-            return  # already consumed; a second read would block the response
-        try:
-            length = int(self.headers.get("Content-Length") or 0)
-        except ValueError:
-            return
-        if length > 0:
-            self.rfile.read(length)
-
-    def _body(self):
-        """Parse the JSON request body. Never returned to a log or an error."""
-        try:
-            length = int(self.headers.get("Content-Length") or 0)
-        except ValueError:
-            raise ValueError("Content-Length is not a number")
-        raw = self.rfile.read(length) if length > 0 else b""
-        self._body_read = True
-        if not raw:
-            raise ValueError("expected a JSON body")
-        parsed = json.loads(raw.decode("utf-8"))
-        if not isinstance(parsed, dict):
-            raise ValueError("expected a JSON object")
-        return parsed
-
-    # -- routes ------------------------------------------------------------
-    def do_GET(self, head=False):
-        # HEAD shares the route table: `wget --spider` (the fleet's probe
-        # shape) asks for HEAD, and a 501 there would read as a dead service.
-        path = urllib.parse.urlparse(self.path).path
-        cfg = self.config()
-        if path == "/health":
-            self._json(200, {"status": "ok", "backend": cfg.backend,
-                             "voices": list_voices(cfg.voices_dir)}, head)
-        elif path == "/v1/audio/voices":
-            self._json(200, list_voices(cfg.voices_dir), head)
-        else:
-            self._json(404, _error("no such endpoint: %s" % path), head)
-
-    def do_HEAD(self):
-        self.do_GET(head=True)
-
-    def do_POST(self):
-        path = urllib.parse.urlparse(self.path).path
-        if path != "/v1/audio/speech":
-            self._drain()
-            self._json(404, _error("no such endpoint: %s" % path))
-            return
-        cfg = self.config()
-        try:
-            body = self._body()
-        except (ValueError, UnicodeDecodeError) as exc:
-            self._drain()
-            self._json(400, _error("malformed request body: %s" % exc))
-            return
-
-        text = body.get("input")
-        if not isinstance(text, str) or not text.strip():
-            self._json(400, _error("`input` must be a non-empty string"))
-            return
-        if len(text) > cfg.max_chars:
-            self._json(414, _error("input is %d characters, over the %d limit "
-                                   "(SHIM_MAX_CHARS); split it or raise the limit"
-                                   % (len(text), cfg.max_chars)))
-            return
-
-        fmt = str(body.get("response_format") or "wav").strip().lower()
-        if fmt not in ("wav", "mp3"):
-            self._json(400, _error("unsupported response_format %r; this "
-                                   "endpoint serves wav and mp3" % fmt))
-            return
-
-        # `voice_id` is the alias the omp harness forwards for `voice`. A request
-        # that names nothing takes the declared default — the harness has no
-        # setting that feeds a voice, so voice-less calls are the normal shape
-        # there, not a client bug.
-        requested = body.get("voice") or body.get("voice_id")
-        defaulted = not requested and bool(cfg.default_voice)
-        voice = requested or cfg.default_voice
-        reference = resolve_voice(cfg.voices_dir, voice)
-        if reference is None:
-            available = list_voices(cfg.voices_dir)
-            listed = ", ".join(available) if available else "(none: no voice pack with a reference.wav in %s)" % cfg.voices_dir
-            self._json(400, _error("unknown voice %r; available voices: %s"
-                                   % (voice, listed), available))
-            return
-
-        model = str(body.get("model") or cfg.upstream_model)
-        try:
-            if cfg.backend == "omni-http":
-                audio = self._speech_http(cfg, model, text, str(voice), reference)
-            elif cfg.backend == "llama-tts-cli":
-                audio = self._speech_cli(cfg, text, reference)
-            else:
-                raise ShimError(500, "unknown TTS_BACKEND %r (expected omni-http or llama-tts-cli)" % cfg.backend)
-            if fmt == "mp3":
-                audio = to_mp3(audio, cfg.timeout)
-        except ShimError as exc:
-            self._json(exc.status, _error(exc.message))
-            return
-
-        content_type = MP3_CONTENT_TYPE if fmt == "mp3" else WAV_CONTENT_TYPE
-        self._bytes(200, audio, content_type)
-        self.log_message("speech voice=%s%s format=%s chars=%d -> %d bytes %s",
-                         voice, " (default)" if defaulted else "", fmt,
-                         len(text), len(audio), content_type)
-
-    # -- backends ----------------------------------------------------------
-    def _speech_http(self, cfg, model, text, voice, reference):
-        """Forward to the upstream with the cloned voice riding along as
-        base64 ``reference_audio``. The upstream is asked for wav always; mp3
-        is our own re-encode on the way out."""
-        try:
-            payload = {
-                "model": model,
-                "input": text,
-                "voice": voice,
-                "response_format": "wav",
-                "reference_audio": base64.b64encode(reference.read_bytes()).decode("ascii"),
-            }
-        except OSError as exc:
-            raise ShimError(500, "cannot read %s: %s" % (reference, exc))
-
-        url = cfg.upstream + "/v1/audio/speech"
-        request = urllib.request.Request(
-            url, data=json.dumps(payload).encode("utf-8"),
-            headers={"Content-Type": "application/json"}, method="POST")
-        try:
-            with urllib.request.urlopen(request, timeout=cfg.timeout) as response:
-                audio = response.read()
-        except urllib.error.HTTPError as exc:
-            detail = exc.read(512).decode("utf-8", "replace").strip()
-            raise ShimError(502, "upstream %s returned HTTP %d: %s"
-                            % (url, exc.code, detail or "no detail"))
-        except urllib.error.URLError as exc:
-            raise ShimError(502, "upstream %s unreachable: %s" % (url, exc.reason))
-        except (TimeoutError, OSError) as exc:
-            raise ShimError(502, "upstream %s failed: %s" % (url, exc))
-        if not audio:
-            raise ShimError(502, "upstream %s returned no audio" % url)
-        return audio
-
-    def _speech_cli(self, cfg, text, reference):
-        """Run the local TTS binary once, in a temp dir that is always removed
-        — also on failure, or a crash-looping binary fills /tmp."""
-        if not cfg.llama_bin or not cfg.llama_model:
-            raise ShimError(500, "TTS_BACKEND=llama-tts-cli needs LLAMA_TTS_BIN and LLAMA_TTS_MODEL")
-        work = tempfile.TemporaryDirectory(prefix="frank-tts-")
-        try:
-            out_path = Path(work.name) / "speech.wav"
-            argv = ["nice", "-n", "19", "env", "OMP_NUM_THREADS=%d" % cfg.llama_threads,
-                    cfg.llama_bin, "-m", cfg.llama_model]
-            if cfg.llama_mmproj:
-                argv += ["-mm", cfg.llama_mmproj]
-            argv += ["--tts-lang", cfg.llama_lang,
-                     "--tts-speaker-file", str(reference),
-                     "-p", text,
-                     "--output", str(out_path)]
-            try:
-                proc = subprocess.run(argv, cwd=work.name, capture_output=True,
-                                      timeout=cfg.timeout)
-            except FileNotFoundError as exc:
-                raise ShimError(500, "cannot run the TTS command: %s" % exc)
-            except subprocess.TimeoutExpired:
-                raise ShimError(502, "TTS binary timed out after %gs" % cfg.timeout)
-            if proc.returncode != 0:
-                raise ShimError(502, "TTS binary failed (exit %s): %s"
-                                % (proc.returncode, _tail(proc.stderr)))
-            if not out_path.is_file():
-                raise ShimError(502, "TTS binary exited 0 but wrote no %s" % out_path.name)
-            return out_path.read_bytes()
-        finally:
-            work.cleanup()
-
-
-class SpeechServer(ThreadingHTTPServer):
-    """Threads so a health probe answers while a long synthesis runs; daemon
-    threads so SIGTERM is never held hostage by a stuck request."""
-
-    daemon_threads = True
-    allow_reuse_address = True
-    config = None
-
-
-def create_server(config=None, host="0.0.0.0", port=None):
-    """Bind the shim. ``config``/``port`` are the injectable seams: tests pass
-    a Config with a temp VOICES_DIR and ``port=0`` for an ephemeral port."""
-    cfg = Config.from_env() if config is None else config
-    server = SpeechServer((host, cfg.port if port is None else port), SpeechHandler)
-    server.config = cfg
-    return server
-
-
-def main(argv=None):
-    parser = argparse.ArgumentParser(
-        prog="frank-tts", description="OpenAI-compatible speech endpoint for cloned voice packs")
-    parser.add_argument("--host", default="0.0.0.0",
-                        help="bind address (default: all interfaces)")
-    parser.add_argument("--port", type=int, default=None,
-                        help="bind port (default: $SHIM_PORT, 0 = ephemeral)")
-    args = parser.parse_args(argv)
-
-    server = create_server(host=args.host, port=args.port)
-    cfg = server.config
-    bound = server.server_address[1]
-    print("frank-tts: backend=%s voices_dir=%s listening on http://%s:%d"
-          % (cfg.backend, cfg.voices_dir, args.host, bound), flush=True)
-    try:
-        server.serve_forever()
-    except KeyboardInterrupt:
-        pass
-    finally:
-        server.server_close()
-    return 0
-
-
-if __name__ == "__main__":
-    sys.exit(main())
@@ -1,134 +0,0 @@
-# Quadlet unit: frank-tts-engine (rootless, boot-persistent).
-#
-# Install path on dankclaw: ~/.config/containers/systemd/frank-tts-engine.container
-# Unit name: frank-tts-engine.service (container name: frank-tts-engine)
-#
-# WHAT IT IS. The arm A TTS engine: qwentts.cpp (pinned commit
-# 5916bf7c79dbb4e66fc059cdc507a79a23a28bb4) built into an image by
-# dankclaw/voiceservice/Containerfile.engine, running its resident `tts-server`
-# with the 1.7 B base talker Q4_K_M and the Q8_0 12 Hz tokenizer loaded ONCE at
-# startup. It is deliberately NOT a fleet endpoint: the endpoint consumers
-# address is **frank-tts on 7772** ([[Frank TTS]]), which resolves voice names,
-# re-encodes to mp3 and owns the timeout budget. Runbook: [[Frank TTS Engine]].
-#
-# WHY THIS ENGINE (Frank's call, 2026-10-18, criterion "whatever is better for
-# live output"): server-side time-to-first-audio 94-112 ms (86-451 ms end to
-# end through the façade), streaming that stays ahead of real time at warm RTF
-# 0.55-0.67, and a run-to-run spread of +-0.01 — against the mainline CLI path's
-# 0.93-0.99 at +-0.25, which crosses 1.0 often enough to stall a live stream.
-# The A/B that decided it is "Decision: arm A for live output" in
-# [[Local TTS and Voice Cloning]].
-#
-# IMAGE. Built locally by deploy-remote.sh (build_compose_image), like
-# loki/alloy/frank-tts. `Pull=never` plus the `localhost/` prefix are required:
-# this host's quadlet path cannot resolve a locally built tag from local storage
-# otherwise. The build is a network operation (it clones the pinned commit) and
-# is skipped when the Containerfile hash label already matches.
-#
-# NETWORK, AND WHY IT PUBLISHES NOTHING. The container sits on exactly one
-# network: the private `frank-tts-engine.network` (10.89.5.0/24). The façade
-# joins the same network and dials `http://frank-tts-engine:8880` — the
-# container DNS name, on the network that owns it. There is no `PublishPort` of
-# any kind, not even `127.0.0.1:8880`, and the reason is the lesson that 502'd
-# hermes: netavark DNATs published ports only, so a host-loopback publish is not
-# a path to a peer container — the façade could never have used it, and a bridge
-# container cannot reach the host's own loopback bind either (see
-# [[Container Network Namespaces]]). A publish would therefore buy exactly one
-# thing: an extra reachable surface on an engine that has no authentication, on
-# a box whose CPU is also carrying the text endpoints. Diagnostics do not need
-# it either — `podman exec frank-tts-engine curl http://127.0.0.1:8880/health`
-# speaks to the server from inside its own namespace, which is where the address
-# is valid.
-#
-# MEMORY / THE RENDER NODE. Host RSS is ~1.7 GiB, not the size of the GGUFs:
-# the Vulkan backend copies both tensors into device memory at load, so they
-# stop counting in the process's RSS (measured after the 2026-10-19 cutover —
-# `podman stats` / `ps -o rss`). No memory limit is set: the box has the room
-# and a limit would convert a busy moment into an OOM kill mid-utterance.
-#
-# `AddDevice=/dev/dri` is what makes the iGPU visible; without it the Vulkan
-# ICD enumerates nothing and the server dies on device lookup after it has
-# already read both GGUFs. `GroupAdd=keep-groups` is belt-and-braces, not the
-# thing that works here: the units podman spawns carry no supplementary groups
-# at all, and this host grants render access a different way — the udev rule
-# that makes the render nodes world-writable (same rule ComfyUI rides, see
-# dankclaw/comfyui). Measured: `id` inside this container lists only the
-# container's own group, and `tts-server` still holds renderD129.
-# The node is NOT named: this host is hybrid, renderD128 is the NVIDIA card and
-# renderD129 the amdgpu iGPU, so a node number in the unit is a guess that a
-# re-probe can invalidate, while the directory hands the container every render
-# node and the driver picks by name.
-#
-# /dev/kfd stays absent on purpose: that is the ROCm compute interface, owned by
-# the engines that use HIP. A Vulkan endpoint has no use for it, and handing a
-# container a second way onto the GPU is surface for nothing.
-#
-# THE VOICE MOUNT IS READ-ONLY AND IS FOR THE WARMER, NOT THE ENGINE. A voice
-# becomes known to the engine only through POST /v1/audio/voices, and the
-# registration is MEMORY ONLY — it does not survive a restart, which is exactly
-# why the warm hook below runs on every start. That hook runs INSIDE this
-# container (see ExecStartPost), so it reads the packs at /voices; `tts-server`
-# itself never opens the directory, it receives the reference audio as base64 in
-# the registration body. The façade mounts the same host directory read-only for
-# its own name list, so one directory stays the single source of a voice name.
-#
-# REGISTRY LAW, THE REASON EXECSTARTPOST EXISTS. The voice registry is memory
-# only: a restart forgets every voice, and re-registering does not warm one —
-# only synthesizing does — so a boot without this hook leaves the façade
-# answering 400 for every name. The per-voice cold penalty is real but small, and
-# not the trial's 6.2-8.0 s: measured on the built container 2026-10-18 against an
-# empty registry, the first synthesis of a freshly registered voice cost 1,883 ms
-# against 1,226-1,283 ms for the requests after it. The warm pass runs serially
-# (two short syntheses per voice, ~5 s each) and `TimeoutStartSec=900` keeps a growing library from
-# tripping systemd's default 90 s start deadline. The leading `-` on
-# ExecStartPost is load-bearing: with `Restart=always`, a warmer that exits
-# non-zero (a pack missing its reference.wav, the engine still loading past the
-# warmer's own patience) would restart the whole container and crash-loop the
-# endpoint forever. A failed warm must be a journal line, not a restart.
-#
-# The hook is `podman exec`, not a host script, because there is no address the
-# host could dial: this container publishes nothing (above). Exec-ing into it
-# also means the warmer inherits this unit's EnvironmentFile, so every knob it
-# has lives in scripts/frank-tts-engine.env beside the engine's own, and it sees
-# the same /voices and 127.0.0.1:8880 the engine does.
-#
-# HEALTH. `GET /health` answers `{"status":"ok"}` from the HTTP thread, not the
-# synthesis queue, so a busy engine still reads healthy and a dead one does not.
-# HealthStartPeriod covers the first model load. The command is argument-only
-# and needs /bin/sh — the generator wraps HealthCmd as CMD-SHELL — so the image
-# installs curl (see the opposite case in dankclaw/loki, whose distroless base
-# has no shell).
-
-[Unit]
-Description=frank-tts-engine (arm A qwentts.cpp TTS engine, internal :8880)
-After=network-online.target
-Wants=network-online.target
-
-[Container]
-ContainerName=frank-tts-engine
-Image=localhost/franksdank/frank-tts-engine:latest
-Pull=never
-Network=frank-tts-engine.network
-# No PublishPort, at all: the only reachable address this server has is the
-# bridge address the façade resolves by name. Diagnostics use `podman exec`.
-EnvironmentFile=/home/dank/serverAppConfigs/scripts/frank-tts-engine.env
-Volume=/home/dank/models/Qwen3-TTS-GGUF:/models:ro
-Volume=/home/dank/voices:/voices:ro
-# Render access for the iGPU (see MEMORY / THE RENDER NODE above).
-AddDevice=/dev/dri
-GroupAdd=keep-groups
-HealthCmd=curl -fsS http://127.0.0.1:8880/health
-HealthInterval=30s
-HealthTimeout=10s
-HealthRetries=3
-HealthStartPeriod=120s
-
-[Service]
-Restart=always
-RestartSec=10
-# Model load (tens of seconds) plus two short syntheses per registered voice.
-TimeoutStartSec=900
-ExecStartPost=-/usr/bin/podman exec frank-tts-engine /app/warm-voices.sh
-
-[Install]
-WantedBy=default.target
@@ -1,52 +0,0 @@
-# Quadlet network: the private bridge of the voice pair (rootless).
-#
-# Install path on dankclaw: ~/.config/containers/systemd/frank-tts-engine.network
-# Unit name: frank-tts-engine-network.service (podman network name:
-# systemd-frank-tts-engine — quadlet prefixes "systemd-", so probe the runtime
-# name, never the filename; see [[Container Network Namespaces]]).
-#
-# WHY A PRIVATE BRIDGE FOR A PAUSE THAT LOOKS PUBLIC-FACING. frank-tts (:7772,
-# published, the one endpoint consumers address) and frank-tts-engine (:8880,
-# never published, the arm A qwentts.cpp server behind it) could both have
-# stayed on the fleet-wide dankclaw_default and resolved each other by name.
-# They are split off anyway, for two reasons:
-#
-# 1. THE HOST-LOOPBACK TRAP. The phase-1 shape in `scripts/frank-tts.env` still
-#    names `host.containers.internal:8880` as the upstream, on the assumption the
-#    engine runs on the host and publishes `127.0.0.1:8880`. That cannot work:
-#    netavark DNATs published ports only, a published port is not on any podman
-#    bridge, and a container cannot reach the host's own loopback bind — the same
-#    shape that 502'd hermes, and the reason the signal-relay hop needs a host LAN
-#    bind instead ([[Container Network Namespaces]], [[Homepage Status Probes]]).
-#    Putting both containers on one committed network removes the guess: the
-#    façade dials the engine's container DNS name on the network that owns it, and
-#    the engine publishes nothing at all.
-#
-# 2. THE ENGINE HAS NO AUTH AND NO QUEUE LIMIT WORTH THE NAME. `tts-server`
-#    answers anyone who can open a socket, and at `--max-batch 1` a second
-#    caller simply waits behind the first. Left on dankclaw_default every
-#    container on that net — open-webui, hermes, the *arr proxies — could drive
-#    a CPU synthesiser that occupies every other core on a box that also serves
-#    two text endpoints. On this bridge the only dialer is frank-tts: the engine
-#    publishes no port, so not even the host has an address to reach it at, and
-#    diagnostics go through `podman exec` instead.
-#
-# Kernel bridge, not slirp4netns: the 2026-08-22 open-webui incident (a dead
-# per-container userspace helper left a container that accepted TCP and reset
-# every request for two days, logging nothing). A bridge has no such helper.
-#
-# Subnet registry on dankclaw (rootless):
-#   10.88.0.0/16  default "podman" bridge
-#   10.89.0.0/24  gitea (systemd-gitea)
-#   10.89.1.0/24  compose stack (dankclaw_default)
-#   10.89.2.0/24  searxng (systemd-searxng)
-#   10.89.3.0/24  open-webui (systemd-open-webui)
-#   10.89.4.0/24  open-notebook (systemd-open-notebook)
-#   10.89.5.0/24  voice pair (systemd-frank-tts-engine)  <- this file
-# Pinned so the two container IPs stay deterministic across reboots, like the
-# other pinned nets; 10.89.5.0/24 was the next unclaimed /24 in this registry.
-
-[Network]
-Driver=bridge
-Subnet=10.89.5.0/24
-Gateway=10.89.5.1
@@ -1,84 +0,0 @@
-# Quadlet unit: frank-tts (rootless, boot-persistent).
-#
-# Install path on dankclaw: ~/.config/containers/systemd/frank-tts.container
-# Unit name: frank-tts.service (container name: frank-tts)
-#
-# WHAT IT IS. The fleet's one OpenAI-compatible speech endpoint: consumers
-# (omp, OpenWebUI, open-notebook) POST /v1/audio/speech with voice=<name> and
-# get audio in that cloned voice. A voice is a directory under
-# /home/dank/voices holding reference.wav; the shim refuses any name without
-# one (400 + the list that does exist) rather than narrating in a fallback
-# voice nobody chose. The engine behind it is a separate decision —
-# TTS_BACKEND in scripts/frank-tts.env picks between forwarding to the
-# OpenAI-shaped engine (`frank-tts-engine`, [[Frank TTS Engine]]) and running a
-# local CLI binary per request.
-#
-# IMAGE. Built locally from dankclaw/voiceservice/Containerfile by
-# deploy-remote.sh (build_compose_image), like loki/alloy: python 3.12-alpine
-# pinned by digest + ffmpeg, app.py copied in, running as uid 10000.
-# Pull=never and the localhost/ prefix are required — this host's quadlet
-# path cannot resolve a locally built tag from local storage otherwise.
-#
-# NETWORK. TWO attached networks, and the order is the one the open-notebook
-# pair uses: `dankclaw_default` first (the default route, and where consumers
-# live — they reach the published 7772, and open-notebook dials
-# `http://frank-tts:7772/v1` by container name), then the voice pair's private
-# bridge, `frank-tts-engine.network`, where the arm A engine lives. The shim's
-# upstream is therefore the engine's container DNS name
-# (`http://frank-tts-engine:8880`, scripts/frank-tts.env) and NOT
-# host.containers.internal: netavark DNATs published ports only, so a bridge
-# container cannot reach a host-loopback bind — the shape that 502'd hermes —
-# and the engine publishes nothing at all. That file MUST be installed beside
-# this unit (the engine's handler does it; the façade handler re-installs it for
-# `--only frank-tts`), because with a referenced Network= file missing the v1
-# generator refuses to produce this container unit at all.
-#
-# ORDERING. Wants+After on frank-tts-engine.service, not Requires: the façade
-# has a second backend (`TTS_BACKEND=llama-tts-cli`) that needs no upstream, and
-# an engine that will not start must not take this endpoint's unit down with it.
-# When both boot, this comes up after the engine, whose ExecStartPost warm pass
-# is part of its start — so the first request of the day meets a REGISTERED voice
-# (the registry is memory-only) rather than a 400, and skips the ~0.6 s penalty
-# the first synthesis of an unwarmed voice costs.
-#
-# VOICES. /home/dank/voices is host state, never repo state, and is mounted
-# READ-ONLY: this endpoint has no business writing voice packs. Because the
-# container runs as uid 10000, the packs must be world-readable (644 files,
-# 755 directories), which is how the pack builder writes them.
-#
-# NO GPU. The shim is a façade: audio conversion via ffmpeg on the CPU and,
-# for the CLI backend, whatever CPU threads LLAMA_TTS_THREADS allows, at nice
-# 19. No /dev/kfd, no /dev/dri, no keep-groups — unlike comfyui, whose unit
-# carries them deliberately.
-[Unit]
-Description=frank-tts (cloned-voice OpenAI-compatible TTS facade)
-After=network-online.target frank-tts-engine.service
-Wants=network-online.target
-Wants=frank-tts-engine.service
-
-[Container]
-ContainerName=frank-tts
-Image=localhost/franksdank/frank-tts:latest
-Pull=never
-Network=dankclaw_default
-Network=frank-tts-engine.network
-PublishPort=7772:7772
-EnvironmentFile=/home/dank/serverAppConfigs/scripts/frank-tts.env
-Volume=/home/dank/voices:/voices:ro
-# The probe is a probe, not a voice test: /health answers 200 with the backend
-# and the resolved voice list, so a mounted-but-empty /voices still reads
-# healthy while a dead app does not. busybox wget ships in the alpine base;
-# keep the command argument-only — the podman 5.x generator splits HealthCmd
-# on spaces (see the comfyui unit for the quoting trap).
-HealthCmd=wget --spider -q http://127.0.0.1:7772/health
-HealthInterval=30s
-HealthTimeout=10s
-HealthRetries=3
-HealthStartPeriod=30s
-
-[Service]
-Restart=always
-RestartSec=10
-
-[Install]
-WantedBy=default.target
@@ -1,124 +0,0 @@
-#!/usr/bin/env bash
-# tts-engine-start.sh — the entrypoint of the frank-tts-engine image
-# (dankclaw/voiceservice/Containerfile.engine). It exists for one reason: the
-# knobs of the arm A engine are command-line flags, while the fleet's knobs
-# live in scripts/frank-tts-engine.env and reach the container through
-# EnvironmentFile=. This is the translation layer, and nothing else — no
-# supervision, no retries, no logging of its own. PID 1 stays the server, so
-# `podman stop` and the unit's Restart=always drive it directly.
-#
-# THE FLAG THAT MUST STAY ABSENT:
-#
-#   --threads / -t   does not exist in this build. src/backend.h computes the
-#                    CPU thread count as std::thread::hardware_concurrency()/2
-#                    (the comment there: GEMM shares SIMD units across
-#                    hyperthreads), and there is no flag to override it. So a
-#                    CPU affinity pin (CPUQuota=, taskset, a cpuset) silently
-#                    changes the engine's thread count instead of just its
-#                    placement — pinning is deliberately not done anywhere in
-#                    this stack. See the env file header for the same warning
-#                    aimed at the operator.
-#
-# --device IS here, but conditionally: it reaches the binary only when
-# QWENTTS_DEVICE is non-empty, and empty is the CPU rollback — the binary then
-# takes its own default device, which is CPU because the CPU backend is always
-# built. One line in scripts/frank-tts-engine.env moves the endpoint between the
-# two shapes, with no rebuild. What must never reach argv is a device NAME the
-# loader does not enumerate: that is not refused by the argument parser, it dies
-# during device lookup with the loader's own error, so the journal reads as a
-# load that failed rather than a configuration refusal.
-#
-# Everything defaults to the shape this endpoint runs: 1.7 B talker Q4_K_M with
-# the Q8_0 tokenizer, one request at a time, on the private bridge, on the
-# gfx1151 iGPU. An empty environment still boots that shape.
-set -euo pipefail
-
-# The binary path. Overridable for one reason: the bats seam runs a stub
-# instead of a 1.2 GiB model. Operators never set it — the image ships exactly
-# one binary at /app/tts-server.
-: "${QWENTTS_BIN:=/app/tts-server}"
-
-# Paths are CONTAINER paths (the host directories they come from are named in
-# scripts/frank-tts-engine.env and mounted by the Quadlet unit).
-: "${QWENTTS_MODEL:=/models/qwen-talker-1.7b-base-Q4_K_M.gguf}"
-: "${QWENTTS_CODEC:=/models/qwen-tokenizer-12hz-Q8_0.gguf}"
-
-# Reported by GET /v1/models, and what scripts/frank-tts.env's
-# TTS_UPSTREAM_MODEL names. A model id, not a filename: the file name would
-# leak a quantization choice into every consumer's config.
-: "${QWENTTS_ALIAS:=frank-tts-engine}"
-
-# 0.0.0.0 inside the container is not "public". tts-server defaults to
-# 127.0.0.1 (its own usage prints `--host <ip> Listen address (default:
-# 127.0.0.1)`), and inside a container netns a loopback bind is reachable by
-# nothing outside that namespace — the façade dials this container by its bridge
-# address on the private voice network, which is why the address has to be
-# broad. The unit PUBLISHES NOTHING (see
-# dankclaw/voiceservice/quadlet/frank-tts-engine.container: no PublishPort at
-# all, not even 127.0.0.1:8880), so the only namespaces that can reach
-# 0.0.0.0:8880 are that bridge and this container's own loopback — and its own
-# loopback is exactly where the boot warmer and the health probe speak.
-: "${QWENTTS_HOST:=0.0.0.0}"
-: "${QWENTTS_PORT:=8880}"
-
-# The language label used when a request omits one. `auto` is the upstream
-# default and what the trial ran.
-: "${QWENTTS_LANG:=auto}"
-
-# One request at a time. The server-side queue is a serialised FIFO at this
-# setting, which is what keeps the per-voice timings predictable. A second
-# concurrent synthesis shares the device with the first — and on this box that
-# device is also carrying frank-dragoon — so it buys no throughput and it
-# doubles the wait for the caller behind it. The engine's own default is also 1;
-# written here because it is a decision, not an accident.
-: "${QWENTTS_MAX_BATCH:=1}"
-
-# The compute device, NAMED and not numbered: `Vulkan0` is how ggml enumerates
-# the gfx1151 iGPU on this host. An index would follow whatever order the
-# loader's device list returns; a name says which device this endpoint means.
-# The container sees exactly one Vulkan device because the image carries the
-# Mesa ICD and no NVIDIA one, so the 3090 is not in the enumeration at all.
-#
-# EMPTY IS THE CPU ROLLBACK: a value that is set but EMPTY passes no `--device`,
-# the binary picks its own default, and the CPU backend is always built — which
-# is why moving this endpoint between the iGPU and the CPU is one line in
-# scripts/frank-tts-engine.env and not a rebuild.
-#
-# That is also why the default below is NOT `${QWENTTS_DEVICE:=Vulkan0}`: `:=`
-# fires on a set-but-empty variable as well as an unset one, so the rollback line
-# `QWENTTS_DEVICE=` would be rewritten back to Vulkan0 and the endpoint would sit
-# on the shared GPU with no way out short of a rebuild. Only UNSET takes the
-# default.
-if [[ ! -v QWENTTS_DEVICE ]]; then
-    QWENTTS_DEVICE=Vulkan0
-fi
-
-# Fail loudly BEFORE binding, so the journal says which artifact is missing
-# instead of the unit reporting a container that starts and dies. This is the
-# self-healing path: Restart=always re-runs it every RestartSec, and the same
-# sentence repeats until the operator fetches the weights
-# (scripts/download-frank-tts-engine-qwentts.sh).
-for artifact in "$QWENTTS_MODEL" "$QWENTTS_CODEC"; do
-    if [[ ! -r "$artifact" ]]; then
-        echo "frank-tts-engine: cannot read ${artifact}; run scripts/download-frank-tts-engine-qwentts.sh on the host (the /models mount is read-only and holds the two GGUF files)" >&2
-        exit 1
-    fi
-done
-
-ARGS=(
-    --model "$QWENTTS_MODEL"
-    --codec "$QWENTTS_CODEC"
-    --alias "$QWENTTS_ALIAS"
-    --host "$QWENTTS_HOST"
-    --port "$QWENTTS_PORT"
-    --lang "$QWENTTS_LANG"
-    --max-batch "$QWENTTS_MAX_BATCH"
-)
-
-# Only when a device is named. An empty QWENTTS_DEVICE has to produce NO flag,
-# because "the rollback shape" is argv identical to the pre-iGPU endpoint.
-if [[ -n "$QWENTTS_DEVICE" ]]; then
-    ARGS+=(--device "$QWENTTS_DEVICE")
-fi
-
-exec "$QWENTTS_BIN" "${ARGS[@]}"
@@ -1,222 +0,0 @@
-#!/usr/bin/env bash
-# tts-engine-warm.sh — the boot warmer of the arm A TTS engine, shipped inside
-# the frank-tts-engine image at /app/warm-voices.sh
-# (dankclaw/voiceservice/Containerfile.engine).
-#
-# WHO RUNS IT. Two callers, same binary:
-#
-#   1. The Quadlet unit, as
-#      `ExecStartPost=-/usr/bin/podman exec frank-tts-engine /app/warm-voices.sh`
-#      — after every start of the container, including every boot.
-#      `podman exec` rather than a host script because the engine publishes no
-#      port at all, so there is no address the host could dial; and because an
-#      exec inherits the unit's EnvironmentFile, which is what puts every knob of
-#      this script in scripts/frank-tts-engine.env next to the engine's own.
-#   2. scripts/deploy-remote.sh's deploy_frank_tts_engine, once per deploy, so a
-#      voice that did not warm becomes a deploy result instead of only a journal
-#      line (the unit's `-` prefix deliberately cannot fail the unit).
-#
-# WHY IT EXISTS AT ALL (the registry law). The voice registry is MEMORY ONLY: a
-# restart forgets every voice, and re-registering a voice does not warm it — only
-# synthesizing does. A boot that registers nothing therefore leaves frank-tts
-# answering 400 for every name, so this script registers every pack it finds. The
-# per-voice cold penalty is real but small, and smaller than the trial reported:
-# measured on the deployed build 2026-10-18 against an empty registry, the FIRST
-# synthesis of a freshly registered voice cost 1,883 ms against 1,226-1,283 ms for
-# the requests after it — about 0.6 s, not the trial's 6.2-8.0 s. Each voice gets
-# two identical requests so the journal carries both numbers, `voice <name>
-# cold=…ms warm=…ms`; cold ≈ warm is the expected shape, and a large gap is what
-# would mean the snapshot path is broken.
-#
-# THE VOICES. /voices — the host directory /home/dank/voices mounted read-only by
-# the unit, the same packs frank-tts mounts for its own name list. A pack is a
-# directory that contains reference.wav; everything else in it (manifest.json, the
-# engine's reference.spk / reference.rvq extraction, a 16 kHz variant) is not read
-# here. The registration payload is `{"name": <dir>, "wav_b64": <base64 of
-# reference.wav>}` with NO ref_text, which selects the embedding-only (x-vector)
-# voice — the shipped default, warm RTF ~0.57-0.63 against ICL's 0.67, and the one
-# this fleet can produce honestly because it has no trustworthy transcript for the
-# audition crops. Adding ref_text (with the codes) would flip that voice to
-# in-context-learning mode: better fidelity on a good transcript, slower, and only
-# worth it for batch work like a podcast render. That is a per-voice opt-in this
-# script deliberately does NOT exercise, and the door stays open: a pack that
-# wants ICL needs its text in the pack plus one key in the payload below.
-#
-# A lone `spk_b64` is a 400 (src/tts-server.h:480 — it is `wav_b64` or BOTH
-# `spk_b64` and `rvq_b64`), which is why the WAV travels and the .spk stays home.
-#
-# NOTHING HERE IS A SECRET, and nothing is logged that could be: the audio bytes
-# are discarded, the synthesized text comes from the env file, and the reference
-# clip is base64'd into a request body that is written to a temporary file and
-# handed to curl as `@file` — never echoed, and never placed on a command line,
-# because of the size law explained at the mktemp below.
-set -euo pipefail
-
-# The engine's own address, from inside its namespace. Loopback is valid HERE:
-# this script runs in the container, next to the server. There is deliberately no
-# host-visible address to use instead — the unit publishes nothing, because a
-# host-loopback publish would be unreachable from the façade anyway (netavark
-# DNATs published ports only) while adding a surface on an unauthenticated
-# server.
-: "${ENGINE_URL:=http://127.0.0.1:8880}"
-
-# Container path; the host directory behind it is named in the Quadlet unit.
-: "${VOICES_DIR:=/voices}"
-
-# How long to wait for the engine to answer /health. Model load dominates: the
-# two GGUFs (1.2 GiB + 291 MB) are read and the context built before the HTTP
-# thread binds, and a boot races every other unit on this box.
-: "${HEALTH_WAIT:=300}"
-
-# One short sentence, long enough to exercise the codec snapshot and short enough
-# that warming a library of voices is seconds, not minutes. Never varied per run:
-# the cold/warm numbers in the journal should mean the same thing every boot. The
-# synthesis asks for `wav`, the buffered path — the streamed path (`pcm`) is
-# different audio at the same seed, and `wav` is also what frank-tts asks for.
-: "${WARM_TEXT:=Warm up, one two three.}"
-
-# Per-request ceiling. A cold voice is 6-8 s by measurement; anything approaching
-# this is a stalled engine and should be reported rather than waited on forever —
-# systemd counts the whole warm pass against the unit's start deadline.
-: "${REQUEST_TIMEOUT:=120}"
-
-# The warm text goes into a JSON string literal by interpolation. Two steps make
-# whatever the two readers of scripts/frank-tts-engine.env deliver arrive as the
-# sentence itself:
-#
-# 1. Unwrap ONE shell-style quoting layer. bash (`set -a; source`) strips the
-#    quotes around a value with spaces; podman's `--env-file` does NOT — measured
-#    2026-10-18 against the image's pinned base, `WARM_TEXT="…"` reaches the
-#    container WITH its quotes. Unwrapped only when the whole value is exactly one
-#    quoted string (starts AND ends with `"` and holds no quote inside), so a
-#    sentence that legitimately contains quotes is never mangled.
-# 2. Escape the two characters that would corrupt or end a JSON string: a
-#    backslash, and a double quote.
-#
-# Without step 1 the body is `{"input":""Warm up…"",…}`, the engine's JSON parser
-# rejects it, every voice stays cold, and the boot still reads healthy. Voice NAMES
-# need neither step: the character class checked below admits no quote, no
-# backslash and no control byte.
-warm_text="$WARM_TEXT"
-if (( ${#warm_text} >= 2 )) && [[ "$warm_text" == \"*\" ]]; then
-    inner="${warm_text:1:${#warm_text}-2}"
-    [[ "$inner" != *'"'* ]] && warm_text="$inner"
-fi
-warm_json="${warm_text//\\/\\\\}"
-warm_json="${warm_json//\"/\\\"}"
-
-# The registration body is a FILE, never an argument. The kernel caps a single
-# command-line string at 32 pages — MAX_ARG_STRLEN = 131,072 bytes — and a shipped
-# reference clip base64's to about 1.27 MB, so passing the body as one argument
-# fails with E2BIG before curl is even executed. The first real boot of the unit
-# (2026-10-18, the three shipped packs) proved it end to end: `/usr/bin/curl:
-# Argument list too long` for every voice, reported as the misleading
-# "registration failed (HTTP no response)". `mktemp` lands in the container's
-# /tmp as the unprivileged `tts` user; one file for the whole pass, overwritten
-# per voice, removed on any exit — it is container scratch, not host state.
-body_file="$(mktemp -t frank-tts-warm.XXXXXX)"
-trap 'rm -f "$body_file"' EXIT
-
-fail=0
-log() { printf 'frank-tts-engine-warm: %s\n' "$*"; }
-
-# Wait for /health. It answers from the HTTP thread, so a 200 means the server is
-# listening; it says nothing about voices, which is what the rest of this script
-# is for.
-ready=0
-for ((i = 1; i <= HEALTH_WAIT; i++)); do
-    if curl -fsS --max-time 5 "${ENGINE_URL}/health" >/dev/null 2>&1; then
-        ready=1
-        break
-    fi
-    sleep 1
-done
-if (( ready == 0 )); then
-    log "engine did not answer ${ENGINE_URL}/health within ${HEALTH_WAIT}s; nothing warmed"
-    exit 1
-fi
-
-if [[ ! -d "$VOICES_DIR" ]]; then
-    log "no voice directory at ${VOICES_DIR}; nothing to warm (frank-tts will answer 400 with an empty voice list until packs are built)"
-    exit 0
-fi
-
-warmed=0
-for dir in "$VOICES_DIR"/*/; do
-    [[ -d "$dir" ]] || continue
-    name="${dir%/}"
-    name="${name##*/}"
-    if [[ ! -f "${dir}reference.wav" ]]; then
-        continue          # not a pack; frank-tts refuses this name too
-    fi
-    # The engine lowercases every voice name (tts_voice_name) while frank-tts
-    # matches the directory name exactly, so a mixed-case pack would work by
-    # accident on one side and not the other. Refuse it loudly instead. Pack
-    # names are also JSON string content here, so the character class is what
-    # keeps a directory name from writing structure into the request body.
-    if [[ ! "$name" =~ ^[a-z0-9][a-z0-9._-]*$ ]]; then
-        log "voice '${name}': name must be lowercase [a-z0-9._-] (the engine lowercases names; the shim does not) — skipped"
-        fail=1
-        continue
-    fi
-
-    # The audio enters through a redirection rather than a command substitution:
-    # the file is several hundred KB, and both a shell variable and an argument of
-    # that size are what this script used to die on.
-    {
-        printf '{"name":"%s","wav_b64":"' "$name"
-        base64 -w0 "${dir}reference.wav"
-        printf '"}'
-    } > "$body_file"
-    if ! code="$(curl -sS --max-time "$REQUEST_TIMEOUT" -o /dev/null -w '%{http_code}' \
-            -H 'Content-Type: application/json' --data-binary "@${body_file}" \
-            "${ENGINE_URL}/v1/audio/voices")" || [[ "$code" != "200" ]]; then
-        log "voice '${name}': registration failed (HTTP ${code:-no response}) — the shim will answer 400 for this name"
-        fail=1
-        continue
-    fi
-
-    # The cold pass. Its output is discarded; the point is the codec snapshot it
-    # leaves inside the engine.
-    start=$(date +%s%3N)
-    if ! curl -fsS --max-time "$REQUEST_TIMEOUT" -o /dev/null \
-            -H 'Content-Type: application/json' \
-            --data-binary "{\"input\":\"${warm_json}\",\"voice\":\"${name}\",\"response_format\":\"wav\"}" \
-            "${ENGINE_URL}/v1/audio/speech"; then
-        log "voice '${name}': registered, but the warm synthesis failed — first live request stays cold"
-        fail=1
-        continue
-    fi
-    cold=$(( $(date +%s%3N) - start ))
-
-    # The warm pass, same request. This is the number the fleet actually
-    # experiences, and it is what "the second request answers under a second" is
-    # measured against.
-    start=$(date +%s%3N)
-    if ! curl -fsS --max-time "$REQUEST_TIMEOUT" -o /dev/null \
-            -H 'Content-Type: application/json' \
-            --data-binary "{\"input\":\"${warm_json}\",\"voice\":\"${name}\",\"response_format\":\"wav\"}" \
-            "${ENGINE_URL}/v1/audio/speech"; then
-        log "voice '${name}': warm synthesis failed after a cold one that worked"
-        fail=1
-        continue
-    fi
-    warm=$(( $(date +%s%3N) - start ))
-
-    log "voice ${name} cold=${cold}ms warm=${warm}ms"
-    warmed=$(( warmed + 1 ))
-done
-
-# Failure first, emptiness second, and not the other way round: a pass where
-# every voice failed also has `warmed == 0`, and reading that as "nothing to do,
-# success" is how an endpoint that answers 400 for every voice name gets deployed
-# on a Friday.
-if (( fail )); then
-    log "${warmed} voice(s) warm, at least one failed (see the lines above)"
-    exit 1
-fi
-if (( warmed == 0 )); then
-    log "no voice with a reference.wav under ${VOICES_DIR}; engine is up but speaks nothing"
-    exit 0
-fi
-log "${warmed} voice(s) registered and warm"
@@ -0,0 +1,156 @@
+#!/usr/bin/env python3
+"""HTTP front door for the frank-medium single-GPU replicas.
+
+Why two replicas instead of TP=2: one vLLM process spanning both R9700s needs
+RCCL collectives, and RCCL's kernels require PCIe atomics on every
+participating GPU. The second card sits behind chipset bridges advertising
+`AtomicOpsCap: Routing-`, which no software can enable (see the runbook), so
+no collective can launch on it. Two single-GPU replicas use both cards; this
+service keeps the fleet contract of one endpoint on :7770, streaming included.
+
+Routing decisions live in routing.py — spread single-turn requests, pin
+continuations to the replica whose prefix cache holds their context. This file
+only moves bytes: parse enough of the request to route it, forward it verbatim,
+stream the answer back, and report replica health on /health.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import os
+from contextlib import suppress
+from aiohttp import ClientSession, ClientTimeout, web
+
+import routing
+
+PORT = int(os.environ.get("PORT", "7770"))
+HEALTH_TTL = float(os.environ.get("HEALTH_TTL", "10"))
+# Hop-by-hop headers must not be forwarded (RFC 9110 §7.6.1) — forwarding
+# transfer-encoding or content-length of a rewritten body breaks clients.
+HOP_BY_HOP = {
+    "host", "content-length", "transfer-encoding", "connection",
+    "keep-alive", "proxy-authenticate", "proxy-authorization", "te",
+    "trailers", "upgrade",
+}
+
+UPSTREAMS = routing.UPSTREAMS
+# Replica health, refreshed by the watcher. Module state rather than aiohttp app
+# state: the application is frozen once the site is up.
+_UP: list[bool] = [True] * len(UPSTREAMS)
+
+
+async def health(session: ClientSession) -> web.Response:
+    results = await asyncio.gather(
+        *(session.get(u + "/health", timeout=ClientTimeout(total=5)) for u in UPSTREAMS),
+        return_exceptions=True,
+    )
+    lines, up = [], 0
+    for url, res in zip(UPSTREAMS, results):
+        if isinstance(res, Exception) or res.status != 200:
+            lines.append(f"{url} DOWN")
+        else:
+            up += 1
+            lines.append(f"{url} UP")
+    return web.Response(status=200 if up else 503, text="\n".join(lines) + "\n")
+
+
+async def watch_health() -> None:
+    """Refresh the replica up/down table so routing never targets a dead one.
+
+    One HTTP request per replica per interval, in exchange for never handing a
+    client a 502 after a restart.
+    """
+    global _UP
+    session = ClientSession()
+    try:
+        while True:
+            res = await asyncio.gather(
+                *(session.get(u + "/health", timeout=ClientTimeout(total=5)) for u in UPSTREAMS),
+                return_exceptions=True,
+            )
+            _UP = [not isinstance(r, Exception) and r.status == 200 for r in res]
+            await asyncio.sleep(HEALTH_TTL)
+    finally:
+        await session.close()
+
+
+async def proxy(request: web.Request) -> web.StreamResponse:
+    session: ClientSession = request.app["session"]
+    if request.path == "/health":
+        return await health(session)
+
+    body = await request.read()
+    headers = {k: v for k, v in request.headers.items() if k.lower() not in HOP_BY_HOP}
+
+    # Fail over before anything is written to the client. A replica that
+    # restarts is invisible to the health watcher for up to HEALTH_TTL, and a
+    # deploy restarts both replicas — without this retry a 10 s window of
+    # requests would 502 on a service that has a live card the whole time.
+    # vLLM's OpenAI endpoints are stateless per request, so re-sending a
+    # request that never reached an engine is safe; a request that failed
+    # mid-stream is not retried (that path is below).
+    tried: set[int] = set()
+    upstream = None
+    last_error = "no healthy frank-medium replica"
+    while len(tried) < len(UPSTREAMS):
+        idx = routing.route(body, request.remote, _UP)
+        if idx < 0 or idx in tried:
+            break
+        tried.add(idx)
+        target = UPSTREAMS[idx] + request.path_qs
+        try:
+            upstream = await session.request(
+                request.method, target, headers=headers, data=body or None,
+                timeout=ClientTimeout(total=None, sock_connect=10, sock_read=3600),
+            )
+            break
+        except Exception as exc:
+            # Demote at once: the watcher will confirm, but every request until
+            # then should skip this replica rather than pay the same timeout.
+            # Logged because a silent failover hides a card that is down.
+            _UP[idx] = False
+            last_error = f"{target} unreachable: {exc}"
+            print(f"proxy: failover — {target} unreachable ({exc}); retrying "
+                  f"another replica", flush=True)
+    if upstream is None:
+        # 503 when no replica was even reachable through the routing table,
+        # 502 when replicas were tried and all of them failed.
+        status = 502 if tried else 503
+        return web.Response(status=status, text=f"proxy: {last_error}\n")
+
+    out = web.StreamResponse(status=upstream.status, reason=upstream.reason)
+    for k, v in upstream.headers.items():
+        if k.lower() not in HOP_BY_HOP:
+            out.headers[k] = v
+    await out.prepare(request)
+    try:
+        async for chunk in upstream.content.iter_any():
+            await out.write(chunk)
+    except Exception:
+        # The client went away mid-stream (an aborted generation). Nothing to
+        # report: the engine sees its own cancellation.
+        pass
+    finally:
+        upstream.release()
+    with suppress(Exception):
+        await out.write_eof()
+    return out
+
+
+async def main() -> None:
+    app = web.Application(client_max_size=1024 ** 3)
+    app["session"] = ClientSession()
+    app.router.add_route("*", "/{tail:.*}", proxy)
+    runner = web.AppRunner(app)
+    await runner.setup()
+    await web.TCPSite(runner, "0.0.0.0", PORT).start()
+    print(f"proxy: :{PORT} -> {', '.join(UPSTREAMS)} routing={routing.ROUTING} "
+          f"key={routing.AFFINITY_BYTES}B known_max={routing.KNOWN_MAX}", flush=True)
+    watcher = asyncio.create_task(watch_health())
+    with suppress(asyncio.CancelledError):
+        await asyncio.gather(watcher)
+
+
+if __name__ == "__main__":
+    with suppress(KeyboardInterrupt):
+        asyncio.run(main())
@@ -0,0 +1,98 @@
+# Quadlet unit: frank-medium replica A (rootless, boot-persistent).
+#
+# Install path on dankserver: ~/.config/containers/systemd/frank-medium-a.container
+# Unit name: frank-medium-a.service (container name: frank-medium-a)
+#
+# WHY TWO REPLICAS INSTEAD OF TP=2. The intended shape for frank-medium is one
+# tensor-parallel process across both R9700s (see frank-medium.container, kept
+# as the target). That shape cannot run on this host: RCCL's collective kernels
+# carry a hidden hostcall-buffer parameter, and the ROCr runtime refuses to
+# dispatch them on a device without PCIe atomics ("Pcie atomics not enabled,
+# hostcall not supported" -> AQL dispatch failed -> hipErrorIllegalState). The
+# second R9700 sits behind the chipset bridges 05:00.0/06:09.0, which advertise
+# `AtomicOpsCap: Routing-` — a read-only capability, so no software can enable
+# atomics on that path; the amdgpu driver's pci_enable_atomic_ops_to_root()
+# therefore fails for that GPU and KFD marks the CPU<->GPU2 link NO_ATOMICS.
+# The first R9700 sits on CPU root port 00:01.1 (Routing+) and does have
+# atomics. Result: every collective that touches GPU 2 dies at launch.
+# Two independent single-GPU replicas use both cards instead; the proxy unit
+# keeps the single :7770 endpoint. Full evidence: the frank-medium runbook.
+#
+# SHAPE. vLLM binds 0.0.0.0:7771 on the host network directly; no PublishPort.
+# MAXLEN is 114688, not the TP=2 deployment's 262144: one card carries the
+# whole model (~20 GiB of weights + drafter), leaving ~5.2 GiB of KV, and
+# 131072 does not fit (measured: needs 5.58 GiB). KV_MEM=0 lets vLLM profile
+# this shape instead of using the 2x7551-32624 pin, which is only valid at
+# TP=2/262144.
+#
+# DEVICES. /dev/kfd + /dev/dri; HIP_VISIBLE_DEVICES=0 (the first R9700). The
+# 2 GiB Raphael iGPU enumerates as index 2 and is excluded by the pin.
+# Only HIP_VISIBLE_DEVICES is set, never ROCR_VISIBLE_DEVICES: ROCR filters
+# first and HIP indexes into the filtered list, so a single-valued pair
+# (ROCR=1,HIP=1) leaves an empty device set.
+#
+# DO NOT name an environment variable TEMP, HOST or GPU_MAX_HEAP_SIZE: the
+# HIP runtime reads bare TEMP for its JIT cache path (2026-08-13 incident).
+[Unit]
+Description=frank-medium replica A — vllm-radiance MXFP4 Qwen3.8-27B on R9700 #1
+After=network-online.target
+
+[Container]
+ContainerName=frank-medium-a
+Image=localhost/radiance-vllm:0.9.3
+AddDevice=/dev/kfd
+AddDevice=/dev/dri
+GroupAdd=keep-groups
+# The upstream launcher's exact podman posture (rootless).
+PodmanArgs=--privileged --ipc=host --security-opt seccomp=unconfined --cap-add SYS_PTRACE --network=host
+Environment=HIP_VISIBLE_DEVICES=0
+Environment=SNAP=/models/Qwen3.8-27B-Uncensored-MXFP4-awq
+Environment=DRAFTER=/models/Qwen3.8-27B-DFlash2-FP8
+Environment=CHAT_TEMPLATE=/patches/qwen-fixed-v22.3.jinja
+Environment=SERVED_NAMES=frank-medium
+Environment=PORT=7771
+Environment=TP=1
+Environment=MAXSEQS=4
+Environment=CHUNK=8192
+Environment=MAXLEN=114688
+Environment=GPU_UTIL=0.98
+Environment=KV_MEM=0
+Environment=SPEC_METHOD=dflash
+Environment=SPEC=7
+# Checkpoints (rw: nothing downloads at runtime, but vLLM mmaps the shards).
+Volume=/home/dank/models:/models
+# Tokenizer files ride the HF cache; HF_HUB_OFFLINE=1 is set in the entrypoint.
+Volume=/home/dank/.cache/huggingface:/root/.cache/huggingface
+# Compile cache. Shared with the other replica and with the TP=2 target: the
+# entries are content-addressed, so shapes simply add subdirectories.
+Volume=/home/dank/.radiance-cache-w4a8-093-gdnm-nqft-fp8s-gnq:/cache
+# The pinned radiance clone: patches applied at every container start.
+Volume=/home/dank/GitPackages/radiance-vllm-mxfp4:/patches:z
+# The patched libr4d (pin b9e42ab + repo extras, key b9e42ab-rx6). The image's
+# shipped r4d.so predates the GDN overflow fix and NaNs this model.
+Volume=/home/dank/.cache/radiance-libr4d/b9e42ab-rx6:/r4d:z
+# Entrypoint + healthcheck from this checkout, mounted read-only.
+Volume=/home/dank/serverAppConfigs/dankserver/frank-medium/radiance-entrypoint.sh:/usr/local/bin/frank-medium-entrypoint.sh:ro
+Volume=/home/dank/serverAppConfigs/dankserver/frank-medium/frank-medium-healthcheck.py:/usr/local/bin/frank-medium-healthcheck.py:ro
+# The entrypoint must be the container command; the image's own entrypoint
+# (which the patched script execs at the end) must not run unpatched.
+PodmanArgs=--entrypoint /usr/local/bin/frank-medium-entrypoint.sh
+# 200 only once the engine serves; during first-start compilation the
+# container is legitimately unhealthy. HealthCmd is split on spaces into an
+# argv array by the podman 5.x v1 generator, hence the file-based probe.
+HealthCmd=python3 /usr/local/bin/frank-medium-healthcheck.py
+HealthInterval=60s
+HealthTimeout=30s
+HealthRetries=3
+HealthStartPeriod=1800s
+Pull=never
+
+[Service]
+# First start compiles Triton/inductor kernels (several minutes) before the
+# engine loads ~20 GiB of weights.
+TimeoutStartSec=1800
+Restart=on-failure
+RestartSec=10
+
+[Install]
+WantedBy=default.target
@@ -0,0 +1,98 @@
+# Quadlet unit: frank-medium replica B (rootless, boot-persistent).
+#
+# Install path on dankserver: ~/.config/containers/systemd/frank-medium-b.container
+# Unit name: frank-medium-b.service (container name: frank-medium-a)
+#
+# WHY TWO REPLICAS INSTEAD OF TP=2. The intended shape for frank-medium is one
+# tensor-parallel process across both R9700s (see frank-medium.container, kept
+# as the target). That shape cannot run on this host: RCCL's collective kernels
+# carry a hidden hostcall-buffer parameter, and the ROCr runtime refuses to
+# dispatch them on a device without PCIe atomics ("Pcie atomics not enabled,
+# hostcall not supported" -> AQL dispatch failed -> hipErrorIllegalState). The
+# second R9700 sits behind the chipset bridges 05:00.0/06:09.0, which advertise
+# `AtomicOpsCap: Routing-` — a read-only capability, so no software can enable
+# atomics on that path; the amdgpu driver's pci_enable_atomic_ops_to_root()
+# therefore fails for that GPU and KFD marks the CPU<->GPU2 link NO_ATOMICS.
+# The first R9700 sits on CPU root port 00:01.1 (Routing+) and does have
+# atomics. Result: every collective that touches GPU 2 dies at launch.
+# Two independent single-GPU replicas use both cards instead; the proxy unit
+# keeps the single :7770 endpoint. Full evidence: the frank-medium runbook.
+#
+# SHAPE. vLLM binds 0.0.0.0:7772 on the host network directly; no PublishPort.
+# MAXLEN is 114688, not the TP=2 deployment's 262144: one card carries the
+# whole model (~20 GiB of weights + drafter), leaving ~5.2 GiB of KV, and
+# 131072 does not fit (measured: needs 5.58 GiB). KV_MEM=0 lets vLLM profile
+# this shape instead of using the 2x7551-32624 pin, which is only valid at
+# TP=2/262144.
+#
+# DEVICES. /dev/kfd + /dev/dri; HIP_VISIBLE_DEVICES=1 (the second R9700). The
+# 2 GiB Raphael iGPU enumerates as index 2 and is excluded by the pin.
+# Only HIP_VISIBLE_DEVICES is set, never ROCR_VISIBLE_DEVICES: ROCR filters
+# first and HIP indexes into the filtered list, so a single-valued pair
+# (ROCR=1,HIP=1) leaves an empty device set.
+#
+# DO NOT name an environment variable TEMP, HOST or GPU_MAX_HEAP_SIZE: the
+# HIP runtime reads bare TEMP for its JIT cache path (2026-08-13 incident).
+[Unit]
+Description=frank-medium replica B — vllm-radiance MXFP4 Qwen3.8-27B on R9700 #2
+After=network-online.target
+
+[Container]
+ContainerName=frank-medium-b
+Image=localhost/radiance-vllm:0.9.3
+AddDevice=/dev/kfd
+AddDevice=/dev/dri
+GroupAdd=keep-groups
+# The upstream launcher's exact podman posture (rootless).
+PodmanArgs=--privileged --ipc=host --security-opt seccomp=unconfined --cap-add SYS_PTRACE --network=host
+Environment=HIP_VISIBLE_DEVICES=1
+Environment=SNAP=/models/Qwen3.8-27B-Uncensored-MXFP4-awq
+Environment=DRAFTER=/models/Qwen3.8-27B-DFlash2-FP8
+Environment=CHAT_TEMPLATE=/patches/qwen-fixed-v22.3.jinja
+Environment=SERVED_NAMES=frank-medium
+Environment=PORT=7772
+Environment=TP=1
+Environment=MAXSEQS=4
+Environment=CHUNK=8192
+Environment=MAXLEN=114688
+Environment=GPU_UTIL=0.98
+Environment=KV_MEM=0
+Environment=SPEC_METHOD=dflash
+Environment=SPEC=7
+# Checkpoints (rw: nothing downloads at runtime, but vLLM mmaps the shards).
+Volume=/home/dank/models:/models
+# Tokenizer files ride the HF cache; HF_HUB_OFFLINE=1 is set in the entrypoint.
+Volume=/home/dank/.cache/huggingface:/root/.cache/huggingface
+# Compile cache. Shared with the other replica and with the TP=2 target: the
+# entries are content-addressed, so shapes simply add subdirectories.
+Volume=/home/dank/.radiance-cache-w4a8-093-gdnm-nqft-fp8s-gnq:/cache
+# The pinned radiance clone: patches applied at every container start.
+Volume=/home/dank/GitPackages/radiance-vllm-mxfp4:/patches:z
+# The patched libr4d (pin b9e42ab + repo extras, key b9e42ab-rx6). The image's
+# shipped r4d.so predates the GDN overflow fix and NaNs this model.
+Volume=/home/dank/.cache/radiance-libr4d/b9e42ab-rx6:/r4d:z
+# Entrypoint + healthcheck from this checkout, mounted read-only.
+Volume=/home/dank/serverAppConfigs/dankserver/frank-medium/radiance-entrypoint.sh:/usr/local/bin/frank-medium-entrypoint.sh:ro
+Volume=/home/dank/serverAppConfigs/dankserver/frank-medium/frank-medium-healthcheck.py:/usr/local/bin/frank-medium-healthcheck.py:ro
+# The entrypoint must be the container command; the image's own entrypoint
+# (which the patched script execs at the end) must not run unpatched.
+PodmanArgs=--entrypoint /usr/local/bin/frank-medium-entrypoint.sh
+# 200 only once the engine serves; during first-start compilation the
+# container is legitimately unhealthy. HealthCmd is split on spaces into an
+# argv array by the podman 5.x v1 generator, hence the file-based probe.
+HealthCmd=python3 /usr/local/bin/frank-medium-healthcheck.py
+HealthInterval=60s
+HealthTimeout=30s
+HealthRetries=3
+HealthStartPeriod=1800s
+Pull=never
+
+[Service]
+# First start compiles Triton/inductor kernels (several minutes) before the
+# engine loads ~20 GiB of weights.
+TimeoutStartSec=1800
+Restart=on-failure
+RestartSec=10
+
+[Install]
+WantedBy=default.target
@@ -0,0 +1,61 @@
+# Quadlet unit: frank-medium proxy (rootless, boot-persistent).
+#
+# Install path on dankserver: ~/.config/containers/systemd/frank-medium-proxy.container
+# Unit name: frank-medium-proxy.service (container name: frank-medium-proxy)
+#
+# WHAT IT IS. A prefix-affinity reverse proxy in front of the two single-GPU
+# frank-medium replicas (:7771 on R9700 #1, :7772 on R9700 #2). It preserves
+# the fleet contract: one endpoint on host port 7770 answering as the served
+# model name frank-medium, OpenAI /v1, streaming included.
+#
+# WHY IT EXISTS. TP=2 — one vLLM process spanning both cards — is impossible
+# on this host (see frank-medium-a.container and the runbook: RCCL's kernels
+# need PCIe atomics, and the second card's chipset path advertises
+# AtomicOpsCap: Routing-). Two replicas are how both cards get used.
+#
+# ROUTING. Requests are keyed on a hash of their leading 4 KiB of prompt text,
+# so a conversation family — an agent loop re-sending the same system prompt,
+# tool list and file context every turn — keeps landing on the same replica,
+# where --enable-prefix-caching has that prefix warm. Plain round-robin threw
+# half of those cache hits away. A replica that is down is skipped, and
+# ROUTING=roundrobin restores the old behaviour for A/B measurement. /health
+# answers 200 while at least one replica is up and names the ones that are
+# down. The proxy runs on host networking so it can bind 7770 and reach the
+# replicas on loopback.
+#
+# The script is a few dozen lines of aiohttp, mounted read-only out of
+# /home/dank/serverAppConfigs (deploy.sh pulls that checkout before dispatch).
+# It reuses the same image as the replicas: aiohttp is already there and no
+# extra image has to be pulled on this host.
+[Unit]
+Description=frank-medium proxy — prefix-affinity in front of the two single-GPU replicas
+After=network-online.target
+Wants=frank-medium-a.service frank-medium-b.service
+After=frank-medium-a.service frank-medium-b.service
+
+[Container]
+ContainerName=frank-medium-proxy
+Image=localhost/radiance-vllm:0.9.3
+PodmanArgs=--network=host
+Environment=PORT=7770
+Environment=UPSTREAMS=http://127.0.0.1:7771,http://127.0.0.1:7772
+# The proxy is two files: routing.py holds the replica-selection policy, the
+# server holds the HTTP plumbing. Both are mounted because the server imports
+# the policy by name from its own directory (/usr/local/bin is sys.path[0]).
+Volume=/home/dank/serverAppConfigs/dankserver/frank-medium/proxy/routing.py:/usr/local/bin/routing.py:ro
+Volume=/home/dank/serverAppConfigs/dankserver/frank-medium/proxy/affinity-proxy.py:/usr/local/bin/affinity-proxy.py:ro
+Volume=/home/dank/serverAppConfigs/dankserver/frank-medium/frank-medium-healthcheck.py:/usr/local/bin/frank-medium-healthcheck.py:ro
+PodmanArgs=--entrypoint /usr/local/bin/affinity-proxy.py
+HealthCmd=python3 /usr/local/bin/frank-medium-healthcheck.py
+HealthInterval=60s
+HealthTimeout=30s
+HealthRetries=3
+HealthStartPeriod=60s
+Pull=never
+
+[Service]
+Restart=on-failure
+RestartSec=10
+
+[Install]
+WantedBy=default.target
@@ -0,0 +1,129 @@
+#!/usr/bin/env python3
+"""Routing policy for the frank-medium proxy: which replica serves a request.
+
+Kept apart from the HTTP plumbing in affinity-proxy.py so the decisions can be
+tested without aiohttp, a network or a GPU — the first version of these rules
+shipped with a bug that only appeared on a continuation request, and a test
+that loads this module catches that class of mistake directly.
+
+Two cards, two behaviours, and the workload decides which is right:
+
+* A **single-turn** request is one member of a batch (a benchmark run, a burst
+  of parallel subagents). Spreading those over both replicas lets both cards
+  prefill at once; measured on this host, a 16-request burst of 15k-token
+  prompts finishes in 18.6 s spread versus 30.1 s pinned.
+* A **continuation** re-sends a prefix the replica has already prefilled. Pinned
+  to that replica the turn costs a cache lookup; spread, half the turns pay a
+  cold prefill again (measured: 1096 ms versus 2676 ms mean warm-turn TTFT on
+  90k-token contexts).
+
+ROUTING picks the strategy; the default `adaptive` applies both rules above.
+`affinity` and `roundrobin` exist so the choice can be re-measured, and are not
+production settings.
+"""
+
+from __future__ import annotations
+
+import collections
+import hashlib
+import itertools
+import json
+import os
+
+UPSTREAMS = [u.strip() for u in os.environ.get(
+    "UPSTREAMS", "http://127.0.0.1:7771,http://127.0.0.1:7772").split(",") if u.strip()]
+ROUTING = os.environ.get("ROUTING", "adaptive").strip().lower()
+# Leading bytes of prompt text that identify a conversation's shared prefix.
+# 4 KiB covers a system prompt plus the start of the first user turn, which is
+# where agent loops keep the text they resend every turn.
+AFFINITY_BYTES = int(os.environ.get("AFFINITY_BYTES", "4096"))
+# Prefixes remembered for pinning. Bounded so a long-lived proxy cannot grow
+# without limit; 4096 conversations is far beyond what this fleet runs at once.
+KNOWN_MAX = int(os.environ.get("KNOWN_MAX", "4096"))
+
+_counter = itertools.count()
+# digest -> replica index, insertion-ordered for O(1) LRU eviction. A prefix is
+# remembered as soon as any request carries it; only a continuation is pinned
+# there, because only a continuation can profit from the cache entry.
+_known: "collections.OrderedDict[bytes, int]" = collections.OrderedDict()
+
+
+def affinity_key(body: bytes, remote: str | None) -> tuple[bytes, bool]:
+    """Key a request by its shared prefix, and report whether it continues one.
+
+    Chat and completions bodies both carry their prompt text in JSON, so the
+    leading bytes of that text are the prefix; a body we cannot parse (or an
+    endpoint with no prompt, e.g. /v1/models) falls back to the client address,
+    which still beats nothing for a single-client loop.
+    """
+    fallback = (remote or "anonymous").encode()
+    try:
+        payload = json.loads(body)
+    except Exception:
+        return fallback, False
+
+    text = ""
+    continuation = False
+    if isinstance(payload, dict):
+        if isinstance(payload.get("prompt"), str):
+            text = payload["prompt"]
+        elif isinstance(payload.get("messages"), list):
+            continuation = len(payload["messages"]) > 1
+            parts = []
+            for msg in payload["messages"]:
+                if isinstance(msg, dict):
+                    content = msg.get("content")
+                    if isinstance(content, str):
+                        parts.append(content)
+                    elif isinstance(content, list):  # multimodal content blocks
+                        parts.extend(
+                            blk.get("text", "") for blk in content
+                            if isinstance(blk, dict) and isinstance(blk.get("text"), str)
+                        )
+            text = "\n".join(parts)
+    if not text:
+        return fallback, continuation
+    return text[:AFFINITY_BYTES].encode("utf-8", "replace"), continuation
+
+
+def pick(key: bytes, continuation: bool, up: list[bool]) -> int:
+    """Replica for a request, or -1 when none is available.
+
+    Down replicas are never selected — a rollout must not send traffic to an
+    engine that is still compiling its graphs — and a pin whose replica is down
+    is re-established on a live one rather than kept pointing at a corpse.
+    """
+    digest = hashlib.sha1(key).digest()
+
+    if ROUTING == "roundrobin":
+        start = next(_counter) % len(UPSTREAMS)
+    elif ROUTING == "adaptive":
+        entry = _known.get(digest)
+        if continuation and entry is not None and up[entry]:
+            _known.move_to_end(digest)
+            return entry
+        # Anything that is not a continuation spreads by counter — including a
+        # batch of requests that happen to share one prefix. Hashing the prefix
+        # would send every member of such a batch to whichever card the hash
+        # names (measured: 29.8 s for a 16-request burst, against 18.6 s when
+        # spread). The counter is only advanced by these requests, so batch
+        # members alternate across the cards.
+        start = next(_counter) % len(UPSTREAMS)
+    else:  # affinity
+        start = int.from_bytes(digest[:8], "big") % len(UPSTREAMS)
+
+    for offset in range(len(UPSTREAMS)):
+        idx = (start + offset) % len(UPSTREAMS)
+        if up[idx]:
+            if ROUTING == "adaptive":
+                _known[digest] = idx
+                _known.move_to_end(digest)
+                if len(_known) > KNOWN_MAX:
+                    _known.popitem(last=False)
+            return idx
+    return -1
+
+
+def route(body: bytes, remote: str | None, up: list[bool]) -> int:
+    """Convenience wrapper: decode the request body and choose a replica."""
+    return pick(*affinity_key(body, remote), up)
@@ -43,7 +43,7 @@ After=network.target

 [Container]
 ContainerName=calibre-server
-Image=lscr.io/linuxserver/calibre:9.17.0@sha256:b198fc03a6abc6be7e251a9958e50e74634b00e447c7b25e92c191c8077ba378
+Image=lscr.io/linuxserver/calibre:9.15.0
 Network=pasta
 EnvironmentFile=/home/dank/Containers/calibre-server/env
 Environment=PUID=1000
@@ -16,7 +16,7 @@ After=network.target

 [Container]
 ContainerName=calibre-web
-Image=lscr.io/linuxserver/calibre-web:0.6.27-ls407@sha256:470b4ed9a7d6ea1475fef17dbcf957cb60880910296d226bd0dca652c05f68ca
+Image=lscr.io/linuxserver/calibre-web:0.6.27-ls402
 Network=pasta
 Environment=DOCKER_MODS=linuxserver/mods:universal-calibre
 Environment=LANG=en_US.UTF-8
@@ -94,12 +94,34 @@ services:
      timeout: 10s
      retries: 3

-  # Jellyfin left this file when it became a Quadlet: the container is owned by
-  # dankserver/jellyfin/quadlet/jellyfin.container, its data roots are under
-  # /home/dank/Containers/jellyfin, and its library arrives through the SSD
-  # overlay at /var/lib/media-meta/view (scripts/media-view.sh). The block that
-  # was here named /opt/jellyfin/* and a read-only array bind — a second owner for
-  # the container name `jellyfin`, wrong on every path. One owner per container.
+  # ═══════════════════════════════════════════════
+  # Jellyfin — media server (movies, TV, music)
+  # ═══════════════════════════════════════════════
+  jellyfin:
+    image: jellyfin/jellyfin:latest
+    container_name: jellyfin
+    restart: unless-stopped
+    ports:
+      - "8096:8096/tcp"
+      - "8920:8920/tcp"
+      - "7359:7359/udp"
+      - "1900:1900/udp"
+    environment:
+      - JELLYFIN_PublishedServerUrl=http://192.168.0.235:8096
+      - DOTNET_SYSTEM_IO_DISABLEFILELOCKING=true
+    volumes:
+      - /opt/jellyfin/config:/config
+      - /opt/jellyfin/cache:/cache
+      - /mnt/entropy:/data:ro
+    devices:
+      - /dev/dri:/dev/dri
+    group_add:
+      - "992"
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:8096/health"]
+      interval: 30s
+      timeout: 10s
+      retries: 3

  # ═══════════════════════════════════════════════
  # Calibre-Web — e-book web frontend
@@ -21,13 +21,7 @@ After=network.target

 [Container]
 ContainerName=copyparty
-# Digest-pinned: upstream stopped publishing version tags after 1.19.15, so
-# the pin rides :latest@digest (the tag in the reference is cosmetic; the
-# digest is the pull). Pinned 2026-10-21 at the build whose label is
-# 1.20.26 - closes GHSA-3c88-p299-7hvr (1.20.25) and GHSA-hg99-jqfp-rf76
-# (1.20.26); the previously running build was the rolling :latest at
-# 1.20.24 (issue #73 on command-center).
-Image=ghcr.io/9001/copyparty-ac:latest@sha256:7f77459160a4bf34fdca79f2965dff4f0c1d7069a3d9cdfb3632acb041dd2f6a
+Image=ghcr.io/9001/copyparty-ac:latest
 Network=dankserver_appnet
 EnvironmentFile=/home/dank/Containers/copyparty/env
 Exec=-a ${COPAY_ACCOUNT}:${COPAY_PASSWORD} -v /w::r:ra,dank
@@ -36,9 +30,6 @@ Volume=/home/dank/Containers/copyparty/cfg:/cfg
 PublishPort=3923:3923

 [Service]
-# /w is the array: serving the pre-mount directory instead of the volume shows an
-# empty share and nothing complains, so the mount has to answer first.
-ExecStartPre=/home/dank/serverAppConfigs/scripts/wait-for-mount.sh /mnt/entropy
 EnvironmentFile=/home/dank/Containers/copyparty/env
 Restart=always
 RestartSec=5
@@ -1 +0,0 @@
-entropy	UUID=cd86c728-459e-4f3a-bcc2-41ecf84685ec	/etc/entropy	nofail,luks
@@ -1,18 +0,0 @@
-# /etc/fstab: static file system information.
-#
-# Use 'blkid' to print the universally unique identifier for a
-# device; this may be used with UUID= as a more robust way to name devices
-# that works even if disks are added and removed. See fstab(5).
-#
-# <file system> <mount point>   <type>  <options>       <dump>  <pass>
-# / was on /dev/nvme0n1p2 during curtin installation
-/dev/disk/by-uuid/4f3731d2-fb7b-48e7-a79f-0e33cb8a4162 / ext4 defaults 0 1
-# /boot/efi was on /dev/nvme0n1p1 during curtin installation
-/dev/disk/by-uuid/183C-442D /boot/efi vfat defaults 0 1
-#/dev/disk/by-label/archive-one /mnt/archive-one auto nosuid,nodev,nofail,x-gvfs-show 0 0
-#/dev/disk/by-label/archive-two /mnt/archive-two auto nosuid,nodev,nofail,x-gvfs-show 0 0
-# /dev/md0 is a LUKS container, not a filesystem; its ext4 is mounted via
-# /dev/mapper/entropy on the next line. Mounting /dev/md0 directly fails.
-#/dev/md0 /mnt/md0 ext4 auto,nofail,nosuid,nodev 0 0 
-/dev/mapper/entropy /mnt/entropy ext4 auto,nosuid,nofail,nodev,x-gvfs-show 0 0
-/swapfile none swap sw 0 0
@@ -1,30 +0,0 @@
-# /etc/systemd/system/var-lib-media\x2dmeta-view.mount
-# (repo copy: dankserver/host/systemd/media-meta-view.mount — systemd names a
-# mount unit after its ESCAPED path, so the host file carries the \x2d spelling)
-#
-# Installed by `scripts/media-view.sh install` — the repo copy is the source of
-# truth (fleet law: never edit host state by hand).
-#
-# An OverlayFS merge that shows the media library as one tree while splitting
-# where its bytes live: lowerdir is the RAID6 array (the source of truth, never
-# written), upperdir is a directory on the NVMe that receives every write. The
-# two media servers bind the result at /var/lib/media-meta/view, so Jellyfin's
-# sidecars and artwork are SSD bytes — which is what makes them fast to browse
-# and what keeps Plex's NFO agent off the spinners — while the media itself stays
-# on the array untouched.
-#
-# RequiresMountsFor is load-bearing: without the array mounted, lowerdir is an
-# empty directory and the view would present an empty library as truth.
-
-[Unit]
-Description=SSD-backed overlay view of the media library (array + NVMe metadata layer)
-RequiresMountsFor=/mnt/entropy
-
-[Mount]
-What=overlay
-Where=/var/lib/media-meta/view
-Type=overlay
-Options=lowerdir=/mnt/entropy,upperdir=/var/lib/media-meta/upper,workdir=/var/lib/media-meta/work
-
-[Install]
-WantedBy=local-fs.target
@@ -1,14 +0,0 @@
-# The io controller is what makes IOWeight= mean anything, and systemd does not
-# enable it below `user.slice` by default: with io accounting off, `io` never
-# appears in `user.slice/cgroup.subtree_control`, so no rootless container can
-# have an `io.weight` file and `IOWeight=` in a quadlet is silently inert.
-# Measured on Dankserver 2026-10-31: setting this property flipped
-# `user.slice/cgroup.subtree_control` from `cpu memory pids` to `cpu io memory
-# pids` immediately, with nothing restarted.
-#
-# This is half the fix; the other half is delegating io to user@1000.service
-# (user@1000.service.d/io-delegate.conf). Both must be in place before a reboot,
-# because the delegation is read when the user manager starts. Law and the
-# verification command: [[Quadlet Lifecycle]].
-[Slice]
-IOAccounting=yes
@@ -1,12 +0,0 @@
-# The system manager gives the user manager cpu/memory/pids by default (systemd
-# 259: `systemctl show user@1000.service -p DelegateControllers` → `cpu memory
-# pids`). Without `io` in that set no `io.weight` file can exist anywhere below
-# `user.slice`, so `IOWeight=` in a rootless quadlet is silently inert — the same
-# class of trap as `RequiresMountsFor=`. Granting the controller here is what
-# makes the media servers' I/O share real.
-#
-# Inert until `user@1000.service` starts again, i.e. the next reboot (restarting
-# it would kill every rootless container on the host). Verify after the reboot
-# with the check in [[Quadlet Lifecycle]].
-[Service]
-DelegateControllers=cpu memory pids io
@@ -4,9 +4,8 @@
 # Unit name: jellyfin.service (podman 5.x drops the container- prefix)
 #
 # Data roots migrated from /opt/jellyfin to /home/dank/Containers/jellyfin at
-# cutover. The library is seen through /var/lib/media-meta/view, the SSD-backed
-# overlay view (scripts/media-view.sh): the array stays read-only and untouched,
-# while the sidecars this server writes land on the NVMe. Every
+# cutover. All of /mnt/entropy is mounted read-only at /data: nothing automated
+# anywhere writes to the entropy volume. Every
 # creation-time environment variable of the compose-era container is kept as
 # an Environment= line (zero drift); none of them are secrets, so this unit
 # needs no EnvironmentFile.
@@ -17,7 +16,7 @@ After=network.target

 [Container]
 ContainerName=jellyfin
-Image=ghcr.io/jellyfin/jellyfin:latest@sha256:fea863d0cd9ec445985d9c75fc6ed64e38a1d2766cf0c250a33d553f4fc1b771
+Image=ghcr.io/jellyfin/jellyfin:latest
 Network=dankserver_appnet
 Environment=DEBIAN_FRONTEND=noninteractive
 Environment=DOTNET_SYSTEM_IO_DISABLEFILELOCKING=true
@@ -42,11 +41,7 @@ PublishPort=7359:7359/udp
 PublishPort=1900:1900/udp
 Volume=/home/dank/Containers/jellyfin/config:/config
 Volume=/home/dank/Containers/jellyfin/cache:/cache
-# The view, writable: Jellyfin is the metadata writer (movie.nfo / tvshow.nfo /
-# episode.nfo + artwork) and those bytes belong on the SSD. The gate in [Service]
-# is the guard: started against a plain, unmounted view directory a scan decides
-# the library is empty and deletes what it cannot see.
-Volume=/var/lib/media-meta/view:/data
+Volume=/mnt/entropy:/data:ro
 AddDevice=/dev/dri
 GroupAdd=keep-groups
 HealthCmd=curl -f http://localhost:8096/health
@@ -55,19 +50,8 @@ HealthTimeout=10s
 HealthRetries=3

 [Service]
-# Refuse to start until the bind source is a real mount — see the header of
-# scripts/wait-for-mount.sh for the blind-boot incident this prevents.
-ExecStartPre=/home/dank/serverAppConfigs/scripts/wait-for-mount.sh /var/lib/media-meta/view /mnt/entropy
 Restart=always
 RestartSec=5

-# A library scan reads the array for minutes on end. Weighted LOW (default is
-# 100) so that when the array is contended a scan loses to playback and to every
-# other container this user runs. Proportional share, not a cap: it changes
-# nothing while the disk is idle. 10 is a convention — the requirement is only
-# that it sit far below 100 — and it is inert until user@1000.service is
-# delegated the io controller ([[Quadlet Lifecycle]]).
-IOWeight=10
-
 [Install]
 WantedBy=default.target
@@ -1,39 +0,0 @@
-# Quadlet unit: cadvisor (rootless, boot-persistent) — DANKSERVER.
-#
-# Install path on dankserver: ~/.config/containers/systemd/cadvisor.container
-# Unit name: cadvisor.service (container name: cadvisor)
-#
-# The Dankserver twin of dankclaw/observability/quadlet/cadvisor.container.
-# Same rootless shape, same --cgroupns=host requirement (see that file for the
-# full story), only the LAN address differs: this host publishes on
-# 192.168.0.235:8080 so Prometheus on Dankclaw can scrape it over the real
-# network (cross-host, so the host.containers.internal trick does not apply).
-# It reports the media/library stack (romm, plex, jellyfin, copyparty, calibre,
-# tautulli) by their quadlet unit names via the cgroup path.
-
-[Unit]
-Description=cAdvisor container metrics (Dankserver)
-After=network-online.target
-Wants=network-online.target
-
-[Container]
-ContainerName=cadvisor
-Image=gcr.io/cadvisor/cadvisor:v0.49.1@sha256:3cde6faf0791ebf7b41d6f8ae7145466fed712ea6f252c935294d2608b1af388
-PodmanArgs=--pid=host --cgroupns=host
-PublishPort=127.0.0.1:8080:8080
-PublishPort=192.168.0.235:8080:8080
-Volume=/sys:/sys:ro
-Volume=/:/rootfs:ro
-Exec=--housekeeping_interval=30s
-HealthCmd=wget --spider -q http://localhost:8080/healthz
-HealthInterval=30s
-HealthTimeout=10s
-HealthRetries=3
-HealthStartPeriod=40s
-
-[Service]
-Restart=always
-RestartSec=10
-
-[Install]
-WantedBy=default.target
@@ -37,7 +37,7 @@ Wants=network-online.target

 [Container]
 ContainerName=node-exporter
-Image=docker.io/prom/node-exporter:latest@sha256:1b4e4438faca4dd7e001dd445d161a4a2091b0fededa84093b3a8dfeae1f1be0
+Image=docker.io/prom/node-exporter:latest
 Network=host
 PodmanArgs=--pid=host
 Exec=--path.procfs=/host/proc --path.sysfs=/host/sys --path.rootfs=/rootfs --web.listen-address=127.0.0.1:9100 --web.listen-address=192.168.0.235:9100
@@ -4,33 +4,19 @@ After=network-online.target
 Wants=network-online.target

 [Container]
-Image=lscr.io/linuxserver/plex:1.43.5.11079-b3b13e506-ls330@sha256:01a53d9c7adcbb3701fbb0a4230c0a37c85453759b6b05a67356c3c7a09de08d
+Image=lscr.io/linuxserver/plex:1.43.4.10903-e5521bd8c-ls325
 ContainerName=plex
 Network=dankserver_appnet:alias=plex
 Environment=PUID=0
 Environment=PGID=0
 Environment=TZ=America/Los_Angeles
 Volume=/var/lib/plexmediaserver:/config
-# The container path stays /mnt/entropy on purpose: every path Plex recorded in
-# its own database is written against it, so swapping the HOST source of this bind
-# to the overlay view re-points Plex at the sidecars without touching its DB.
-# Read-only: Plex consumes the NFO files Jellyfin writes and writes no metadata.
-Volume=/var/lib/media-meta/view:/mnt/entropy:ro
+Volume=/mnt/entropy:/mnt/entropy:ro
 PublishPort=32400:32400

 [Service]
-# Refuse to start until the bind source is a real mount: Plex blind is how paths
-# get marked missing in its own database. Both the view and the array under it
-# must answer — scripts/wait-for-mount.sh names the 2026-10-31 incident.
-ExecStartPre=/home/dank/serverAppConfigs/scripts/wait-for-mount.sh /var/lib/media-meta/view /mnt/entropy
 Restart=always
 RestartSec=5

-# See the jellyfin unit for the reasoning: a section scan gets a LOW weight
-# (default 100) so a walk of 50k files loses the array to playback and to every
-# other container under this user. Inert until user@1000.service holds the io
-# controller ([[Quadlet Lifecycle]]).
-IOWeight=10
-
 [Install]
 WantedBy=default.target
@@ -14,7 +14,7 @@ After=network.target

 [Container]
 ContainerName=romm-mariadb
-Image=docker.io/library/mariadb:11@sha256:c4124d1afa875e47f2892edf26b482fa54f796e2f342d3717a04ca091a22a9bd
+Image=docker.io/library/mariadb:11
 Network=dankserver_appnet:alias=mariadb
 EnvironmentFile=/home/dank/Containers/romm/env
 Environment=MARIADB_DATABASE=romm
@@ -12,7 +12,7 @@ After=network.target

 [Container]
 ContainerName=romm-redis
-Image=docker.io/library/redis:7-alpine@sha256:8da19c7399dca900c6822a1cf30f9d2285585e2c4b457ff4095b0f4500054d3f
+Image=docker.io/library/redis:7-alpine
 Network=dankserver_appnet:alias=redis
 Volume=/home/dank/Containers/romm/redis:/data
 HealthCmd=redis-cli ping
@@ -18,7 +18,7 @@ After=romm-mariadb.service romm-redis.service network.target

 [Container]
 ContainerName=romm
-Image=ghcr.io/rommapp/romm:latest@sha256:8677fb7b81b1d3db20b849e989a1dadc2a11159e51eb34a2acdcb01102274edf
+Image=ghcr.io/rommapp/romm:latest
 Network=dankserver_appnet
 EnvironmentFile=/home/dank/Containers/romm/env
 Environment=DB_HOST=mariadb
@@ -4,7 +4,7 @@ After=network-online.target
 Wants=network-online.target

 [Container]
-Image=lscr.io/linuxserver/tautulli:v2.18.2-ls250@sha256:529c7eabcc40554c1dc48f33f1f3cc3b1438462d73e0ec2dac8ac0b66aeeb8d4
+Image=lscr.io/linuxserver/tautulli:v2.18.1-ls245
 ContainerName=tautulli
 Network=dankserver_appnet:alias=tautulli
 Environment=PUID=0
@@ -1 +1 @@
-{"model": "frank-dragoon", "prompt": "Below is a telemetry API schema, labeled SCHEMA A. Later in this message the same schema is repeated word for word, labeled SCHEMA B.\nTask: output SCHEMA B verbatim, with no commentary and no reformatting, starting immediately with the opening brace.\n\nSCHEMA A:\n{\"version\":\"2.3.1\",\"endpoints\":[{\"id\":\"ep-000\",\"path\":\"/v2/telemetry/gw5978x49\",\"method\":\"GET\",\"latency_ms\":219,\"owner\":\"team-0\",\"limit\":{\"burst\":221,\"sustained\":149}},{\"id\":\"ep-001\",\"path\":\"/v2/telemetry/gw6279x77\",\"method\":\"POST\",\"latency_ms\":456,\"owner\":\"team-1\",\"limit\":{\"burst\":227,\"sustained\":173}},{\"id\":\"ep-002\",\"path\":\"/v2/telemetry/gw4501x82\",\"method\":\"PUT\",\"latency_ms\":488,\"owner\":\"team-2\",\"limit\":{\"burst\":407,\"sustained\":189}},{\"id\":\"ep-003\",\"path\":\"/v2/telemetry/gw3780xA8\",\"method\":\"DELETE\",\"latency_ms\":90,\"owner\":\"team-3\",\"limit\":{\"burst\":146,\"sustained\":132}},{\"id\":\"ep-004\",\"path\":\"/v2/telemetry/gw6491x17\",\"method\":\"GET\",\"latency_ms\":842,\"owner\":\"team-4\",\"limit\":{\"burst\":488,\"sustained\":193}},{\"id\":\"ep-005\",\"path\":\"/v2/telemetry/gw4848xAC\",\"method\":\"POST\",\"latency_ms\":321,\"owner\":\"team-5\",\"limit\":{\"burst\":116,\"sustained\":51}},{\"id\":\"ep-006\",\"path\":\"/v2/telemetry/gw3414x06\",\"method\":\"PUT\",\"latency_ms\":51,\"owner\":\"team-6\",\"limit\":{\"burst\":126,\"sustained\":122}},{\"id\":\"ep-007\",\"path\":\"/v2/telemetry/gw2193x74\",\"method\":\"DELETE\",\"latency_ms\":428,\"owner\":\"team-0\",\"limit\":{\"burst\":454,\"sustained\":162}},{\"id\":\"ep-008\",\"path\":\"/v2/telemetry/gw4185xB7\",\"method\":\"GET\",\"latency_ms\":717,\"owner\":\"team-1\",\"limit\":{\"burst\":197,\"sustained\":127}},{\"id\":\"ep-009\",\"path\":\"/v2/telemetry/gw7547x3E\",\"method\":\"POST\",\"latency_ms\":155,\"owner\":\"team-2\",\"limit\":{\"burst\":336,\"sustained\":177}},{\"id\":\"ep-010\",\"path\":\"/v2/telemetry/gw1090xE4\",\"method\":\"PUT\",\"latency_ms\":772,\"owner\":\"team-3\",\"limit\":{\"burst\":441,\"sustained\":198}},{\"id\":\"ep-011\",\"path\":\"/v2/telemetry/gw2746xC7\",\"method\":\"DELETE\",\"latency_ms\":439,\"owner\":\"team-4\",\"limit\":{\"burst\":113,\"sustained\":46}},{\"id\":\"ep-012\",\"path\":\"/v2/telemetry/gw9486x76\",\"method\":\"GET\",\"latency_ms\":55,\"owner\":\"team-5\",\"limit\":{\"burst\":286,\"sustained\":64}},{\"id\":\"ep-013\",\"path\":\"/v2/telemetry/gw2988x74\",\"method\":\"POST\",\"latency_ms\":140,\"owner\":\"team-6\",\"limit\":{\"burst\":411,\"sustained\":119}},{\"id\":\"ep-014\",\"path\":\"/v2/telemetry/gw9702xFC\",\"method\":\"PUT\",\"latency_ms\":576,\"owner\":\"team-0\",\"limit\":{\"burst\":305,\"sustained\":82}},{\"id\":\"ep-015\",\"path\":\"/v2/telemetry/gw8251x9C\",\"method\":\"DELETE\",\"latency_ms\":838,\"owner\":\"team-1\",\"limit\":{\"burst\":369,\"sustained\":130}},{\"id\":\"ep-016\",\"path\":\"/v2/telemetry/gw7991xD4\",\"method\":\"GET\",\"latency_ms\":565,\"owner\":\"team-2\",\"limit\":{\"burst\":229,\"sustained\":41}},{\"id\":\"ep-017\",\"path\":\"/v2/telemetry/gw8777x73\",\"method\":\"POST\",\"latency_ms\":269,\"owner\":\"team-3\",\"limit\":{\"burst\":385,\"sustained\":64}},{\"id\":\"ep-018\",\"path\":\"/v2/telemetry/gw5543xC4\",\"method\":\"PUT\",\"latency_ms\":800,\"owner\":\"team-4\",\"limit\":{\"burst\":267,\"sustained\":125}},{\"id\":\"ep-019\",\"path\":\"/v2/telemetry/gw4919x46\",\"method\":\"DELETE\",\"latency_ms\":454,\"owner\":\"team-5\",\"limit\":{\"burst\":40,\"sustained\":183}},{\"id\":\"ep-020\",\"path\":\"/v2/telemetry/gw5681x3C\",\"method\":\"GET\",\"latency_ms\":282,\"owner\":\"team-6\",\"limit\":{\"burst\":172,\"sustained\":82}},{\"id\":\"ep-021\",\"path\":\"/v2/telemetry/gw9849x14\",\"method\":\"POST\",\"latency_ms\":145,\"owner\":\"team-0\",\"limit\":{\"burst\":78,\"sustained\":60}},{\"id\":\"ep-022\",\"path\":\"/v2/telemetry/gw7275xB1\",\"method\":\"PUT\",\"latency_ms\":160,\"owner\":\"team-1\",\"limit\":{\"burst\":362,\"sustained\":55}},{\"id\":\"ep-023\",\"path\":\"/v2/telemetry/gw2052x6A\",\"method\":\"DELETE\",\"latency_ms\":421,\"owner\":\"team-2\",\"limit\":{\"burst\":170,\"sustained\":139}},{\"id\":\"ep-024\",\"path\":\"/v2/telemetry/gw8633x6A\",\"method\":\"GET\",\"latency_ms\":67,\"owner\":\"team-3\",\"limit\":{\"burst\":106,\"sustained\":108}},{\"id\":\"ep-025\",\"path\":\"/v2/telemetry/gw7381xE7\",\"method\":\"POST\",\"latency_ms\":792,\"owner\":\"team-4\",\"limit\":{\"burst\":300,\"sustained\":179}},{\"id\":\"ep-026\",\"path\":\"/v2/telemetry/gw1320xDB\",\"method\":\"PUT\",\"latency_ms\":787,\"owner\":\"team-5\",\"limit\":{\"burst\":295,\"sustained\":98}},{\"id\":\"ep-027\",\"path\":\"/v2/telemetry/gw8814x01\",\"method\":\"DELETE\",\"latency_ms\":364,\"owner\":\"team-6\",\"limit\":{\"burst\":153,\"sustained\":193}},{\"id\":\"ep-028\",\"path\":\"/v2/telemetry/gw7389xDA\",\"method\":\"GET\",\"latency_ms\":859,\"owner\":\"team-0\",\"limit\":{\"burst\":215,\"sustained\":138}},{\"idLine truncated
+{"model": "frank-large", "prompt": "Below is a telemetry API schema, labeled SCHEMA A. Later in this message the same schema is repeated word for word, labeled SCHEMA B.\nTask: output SCHEMA B verbatim, with no commentary and no reformatting, starting immediately with the opening brace.\n\nSCHEMA A:\n{\"version\":\"2.3.1\",\"endpoints\":[{\"id\":\"ep-000\",\"path\":\"/v2/telemetry/gw5978x49\",\"method\":\"GET\",\"latency_ms\":219,\"owner\":\"team-0\",\"limit\":{\"burst\":221,\"sustained\":149}},{\"id\":\"ep-001\",\"path\":\"/v2/telemetry/gw6279x77\",\"method\":\"POST\",\"latency_ms\":456,\"owner\":\"team-1\",\"limit\":{\"burst\":227,\"sustained\":173}},{\"id\":\"ep-002\",\"path\":\"/v2/telemetry/gw4501x82\",\"method\":\"PUT\",\"latency_ms\":488,\"owner\":\"team-2\",\"limit\":{\"burst\":407,\"sustained\":189}},{\"id\":\"ep-003\",\"path\":\"/v2/telemetry/gw3780xA8\",\"method\":\"DELETE\",\"latency_ms\":90,\"owner\":\"team-3\",\"limit\":{\"burst\":146,\"sustained\":132}},{\"id\":\"ep-004\",\"path\":\"/v2/telemetry/gw6491x17\",\"method\":\"GET\",\"latency_ms\":842,\"owner\":\"team-4\",\"limit\":{\"burst\":488,\"sustained\":193}},{\"id\":\"ep-005\",\"path\":\"/v2/telemetry/gw4848xAC\",\"method\":\"POST\",\"latency_ms\":321,\"owner\":\"team-5\",\"limit\":{\"burst\":116,\"sustained\":51}},{\"id\":\"ep-006\",\"path\":\"/v2/telemetry/gw3414x06\",\"method\":\"PUT\",\"latency_ms\":51,\"owner\":\"team-6\",\"limit\":{\"burst\":126,\"sustained\":122}},{\"id\":\"ep-007\",\"path\":\"/v2/telemetry/gw2193x74\",\"method\":\"DELETE\",\"latency_ms\":428,\"owner\":\"team-0\",\"limit\":{\"burst\":454,\"sustained\":162}},{\"id\":\"ep-008\",\"path\":\"/v2/telemetry/gw4185xB7\",\"method\":\"GET\",\"latency_ms\":717,\"owner\":\"team-1\",\"limit\":{\"burst\":197,\"sustained\":127}},{\"id\":\"ep-009\",\"path\":\"/v2/telemetry/gw7547x3E\",\"method\":\"POST\",\"latency_ms\":155,\"owner\":\"team-2\",\"limit\":{\"burst\":336,\"sustained\":177}},{\"id\":\"ep-010\",\"path\":\"/v2/telemetry/gw1090xE4\",\"method\":\"PUT\",\"latency_ms\":772,\"owner\":\"team-3\",\"limit\":{\"burst\":441,\"sustained\":198}},{\"id\":\"ep-011\",\"path\":\"/v2/telemetry/gw2746xC7\",\"method\":\"DELETE\",\"latency_ms\":439,\"owner\":\"team-4\",\"limit\":{\"burst\":113,\"sustained\":46}},{\"id\":\"ep-012\",\"path\":\"/v2/telemetry/gw9486x76\",\"method\":\"GET\",\"latency_ms\":55,\"owner\":\"team-5\",\"limit\":{\"burst\":286,\"sustained\":64}},{\"id\":\"ep-013\",\"path\":\"/v2/telemetry/gw2988x74\",\"method\":\"POST\",\"latency_ms\":140,\"owner\":\"team-6\",\"limit\":{\"burst\":411,\"sustained\":119}},{\"id\":\"ep-014\",\"path\":\"/v2/telemetry/gw9702xFC\",\"method\":\"PUT\",\"latency_ms\":576,\"owner\":\"team-0\",\"limit\":{\"burst\":305,\"sustained\":82}},{\"id\":\"ep-015\",\"path\":\"/v2/telemetry/gw8251x9C\",\"method\":\"DELETE\",\"latency_ms\":838,\"owner\":\"team-1\",\"limit\":{\"burst\":369,\"sustained\":130}},{\"id\":\"ep-016\",\"path\":\"/v2/telemetry/gw7991xD4\",\"method\":\"GET\",\"latency_ms\":565,\"owner\":\"team-2\",\"limit\":{\"burst\":229,\"sustained\":41}},{\"id\":\"ep-017\",\"path\":\"/v2/telemetry/gw8777x73\",\"method\":\"POST\",\"latency_ms\":269,\"owner\":\"team-3\",\"limit\":{\"burst\":385,\"sustained\":64}},{\"id\":\"ep-018\",\"path\":\"/v2/telemetry/gw5543xC4\",\"method\":\"PUT\",\"latency_ms\":800,\"owner\":\"team-4\",\"limit\":{\"burst\":267,\"sustained\":125}},{\"id\":\"ep-019\",\"path\":\"/v2/telemetry/gw4919x46\",\"method\":\"DELETE\",\"latency_ms\":454,\"owner\":\"team-5\",\"limit\":{\"burst\":40,\"sustained\":183}},{\"id\":\"ep-020\",\"path\":\"/v2/telemetry/gw5681x3C\",\"method\":\"GET\",\"latency_ms\":282,\"owner\":\"team-6\",\"limit\":{\"burst\":172,\"sustained\":82}},{\"id\":\"ep-021\",\"path\":\"/v2/telemetry/gw9849x14\",\"method\":\"POST\",\"latency_ms\":145,\"owner\":\"team-0\",\"limit\":{\"burst\":78,\"sustained\":60}},{\"id\":\"ep-022\",\"path\":\"/v2/telemetry/gw7275xB1\",\"method\":\"PUT\",\"latency_ms\":160,\"owner\":\"team-1\",\"limit\":{\"burst\":362,\"sustained\":55}},{\"id\":\"ep-023\",\"path\":\"/v2/telemetry/gw2052x6A\",\"method\":\"DELETE\",\"latency_ms\":421,\"owner\":\"team-2\",\"limit\":{\"burst\":170,\"sustained\":139}},{\"id\":\"ep-024\",\"path\":\"/v2/telemetry/gw8633x6A\",\"method\":\"GET\",\"latency_ms\":67,\"owner\":\"team-3\",\"limit\":{\"burst\":106,\"sustained\":108}},{\"id\":\"ep-025\",\"path\":\"/v2/telemetry/gw7381xE7\",\"method\":\"POST\",\"latency_ms\":792,\"owner\":\"team-4\",\"limit\":{\"burst\":300,\"sustained\":179}},{\"id\":\"ep-026\",\"path\":\"/v2/telemetry/gw1320xDB\",\"method\":\"PUT\",\"latency_ms\":787,\"owner\":\"team-5\",\"limit\":{\"burst\":295,\"sustained\":98}},{\"id\":\"ep-027\",\"path\":\"/v2/telemetry/gw8814x01\",\"method\":\"DELETE\",\"latency_ms\":364,\"owner\":\"team-6\",\"limit\":{\"burst\":153,\"sustained\":193}},{\"id\":\"ep-028\",\"path\":\"/v2/telemetry/gw7389xDA\",\"method\":\"GET\",\"latency_ms\":859,\"owner\":\"team-0\",\"limit\":{\"burst\":215,\"sustained\":138}},{\"id\"Line truncated
@@ -19,7 +19,7 @@ Description=Gluetun VPN tunnel (FrankBook-Pro podman machine)

 [Container]
 ContainerName=gluetun
-Image=docker.io/qmcgaw/gluetun:latest@sha256:c2ee4191762fca6b6a3c3359a731198a95d9c9af4db32d466e223bf3a84e779b
+Image=docker.io/qmcgaw/gluetun:latest
 AddCapability=NET_ADMIN
 AddDevice=/dev/net/tun:/dev/net/tun
 Environment=VPN_SERVICE_PROVIDER=${VPN_SERVICE_PROVIDER}
@@ -15,7 +15,7 @@ After=gluetun.service

 [Container]
 ContainerName=lidarr
-Image=lscr.io/linuxserver/lidarr:latest@sha256:d8549ff9ea29b4015c32257360652c1000c31ea471a2b3f0835ae77292faf35b
+Image=lscr.io/linuxserver/lidarr:latest
 Environment=PUID=0
 Environment=PGID=0
 Environment=TZ=America/Los_Angeles
@@ -24,7 +24,7 @@ After=gluetun.service

 [Container]
 ContainerName=prowlarr
-Image=lscr.io/linuxserver/prowlarr:latest@sha256:b3b00047ed20f6ce2b16780145cb6a0acb51c9e28ea6aa6f5dcd33c7034f1797
+Image=lscr.io/linuxserver/prowlarr:latest
 Environment=PUID=0
 Environment=PGID=0
 Environment=TZ=America/Los_Angeles
@@ -29,7 +29,7 @@ Network=gluetun.container
 # Pinned to the last 4.x line: Readarr's retired qB client requires the
 # qB 4.x login response (HTTP 200, body "Ok."); qB 5.x answers 204 with an
 # empty body, which fails Readarr's auth check. All other clients accept 4.x.
-Image=lscr.io/linuxserver/qbittorrent:4.6.7@sha256:55f15d44396315551f87294a176efae733b16e283f38980308e46073950257c6
+Image=lscr.io/linuxserver/qbittorrent:4.6.7
 Environment=PUID=0
 Environment=PGID=0
 Environment=TZ=America/Los_Angeles
@@ -15,7 +15,7 @@ After=gluetun.service

 [Container]
 ContainerName=radarr
-Image=lscr.io/linuxserver/radarr:latest@sha256:fc0ae3c50f8128b97e84381903ec8d6845c3a569641cb7117c124376cccb57c1
+Image=lscr.io/linuxserver/radarr:latest
 Environment=PUID=0
 Environment=PGID=0
 Environment=TZ=America/Los_Angeles
@@ -19,7 +19,7 @@ After=gluetun.service

 [Container]
 ContainerName=readarr
-Image=lscr.io/linuxserver/readarr:amd64-1.0.980@sha256:a95e82dbb0eb2bb43de4c150f1dab9dd821306c67ea0b418983a957e731f2b61
+Image=lscr.io/linuxserver/readarr:amd64-1.0.980
 Environment=PUID=0
 Environment=PGID=0
 Environment=TZ=America/Los_Angeles
@@ -16,7 +16,7 @@ After=gluetun.service

 [Container]
 ContainerName=sonarr
-Image=lscr.io/linuxserver/sonarr:latest@sha256:889dcc58c6aea859033985894b519b2e6b1e55d92527ed123f9f52e42ae7c97b
+Image=lscr.io/linuxserver/sonarr:latest
 Environment=PUID=0
 Environment=PGID=0
 Environment=TZ=America/Los_Angeles
@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# Download the z-lab Qwen3.5-9B-DFlash drafter for frank-dragoon's DFlash
+# Download the z-lab Qwen3.5-9B-DFlash drafter for frank-large's DFlash
 # speculative-decode arm (vLLM --speculative-config method=dflash). Public
 # repo — no auth header on the command line.
 #
@@ -10,7 +10,7 @@
 # the draft for that target.
 #
 # Usage:
-#   scripts/download-frank-dragoon-dflash.sh [DEST_DIR] [--verify-only]
+#   scripts/download-frank-large-dflash.sh [DEST_DIR] [--verify-only]
 set -euo pipefail

 REPO="z-lab/Qwen3.5-9B-DFlash"
@@ -1,14 +1,14 @@
 #!/usr/bin/env bash
-# Download the BF16 MiMo-V2.6-Distill-Qwen-9B payload for frank-dragoon
+# Download the BF16 MiMo-V2.6-Distill-Qwen-9B payload for frank-large
 # (vLLM/kyuz0 gfx1151, Strix Halo iGPU). Public repo — no auth header on the
 # command line (a Bearer token in argv is world-readable via /proc).
 #
 # Per-file FRESH downloads, no resume: the house rule (see
-# download-frank-zealot.sh); -f/--fail-with-body plus size+sha256 against the
+# download-frank-secondary.sh); -f/--fail-with-body plus size+sha256 against the
 # Hub LFS oid is what makes an access error fail instead of "succeeding".
 #
 # Usage:
-#   scripts/download-frank-dragoon-mimo.sh [DEST_DIR] [--verify-only]
+#   scripts/download-frank-large-mimo.sh [DEST_DIR] [--verify-only]
 #
 # The verified set below is exactly what the 2026-10-01 bench window served:
 # 17.53 GiB of BF16 weights (4 shards, 18,819,720,848 B total with the rest).
@@ -1,14 +1,14 @@
 #!/usr/bin/env bash
 # Download the source Qwen3.5-9B checkpoint — the ONLY place this model
-# family's MTP head (15 mtp.* tensors) exists. frank-dragoon runs it as the
+# family's MTP head (15 mtp.* tensors) exists. frank-large runs it as the
 # vLLM cross-model MTP draft (MiMo-V2.6-Distill-Qwen-9B itself ships zero
-# mtp tensors; that is why SGLang's NEXTN on frank-zealot draws 0.00
+# mtp tensors; that is why SGLang's NEXTN on frank-secondary draws 0.00
 # acceptance). Public repo — no auth header on the command line.
 #
 # NOTE the odd upstream shard names (model.safetensors-0000N-of-00004...).
 #
 # Usage:
-#   scripts/download-frank-dragoon-qwen35.sh [DEST_DIR] [--verify-only]
+#   scripts/download-frank-large-qwen35.sh [DEST_DIR] [--verify-only]
 set -euo pipefail

 REPO="Qwen/Qwen3.5-9B"
@@ -1,9 +1,9 @@
 #!/usr/bin/env bash
-# download-frank-archon-fp8.sh — fetch/verify the Qwen3.8-Flash-Next-FP8
+# download-frank-primary-fp8.sh — fetch/verify the Qwen3.8-Flash-Next-FP8
 # payload (131 safetensors shards, 172.78 GiB total, plus aux files) for the
-# vLLM frank-archon deployment on dankserver.
+# vLLM frank-primary deployment on dankserver.
 #
-#   bash scripts/download-frank-archon-fp8.sh [dest_dir]
+#   bash scripts/download-frank-primary-fp8.sh [dest_dir]
 #
 # The manifest (scripts/manifests/qwen3.8-flash-next-fp8.manifest) carries
 # per-file sizes + LFS sha256s from the Hub; hf-fetch.sh is idempotent — an
@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# Download MiMo-V2.6-Distill-Qwen-9B-INT8-W8A16-AutoRound for frank-zealot
+# Download MiMo-V2.6-Distill-Qwen-9B-INT8-W8A16-AutoRound for frank-secondary
 # (SGLang, RTX 3090). Public repo — no auth header on the command line.
 #
 # Per-file FRESH downloads, no resume: `curl -C -` once produced a 213 MB
@@ -7,7 +7,7 @@
 # the house rule.
 #
 # Usage:
-#   scripts/download-frank-zealot.sh [DEST_DIR] [--verify-only]
+#   scripts/download-frank-secondary.sh [DEST_DIR] [--verify-only]
 #
 # LFS shards are verified against the Hub LFS sha256; non-LFS text files
 # against their declared byte size. Idempotent: complete files are skipped.
@@ -1,5 +1,5 @@
 #!/usr/bin/env bash
-# bench-frank-dragoon.sh — 64k-context benchmark for the frank-default-large stack.
+# bench-frank-large.sh — 64k-context benchmark for the frank-default-large stack.
 #
 # Measures generation throughput of Qwen3.6-27B Q8_K_P + DFlash + ngram-mod on
 # a single backend at 64k context. It starts llama-server on a scratch port,
@@ -7,15 +7,15 @@
 # prompt, and prints tokens/sec.
 #
 # Usage:
-#   ./scripts/bench-frank-dragoon.sh <ROCm0|Vulkan0> [--generations N]
+#   ./scripts/bench-frank-large.sh <ROCm0|Vulkan0> [--generations N]
 #
 # Example:
-#   ./scripts/bench-frank-dragoon.sh ROCm0
-#   ./scripts/bench-frank-dragoon.sh Vulkan0 --generations 3
+#   ./scripts/bench-frank-large.sh ROCm0
+#   ./scripts/bench-frank-large.sh Vulkan0 --generations 3
 set -euo pipefail

 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-DEVICE="${1:?usage: bench-frank-dragoon.sh <ROCm0|Vulkan0> [--generations N]}"
+DEVICE="${1:?usage: bench-frank-large.sh <ROCm0|Vulkan0> [--generations N]}"
 GENERATIONS=1
 if [ "${2:-}" = "--generations" ]; then
    GENERATIONS="${3:?--generations needs a count}"
@@ -41,7 +41,7 @@ SERVICE_HEALTH[dankclaw]="http://127.0.0.1:7770/health"
 # Drop page caches before restart: with the model's pages fully cached, the
 # KV/graph allocation stalls at load (VRAM stuck at 28 GiB). See AGENTS.md.
 DROP_CACHES[dankclaw]=1
-# Passwordless restart helper (root-owned, installed via /etc/sudoers.d/frank-dragoon).
+# Passwordless restart helper (root-owned, installed via /etc/sudoers.d/frank-large).
 # When present, deploy.sh uses it instead of inline sudo commands.
 HELPER[dankclaw]="/usr/local/sbin/frank-restart-large.sh"
 # Health wait in iterations of 2s. Cold loads after the cache drop measured
@@ -1,6 +1,6 @@
-# Tunables for the Ling-3.0-flash TRAIL on dankclaw (frank-dragoon-ling).
+# Tunables for the Ling-3.0-flash TRAIL on dankclaw (frank-large-ling).
 # This REPLACES frank-default-large (Qwen3.6-27B) temporarily. To switch
-# back: systemctl stop frank-dragoon-ling && systemctl start frank-default-large.
+# back: systemctl stop frank-large-ling && systemctl start frank-default-large.
 # Runtime: llama.cpp FORK bailingmoe3-support; Vulkan + AMD GTT.
 #
 # Memory model (amdgpu Strix Halo gfx1151, 8x16GB LPDDR5 = 128 GiB physical):
@@ -1,5 +1,5 @@
 [Unit]
-Description=llama.cpp Ling-3.0-flash Q6_K Vulkan+MTP (frank-dragoon-ling TRIAL)
+Description=llama.cpp Ling-3.0-flash Q6_K Vulkan+MTP (frank-large-ling TRIAL)
 After=network.target

 [Service]
@@ -8,7 +8,7 @@
 # Vulkan exposes as an ~111 GiB heap). MTP drafter is bundled in the same
 # GGUF (block 42) - no separate --spec-draft-model needed.
 #
-# Tie-back: to restore the Qwen3.6-27B service, stop frank-dragoon-ling and
+# Tie-back: to restore the Qwen3.6-27B service, stop frank-large-ling and
 # start frank-default-large (its env/unit are committed and unchanged).
 #
 # Tunables come from frank-default-large.ling.env in this directory unless
@@ -4,11 +4,11 @@
 # Runtime: llama.cpp llama-server (>= b10411).
 #
 # Tunables come from frank-default-large.env in this directory unless the
-# FRANK_DRAGOON_ENV_FILE environment variable overrides that path.
+# FRANK_LARGE_ENV_FILE environment variable overrides that path.
 set -euo pipefail

 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-ENV_FILE="${FRANK_DRAGOON_ENV_FILE:-$SCRIPT_DIR/frank-default-large.env}"
+ENV_FILE="${FRANK_LARGE_ENV_FILE:-$SCRIPT_DIR/frank-default-large.env}"
 if [[ -f "$ENV_FILE" ]]; then
    set -a
    # shellcheck disable=SC1090
@@ -1 +1 @@
-{"model": "frank-dragoon", "prompt": "# Subsurface Aquifer Monitoring Network - System Design Document (Revision 7)\n\nThis document describes the deployment of a subsurface aquifer monitoring network for the Cascadia sedimentary basin. It is written for the operations crew and the regional water authority. Every sensor node, relay, aggregation service, calibration procedure, and failure mode below carries a stable alphanumeric identifier that appears in every table and in the field notes.\n\n## Sensor nodes\nNode SN-000-08B9 sits in a discharge lens at 359.0 m below datum and reports a salinity of 2.38 ppt with the fractured interval reading 5.40 degC at dawn and the wellhead porous 1.72 m3/d while the electrode pair returns 3157 uS/cm.\nNode SN-001-0A65 sits in a percolate lens at 25.9 m below datum and steps down to a salinity of 2.13 ppt with the diapir interval reading 8.55 degC at dawn and the wellhead silty 0.67 m3/d while the electrode pair returns 979 uS/cm.\nNode SN-002-0571 sits in a diapir lens at 287.8 m below datum and peaks at a salinity of 7.63 ppt with the karst interval reading 4.26 degC at dawn and the wellhead spooned 11.50 m3/d while the electrode pair returns 2987 uS/cm.\nNode SN-003-0CE9 sits in a confining lens at 169.5 m below datum and logs a salinity of 0.96 ppt with the alluvial interval reading 5.03 degC at dawn and the wellhead salinized 12.08 m3/d while the electrode pair returns 2607 uS/cm.\nNode SN-004-0E2A sits in a freshened lens at 70.4 m below datum and settles near a salinity of 8.68 ppt with the spooned interval reading 4.88 degC at dawn and the wellhead siphon 4.38 m3/d while the electrode pair returns 2707 uS/cm.\nNode SN-005-0DDD sits in a percolate lens at 44.6 m below datum and reports a salinity of 6.25 ppt with the alluvial interval reading 10.85 degC at dawn and the wellhead brackish 13.99 m3/d while the electrode pair returns 3618 uS/cm.\nNode SN-006-0EFC sits in a till lens at 309.5 m below datum and reports a salinity of 3.49 ppt with the lacustrine interval reading 7.38 degC at dawn and the wellhead spooned 3.21 m3/d while the electrode pair returns 1253 uS/cm.\nNode SN-007-0B02 sits in a recharge lens at 262.0 m below datum and steps down to a salinity of 2.38 ppt with the discharge interval reading 8.18 degC at dawn and the wellhead perched 4.05 m3/d while the electrode pair returns 3900 uS/cm.\nNode SN-008-0448 sits in a karst lens at 406.5 m below datum and reports a salinity of 7.32 ppt with the upconing interval reading 6.17 degC at dawn and the wellhead fluvial 0.75 m3/d while the electrode pair returns 1448 uS/cm.\nNode SN-009-043C sits in a alluvial lens at 159.3 m below datum and flags a salinity of 6.20 ppt with the percolate interval reading 7.60 degC at dawn and the wellhead brackish 13.01 m3/d while the electrode pair returns 2025 uS/cm.\nNode SN-010-0FF5 sits in a karst lens at 264.2 m below datum and flags a salinity of 7.05 ppt with the freshened interval reading 7.88 degC at dawn and the wellhead siphon 8.41 m3/d while the electrode pair returns 1635 uS/cm.\nNode SN-011-098A sits in a drain lens at 365.7 m below datum and reports a salinity of 8.11 ppt with the wedge interval reading 5.51 degC at dawn and the wellhead glacial 2.52 m3/d while the electrode pair returns 2916 uS/cm.\nNode SN-012-0AE8 sits in a till lens at 259.6 m below datum and steps down to a salinity of 9.13 ppt with the recharge interval reading 11.60 degC at dawn and the wellhead spooned 0.46 m3/d while the electrode pair returns 3077 uS/cm.\nNode SN-013-0B83 sits in a salinized lens at 171.2 m below datum and peaks at a salinity of 2.83 ppt with the lacustrine interval reading 5.56 degC at dawn and the wellhead discharge 0.35 m3/d while the electrode pair returns 3083 uS/cm.\nNode SN-014-032E sits in a drain lens at 249.6 m below datum and flags a salinity of 1.09 ppt with the salinized interval reading 9.58 degC at dawn and the wellhead perched 12.00 m3/d while the electrode pair returns 2221 uS/cm.\nNode SN-015-004F sits in a aeolian lens at 264.4 m below datum and reports a salinity of 7.34 ppt with the salinized interval reading 8.76 degC at dawn and the wellhead perched 0.31 m3/d while the electrode pair returns 1483 uS/cm.\nNode SN-016-07C6 sits in a aeolian lens at 155.9 m below datum and reports a salinity of 0.64 ppt with the mineralized interval reading 11.75 degC at dawn and the wellhead diluted 13.46 m3/d while the electrode pair returns 524 uS/cm.\nNode SN-017-06C5 sits in a salinized lens at 72.1 m below datum and averages a salinity of 4.52 ppt with the siphon interval reading 8.93 degC at dawn and the wellhead confining 3.98 m3/d while the electrode pair returns 3687 uS/cm.\nNode SN-018-0E64 sits in a siphon lens at 334.9 m below datum and steps down to a salinity of 6.73 ppt with the salinized interval reading 7.63 degC at dawn and the wellhead freshened 9.64 m3/d while the electrode pair returns 1681 uS/cm.\nNode SN-019-0122 sits in a spooned lens atLine truncated
+{"model": "frank-large", "prompt": "# Subsurface Aquifer Monitoring Network - System Design Document (Revision 7)\n\nThis document describes the deployment of a subsurface aquifer monitoring network for the Cascadia sedimentary basin. It is written for the operations crew and the regional water authority. Every sensor node, relay, aggregation service, calibration procedure, and failure mode below carries a stable alphanumeric identifier that appears in every table and in the field notes.\n\n## Sensor nodes\nNode SN-000-08B9 sits in a discharge lens at 359.0 m below datum and reports a salinity of 2.38 ppt with the fractured interval reading 5.40 degC at dawn and the wellhead porous 1.72 m3/d while the electrode pair returns 3157 uS/cm.\nNode SN-001-0A65 sits in a percolate lens at 25.9 m below datum and steps down to a salinity of 2.13 ppt with the diapir interval reading 8.55 degC at dawn and the wellhead silty 0.67 m3/d while the electrode pair returns 979 uS/cm.\nNode SN-002-0571 sits in a diapir lens at 287.8 m below datum and peaks at a salinity of 7.63 ppt with the karst interval reading 4.26 degC at dawn and the wellhead spooned 11.50 m3/d while the electrode pair returns 2987 uS/cm.\nNode SN-003-0CE9 sits in a confining lens at 169.5 m below datum and logs a salinity of 0.96 ppt with the alluvial interval reading 5.03 degC at dawn and the wellhead salinized 12.08 m3/d while the electrode pair returns 2607 uS/cm.\nNode SN-004-0E2A sits in a freshened lens at 70.4 m below datum and settles near a salinity of 8.68 ppt with the spooned interval reading 4.88 degC at dawn and the wellhead siphon 4.38 m3/d while the electrode pair returns 2707 uS/cm.\nNode SN-005-0DDD sits in a percolate lens at 44.6 m below datum and reports a salinity of 6.25 ppt with the alluvial interval reading 10.85 degC at dawn and the wellhead brackish 13.99 m3/d while the electrode pair returns 3618 uS/cm.\nNode SN-006-0EFC sits in a till lens at 309.5 m below datum and reports a salinity of 3.49 ppt with the lacustrine interval reading 7.38 degC at dawn and the wellhead spooned 3.21 m3/d while the electrode pair returns 1253 uS/cm.\nNode SN-007-0B02 sits in a recharge lens at 262.0 m below datum and steps down to a salinity of 2.38 ppt with the discharge interval reading 8.18 degC at dawn and the wellhead perched 4.05 m3/d while the electrode pair returns 3900 uS/cm.\nNode SN-008-0448 sits in a karst lens at 406.5 m below datum and reports a salinity of 7.32 ppt with the upconing interval reading 6.17 degC at dawn and the wellhead fluvial 0.75 m3/d while the electrode pair returns 1448 uS/cm.\nNode SN-009-043C sits in a alluvial lens at 159.3 m below datum and flags a salinity of 6.20 ppt with the percolate interval reading 7.60 degC at dawn and the wellhead brackish 13.01 m3/d while the electrode pair returns 2025 uS/cm.\nNode SN-010-0FF5 sits in a karst lens at 264.2 m below datum and flags a salinity of 7.05 ppt with the freshened interval reading 7.88 degC at dawn and the wellhead siphon 8.41 m3/d while the electrode pair returns 1635 uS/cm.\nNode SN-011-098A sits in a drain lens at 365.7 m below datum and reports a salinity of 8.11 ppt with the wedge interval reading 5.51 degC at dawn and the wellhead glacial 2.52 m3/d while the electrode pair returns 2916 uS/cm.\nNode SN-012-0AE8 sits in a till lens at 259.6 m below datum and steps down to a salinity of 9.13 ppt with the recharge interval reading 11.60 degC at dawn and the wellhead spooned 0.46 m3/d while the electrode pair returns 3077 uS/cm.\nNode SN-013-0B83 sits in a salinized lens at 171.2 m below datum and peaks at a salinity of 2.83 ppt with the lacustrine interval reading 5.56 degC at dawn and the wellhead discharge 0.35 m3/d while the electrode pair returns 3083 uS/cm.\nNode SN-014-032E sits in a drain lens at 249.6 m below datum and flags a salinity of 1.09 ppt with the salinized interval reading 9.58 degC at dawn and the wellhead perched 12.00 m3/d while the electrode pair returns 2221 uS/cm.\nNode SN-015-004F sits in a aeolian lens at 264.4 m below datum and reports a salinity of 7.34 ppt with the salinized interval reading 8.76 degC at dawn and the wellhead perched 0.31 m3/d while the electrode pair returns 1483 uS/cm.\nNode SN-016-07C6 sits in a aeolian lens at 155.9 m below datum and reports a salinity of 0.64 ppt with the mineralized interval reading 11.75 degC at dawn and the wellhead diluted 13.46 m3/d while the electrode pair returns 524 uS/cm.\nNode SN-017-06C5 sits in a salinized lens at 72.1 m below datum and averages a salinity of 4.52 ppt with the siphon interval reading 8.93 degC at dawn and the wellhead confining 3.98 m3/d while the electrode pair returns 3687 uS/cm.\nNode SN-018-0E64 sits in a siphon lens at 334.9 m below datum and steps down to a salinity of 6.73 ppt with the salinized interval reading 7.63 degC at dawn and the wellhead freshened 9.64 m3/d while the electrode pair returns 1681 uS/cm.\nNode SN-019-0122 sits in a spooned lens at 1Line truncated
@@ -1,319 +0,0 @@
-# XDNA2 NPU inference runtimes on Linux (Strix Halo) — can any of them serve TTS?
-
-Research slice, 2026-10-19. Worker: `NpuRuntimeLandscape`. Research only, no code.
-Question: every way to execute model inference on an AMD XDNA2 NPU on Linux as of
-2026-10, and whether any of them can serve a speech-synthesis model.
-
-> **Status, same night — the spike that followed this slice settled it.** NO-GO on the
-> NPU (the loader declines the graph, not the silicon); the actionable door is the gfx1151
-> iGPU behind `-DGGML_VULKAN=ON`. Measured contract, exoneration chain and host gates:
-> `tts-npu-spike-verdict-2026-10-19.md`. That page supersedes §8–§9 below.
-
-Convention: vendor claims carry URL + the version/date they hold for.
-**[host]** = Dankclaw state verified on the live machine 2026-10-19 (Main), not a
-vendor claim. `[INFERENCE]` = my deduction from the cited artifacts.
-
-## 0. Answers to the three questions Main ranked
-
-**(1) A runtime that already runs TTS on XDNA2 today: it exists at AMD, not on our
-host.** Ryzen AI Software **1.8.0** (docs `latest` == 1.8.0; released 2026-07-23)
-lists `neutts-air` and `neutts-nano` as model type **TTS**, variant
-*full-optimized* (NPU-only) — <https://ryzenai.docs.amd.com/en/latest/llm_list.html>;
-release note "New TTS model support: neutts-air, neutts-nano" —
-<https://ryzenai.docs.amd.com/en/latest/relnotes.html>. Pre-compiled artifacts are
-published: `amd/neutts-air_rai_1.8.0_npu_4K` (arch `qwen2`, lastModified
-2026-07-23, Apache-2.0, "AWQ / Group 128 / Asymmetric / BFP16 activations / UINT4
-Weights") and `amd/neutts-nano_rai_1.8.0_npu_4K` (arch `llama`).
-**Neither artifact is a TTS engine.** Both file lists are LM-only (`model.onnx`,
-`genai_config.json`, BPE tokenizer, AIE-ML compiler metastate: `dd_metastate_*`
-`.ctrlpkt/.super/.state`, `cache/txn_bins.zip`) with **no codec, no vocoder, no
-reference-audio encoder**. [INFERENCE from HF file lists]
-
-**(2) FastFlowLM: no TTS, and we cannot add one** — see §2. Halogen: no TTS entry
-**[host]**. So "something installed already serves TTS on the NPU" is FALSE.
-
-**(3) Our own ONNX graph compiled for the NPU on Linux:** possible for the plain
-ONNX flows (they compile on Linux), **impossible for a new LLM/OGA model** — model
-generation is Windows-only in 1.8.0 — and gated by an operator table where BF16
-lacks Softmax/LayerNorm/ReLU/Gelu (see §4).
-
-## 1. Landscape: every way to reach the XDNA2 array on Linux
-
-| # | Path | License | Strix Halo + Linux, 2026-10 | Speech synthesis |
-|---|------|---------|------------------------------|------------------|
-| 1 | Ryzen AI SW → ONNX Runtime + Vitis AI / Ryzen AI EP; compile INT8 (`X2`) or BF16 (`VAIML`) ONNX at session creation | AMD proprietary (tarball) | supported: our device ID/rev is in AMD's STX table; compile+run works on Linux (§3, §4) | user-authored ONNX only; no TTS recipe/demo |
-| 2 | Same EP driven by **OGA** (`onnxruntime-genai`), "full fusion" NPU-only LLM flow; runner = AMD `model_benchmark` / OGA C++ API | AMD + MS OGA | Linux: **vendor-precompiled models only**, no hybrid flow | **neutts-air / neutts-nano = TTS** (LM half) |
-| 3 | **FastFlowLM** `flm` | MIT + free proprietary kernels | Strix Halo "Supported"; Ubuntu 24.04/25.10/**26.04**, Arch | none (ASR input only) |
-| 4 | **IRON / mlir-aie** + Peano → hand-written AIE-ML design → XRT/`amdxdna` | Apache-2.0 + LLVM exception | explicitly documented "AMD Ryzen™ AI on Linux"; same driver/XRT stack | anything, hand-written; nothing ships TTS |
-| 5 | Vendor-curated AIE-ML graph sets (halogen `HALOGEN_NPU_MODELS`, FLM packs) | vendor | present **[host]** | none **[host]** |
-| 6 | Windows-only: DirectML, Ryzen AI Windows EP (`ryzenai-llm`), Studio Effects, `whispercpp` NPU | — | not usable | (per Lemonade matrix: Windows only) |
-| 7 | ggml/llama.cpp/**qwentts.cpp**, whisper.cpp, sherpa-onnx, Piper | — | no XDNA backend in ggml/ORT-mainline | CPU / iGPU only |
-| 8 | IREE, TVM, ONNX Runtime mainline EPs (OpenVINO = Intel NPU, MIGraphX/ROCm = AMD GPU) | — | no XDNA2 target; AMD's NPU EP is only "Xilinx Vitis-AI (preview)" in the ORT EP list and ships inside AMD's own build | n/a |
-
-Sources: §2–§6 below.
-
-## 2. FastFlowLM (`/usr/bin/flm` 1.0.1 **[host]**) — catalog + HTTP surface
-
-* Repo moved into AMD's ROCm org **2026-08-11** (`github.com/ROCm/FastFlowLM`,
-  v1.0.0). "Supports all Ryzen™ AI Series chips with XDNA2 NPUs (Strix, Strix Halo,
-  Kraken, and Gorgon Point)."
-  <https://github.com/ROCm/FastFlowLM> (README fetched 2026-10-19).
-* Linux support landed 2026-03; distros: Ubuntu 24.04, 25.10, **26.04**, Arch,
-  generic. Needs NPU firmware ≥ 1.1.0.0, kernel **7.0+** with `amdxdna` (or
-  `amdxdna-dkms`), XRT stack (`ppa:lemonade-team/stable` → `libxrt-npu2`,
-  `amdxdna-dkms`); `flm validate` reads the DRM path (`/dev/accel/accel0`), `flm
-  run` goes through **XRT** and fails with `No such device with index '0'` unless
-  `xrt-smi examine` lists the NPU; `memlock` must be raised.
-  <https://fastflowlm.com/docs/install_lin/> (page dated 2026-03-05, fetched
-  2026-10-19). Static build bundles **XRT v2.21.75** — the same XRT series as the
-  distro package on the host **[host]**.
-* HTTP surface: `flm serve <model>` → OpenAI-compatible REST, **port 52625**, no
-  auth. Maintained endpoints: `v1/models`, `v1/chat/completions`,
-  `v1/audio/transcriptions`, `v1/embeddings`. **There is no
-  `/v1/audio/speech`.**
-  <https://fastflowlm.com/docs/instructions/server/openapi/>
-* Catalog 2026-10-19 (<https://fastflowlm.com/docs/models/>): LLaMA, DeepSeek,
-  Qwen, Gemma, MedGemma, TranslateGemma, gpt-oss, LiquidAI/LFM, Phi, Nanbeige,
-  **Whisper — card states "Type: Speech-to-Text (ASR)"**
-  (<https://fastflowlm.com/docs/models/whisper/>), EmbeddingGemma, SmolVLA,
-  Hy-MT2, MiniCPM. No speech-synthesis entry.
-* Engines are **prebuilt, closed, downloaded from Hugging Face per architecture**
-  ("Internet access to HuggingFace is required to download the optimized model
-  kernels"). The only in-tree engine build is Gemma-4E
-  (`cmake --preset linux-default -DFLM_BUILD_GEMMA4E=ON`).
-  → **A new TTS architecture cannot be added to `flm` by us.**
-
-## 3. AMD Ryzen AI Software 1.8.0 on Linux — install source, device support
-
-* **Install channel (important correction to the pypi.amd.com evidence).** The pip
-  index is Windows-only for the EP runtime (`ryzenai_dynamic_dispatch-1.7.0-…
-  win_amd64.whl`, `onnxruntime_vitisai-1.23.2-…win_amd64.whl`; `flexmlrt` has a
-  single **linux_aarch64** wheel) — but AMD's Linux stack is not on pip. It ships
-  as a tarball that is, as of 2026-10-19, **publicly reachable without an AMD
-  account**: `https://download.amd.com/opendownload/RyzenAI/1.8.0/ryzen_ai-1.8.0.tgz`
-  → HTTP 206 on a range probe, `Content-Range: bytes 0-8/**7819506855**`
-  (**≈7.8 GiB**, magic `1f 8b 08` = gzip). (The docs link it behind the
-  EULA form `account.amd.com/...ryzenai-eula-public-xef.html`; the CDN path
-  serves it directly. Verified by me with curl 2026-10-19. I did not finish
-  downloading 7.8 GiB, so the tarball's *contents* list is unverified —
-  `[INFERENCE]` only where noted.)
-* Driver/XRT package (public, 90,016,137 bytes, `application/zip`):
-  `https://download.amd.com/opendownload/RyzenAI/Driver/RAI_1.8_Linux_NPU_XRT.zip`
-  contains `xrt_202620.2.25.37_24.04-amd64-{base,base-dev,npu}.deb` +
-  `xrt_plugin.2.25.260102.56.release_24.04-amd64-amdxdna.deb` (verified by
-  download + unpack). **The plugin deb ships firmware for
-  `/usr/lib/firmware/amdnpu/17f0_11/npu.dev.sbin`** — exactly the directory our
-  NPU loads **[host]** — plus `1502_00`, `17f0_10`, `17f1_10`, `17f1_13`,
-  `libvxdna.so`, `libxrt_driver_xdna.so.2.25.260102.56`, and the DKMS driver
-  source tarball. **No xclbin/PDI and no compiler binary in this package**: it is
-  driver + XRT shim + firmware only.
-* **Our device is officially STX.** AMD's APU table
-  (<https://ryzenai.docs.amd.com/en/latest/app_development.md#apu-types>):
-  `0x1022:0x17F0 rev 0x11 → STX` (0x00/0x10 also STX, 0x20 = KRK, `0x1502` rev 0
-  = PHX/HPT). Release notes: "Strix and **Strix Halo** as STX", and the STX/KRK
-  column enables CNN INT8, CNN BF16, NLP BF16, LLM (OGA).
-* Vendor driver source (same deb, `share/amdxdna/amdxdna.tar.gz`, source dated
-  2026-06-29) maps `0x17f0 rev 0x11 → dev_npu5_info`, `.fw_path =
-  "amdnpu/17f0_11/"`, `default_vbnv = "RyzenAI-npu5"`, and the revision→name table
-  has `{ AIE2_DEV_REVISION_HALO, "NPU Strix Halo" }`; `xrt-smi examine` therefore
-  reports the device as **NPU Strix Halo**.
-* Linux install recipe (1.8.0 docs
-  <https://ryzenai.docs.amd.com/en/latest/linux.html>): Ubuntu 24.04 packages,
-  **Python 3.12** (`apt install python3.12 python3.12-venv`),
-  `libboost-filesystem1.74.0`, dkms; then `install_ryzen_ai.sh -a yes -p <venv>`;
-  `LD_LIBRARY_PATH=…:${RYZEN_AI_INSTALLATION_PATH}/onnxruntime/lib/`;
-  `quicktest.py` "picks up a simple CNN model, **compiles it** and runs on AMD's
-  Neural Processing Unit" — i.e. **compilation runs on Linux** for the non-LLM
-  flows. "Linux currently supports NPU only flow" (no iGPU EP).
-  Host friction: our box is Ubuntu 26.04 + **Python 3.14** **[host]**, so the venv
-  must be built with a 3.12 interpreter and the packages are 24.04-built
-  `[INFERENCE]`.
-* LLM flow on Linux
-  (<https://ryzenai.docs.amd.com/en/latest/llm_linux.html>): prequantized
-  ready-to-run models from HF, `deployment/lib`,
-  `RYZENAI_EP_PATH=$PWD/deployment/lib/libonnxruntime_providers_ryzenai.so`,
-  run via the `model_benchmark` binary; **"Linux does not support Hybrid flow"**;
-  **"Model generation is not supported in current release"** (repeated in
-  relnotes 1.8 Known Issues: "Models generated on Windows are compatible with
-  Linux"); needs `memlock unlimited`.
-  Sample throughput in that doc: **49.1 tokens/s decode**, 753 tokens/s prefill
-  (Phi-3.5-mini, 4K model).
-* Third-party confirmation of the OS split: Lemonade's backend matrix
-  (<https://raw.githubusercontent.com/lemonade-sdk/lemonade/main/README.md>,
-  fetched 2026-10-19) — `flm`/npu/XDNA2 = **Windows, Linux**;
-  `ryzenai-llm`/npu = **Windows only**; `whispercpp`/npu (STT) = **Windows only**;
-  text-to-speech engines `kokoro` and `acestep` = metal / cuda / vulkan / rocm,
-  **no npu device at all**. Lemonade is the "Server Interface" layer of AMD's own
-  Ryzen AI stack (docs index), so this matrix is the vendor-sanctioned view.
-
-## 4. Can we compile our own graph on Linux? (the no-recipe path)
-
-Mechanics from <https://ryzenai.docs.amd.com/en/latest/modelrun.md> (1.8.0):
-ONNX only, opset 17 recommended; the EP compiles on first `InferenceSession` and
-caches (VitisAI EP cache, or ORT **EP context cache** — `ep.context_enable`, and
-"precompile the EP context model using Python and deploy it using a C++ program"
-is explicitly supported); EP auto-partitions unsupported nodes to CPU;
-BF16 = `target VAIML` (`preferred_data_storage: unvectorized` for transformers),
-BF16 **STX/KRK-or-newer only**; INT8 = `target=X2` with XINT8/A8W8/A16W8 and
-"general asymmetric quantization … to enable third-party quantized models";
-**deployment EP cannot compile BF16 on the fly** — ship a pre-compiled model
-(<https://ryzenai.docs.amd.com/en/latest/app_development.md>);
-cache dirs are not portable across EP or NPU-driver versions.
-
-Operator coverage is the hard gate
-(<https://ryzenai.docs.amd.com/en/latest/ops_support.md>, "operators supported in
-Ryzen AI 1.8.0"; "Y" = broad coverage, some configurations may still fail):
-
-* Good for TTS front-ends/vocoders: `ConvTranspose` **Y in BF16/A16W8/A8W8**
-  (absent only from XINT8), `Conv`, `Resize`, `Upsample`, `Pad`, `LSTM` (BF16),
-  `CumSum` (BF16/XINT8), `STFT` (XINT8 only).
-* **Blockers for a decoder-style transformer in the plain flows:** in **BF16** the
-  table has **no Softmax, no LayerNormalization, no Relu/Gelu**; LayerNorm is
-  XINT8-only and Softmax is A16W8/A8W8/XINT8-only. So attention either lives in
-  the **LLM/OGA flow** (whose op set is exactly
-  `SimplifiedLayerNormalization`, `SkipSimplifiedLayerNormalization`, `MatMulNBits`
-  (W4ABF16/W4ABFP16), `RotaryEmbedding`, **`GroupQueryAttention`**, `Sigmoid`,
-  `Mul`, `Add`) or must be decomposed into a quantized INT8 flow.
-* Missing entirely: `GRU`, `Trilu`, `OneHot`, `EyeLike`, `Multinomial`,
-  `NonMaxSuppression`, `CastLike`, `MelWeightMatrix`/`MelFilterGroup`,
-  `RMSNormalization` → graphs that emit them must be rewritten before they
-  partition onto the NPU. [INFERENCE from absence in the table.]
-* Quantization is a separate tool, not in the driver package: **AMD Quark**
-  (`quark.torch` / `quarkonnx`), which documents `ConvTranspose2d`, `conv1d`,
-  softmax/LayerNorm/BFP quantization (relnotes 1.3/1.2/1.1/0.9) — Quark is *not*
-  on pypi.org under the obvious names **[host probe]** and ships from AMD's
-  gated/pip-private channel.
-* **New LLM-shaped model (our Qwen-Talker analogue): Windows-only today.**
-  OGA model generation is not available on Linux in 1.8.0 (§3). So a *new*
-  speech-LM cannot be self-served onto the NPU on Linux without AMD (or a
-  Windows box) in the loop.
-* Practical shape cost: the neutts artifacts carry **one compiled variant per
-  padded sequence length** (128/256/512/768/1024/1536/2048/2304/3072/4096 for
-  prefill; 256…4096 for token-phase attention mask) — i.e. per-shape compile
-  artifacts are the norm for the LLM flow. [INFERENCE from the HF file list.]
-
-## 5. Does the AMD TTS artifact actually produce speech? (cloned-voice survival)
-
-NeuTTS is **LM + separate neural codec** ("Simple LM + codec architecture"),
-clone from ~3 s of reference audio, and the codec is **NeuCodec (50 Hz, single
-codebook)**: `neuphonic/neucodec` (PyTorch, encode+decode),
-`neucodec-onnx-decoder{,-int8}` (**decode only**), "Decoder-only (ONNX) codecs
-require **pre-encoded references** — these ship with the repo for all bundled
-voices." Phoneme input; 2048-token context ≈ 30 s of audio; NeuTTS-Air ~360 M
-active / 552 M with embeddings; upstream benchmarks are **CPU-only llama.cpp**
-(119 tok/s prefill/decode mix on a Ryzen 9 HX 370) and explicitly exclude the
-codec. <https://raw.githubusercontent.com/neuphonic/neutts/main/README.md>
-(fetched 2026-10-19); `neuphonic/neutts-air` HF metadata (lastModified
-2026-08-26, `qwen2`, GGUF + safetensors).
-
-Consequences for the fleet spike:
-
-1. AMD's NPU artifact = LM only; **waveform decode and reference encoding are CPU
-   work either way**. Our `reference.wav` voices survive only if we run the
-   NeuCodec **encoder** (PyTorch, per the codec table there is no ONNX encoder)
-   once per voice and then feed pre-encoded codes. That is a one-time offline
-   step per voice; after that the compiled graph takes codes/prompt prefix as
-   ordinary inputs, so nothing about the compiled graph forbids a caller-supplied
-   speaker prefix. `[INFERENCE]`
-2. Real-time math: 50 Hz codec means **≥50 tokens/s decode** for 1× RT. AMD's own
-   Linux sample output for a 4K Phi-3.5-mini is **49.1 tok/s** — so budget for
-   roughly RT-1× on one 4K "full fusion" graph, before the codec's CPU cost.
-   `[INFERENCE — different model; must be measured.]`
-3. Host RAM: the same doc reports **peak working set 3.54 GB** for that 4K model.
-   NeuTTS-Air is ~552 M params at W4A-BFP16, so expect a smaller footprint than
-   3.5 GB `[INFERENCE]`; still non-trivial against 14 GiB available **[host]**.
-4. AMD ships **no TTS demo, runner, or server** on any OS: the examples page has
-   only CNN/BF16/DistilBERT/super-resolution/OGA-LLM/VLM/RAG
-   (<https://ryzenai.docs.amd.com/en/latest/examples.md>); the RyzenAI-SW repo
-   tree (669 paths, `main`, fetched 2026-10-19) contains speech demos **only under
-   `Demos/ASR/{Whisper,Parakeet-TDT}` and `Transformer-examples/ASR/Whisper-AI`**
-   — no TTS directory. Serving TTS therefore means writing our own OGA host
-   (or driving `model_benchmark`-style C++ code) plus phonemizer + NeuCodec glue.
-
-## 6. Two graphs on one NPU, and the iGPU hazard
-
-* **Concurrency is a first-class driver feature.** The vendor `amdxdna` source
-  (in `xrt_plugin.2.25.260102.56`, 2026-06-29) sets `.hwctx_limit = 16` for
-  npu4/npu5 — i.e. **16 hardware contexts** on our device — and allocates each
-  context a set of AIE columns with QoS `opc = col_opc (4096) × num_col`
-  (`aie2_ctx.c`, `aie2_solver.h`), with `frame_boundary_preempt` /
-  `fill_preempt_slot` in `aie2_pci.h` (firmware preemption at frame boundaries).
-  So halogen's armed models and a TTS graph can hold separate contexts
-  simultaneously; they contend for columns and OPC share, they are not mutually
-  exclusive.
-* AMD's user-visible history agrees: 1.4 "Support for up to **16 hardware
-  contexts** on Ryzen AI 300 series"; 1.5 "Application concurrency: improves the
-  resource distribution across applications"; 1.3 "Idle detection and suspension
-  of contexts", "Rebalance feature for AIE hardware resource optimization";
-  1.0 known issue "Support for multiple applications is limited to up to eight"
-  (all in <https://ryzenai.docs.amd.com/en/latest/relnotes.html>).
-* **Package conflict, the real risk** [INFERENCE]: the vendor stack wants AMD's
-  XRT 2.25.37 + `xrt_plugin…amdxdna` deb **[host: distro `libxrt-npu2
-  1:2.21.75+dfsg-4` installed]**, and firmware must match the loaded `amdxdna.ko`
-  — the vendor README warns that a stale/mismatched `npu.sbin` causes
-  `ERT_CMD_ABORT`/mailbox timeouts, and that a distro **in-tree** `amdxdna` older
-  than the XRT SHIM makes `GET_ARRAY`/`QUERY_TELEMETRY` return
-  `EOPNOTSUPP`/`EINVAL` (<https://github.com/amd/xdna-driver> README,
-  fetched 2026-10-19). Installing AMD's debs over the distro XRT changes the ABI
-  under FastFlowLM/halogen, so the spike must decide the XRT version **once**, for
-  the whole host.
-* **iGPU / fabric clock:** no vendor document supports a fabric-clock pin, and no
-  vendor document addresses NPU+iGPU concurrency hangs. AMD's Strix Halo
-  optimization doc (ROCm docs, dated 2026-02-20, and its successor "AMD RDNA3.5
-  system optimization", ROCm 10.1.0) covers GPUVM/GART/GTT, the TTM `pages_limit`
-  knob, `amd-ttm`, BIOS VRAM sizing, and minimum kernel versions
-  (<https://rocm.docs.amd.com/en/docs-7.2.0/how-to/system-optimization/strixhalo.html>,
-  <https://rocm.docs.amd.com/en/latest/reference/system-optimization/rdna3-5.html>).
-  The `pp_dpm_fclk` level-7 pin and the GPU+NPU hang observation stay
-  fleet-observed **[host]**, i.e. `[INFERENCE]` territory — no citation exists.
-
-## 7. Paths I checked and ruled out (so nobody re-walks them)
-
-* **Windows-only doors**: DirectML; Ryzen AI Windows EP; Lemonade's
-  `ryzenai-llm` and `whispercpp`-NPU engines (matrix above); AMD's OGA model
-  generation. Not usable on Dankclaw.
-* **Mainline ONNX Runtime** has no AMD-NPU EP: its EP summary lists AMD ROCm and
-  AMD MIGraphX (GPU) and "Xilinx Vitis-AI (*preview*)" only
-  (<https://onnxruntime.ai/docs/execution-providers/>, 2026-10-19); the working
-  EP is AMD's own build (`libonnxruntime_providers_ryzenai.so` in the tarball's
-  deployment dir, per llm_linux).
-* **ggml / llama.cpp / qwentts.cpp**: no XDNA backend → the current engine
-  (`tts-server` on GGUF, CPU-only build **[host]**) cannot be "turned on" for the
-  NPU; migration means a different runtime, not a flag.
-* **`flm`**: catalog/engines are closed per-architecture (§2) — the cheapest door
-  by install cost, and it is not a door for TTS.
-* **IRON / mlir-aie** is the only open toolchain that documents running on the
-  Ryzen AI NPU from Linux ("A close-to-metal Python API for programming AMD Ryzen™
-  AI NPUs, built on an open-source MLIR-based compiler toolchain"; "Getting
-  Started for AMD Ryzen™ AI on Linux", Ubuntu 24.04/24.10, XRT from
-  `ppa:amd-team/xrt`: `libxrt2 libxrt-npu2 … amdxdna-dkms`, `render` group
-  required) — <https://raw.githubusercontent.com/Xilinx/mlir-aie/main/README.md>,
-  fetched 2026-10-19. This is how FLM's kernels were made ("Low-level kernels
-  optimized using IRON + AIE-MLIR", FLM README) — i.e. real, but it is
-  hand-written per-op tile programming, not a model compiler: writing a TTS model
-  here is a multi-month engineering project, not a migration.
-* **Embedded** Ryzen AI P100 support in the same Linux doc is irrelevant to STX.
-
-## 8. Recommendation implied by the above (for Main's spike)
-
-Spike the AMD path as decided, with these constraints planned in: Python 3.12 venv
-+ 24.04-built wheels on Ubuntu 26.04; one host-wide XRT decision (AMD 2.25.37 vs
-distro 2.21.75) with firmware/driver pairing checked via `xrt-smi examine` →
-"NPU Strix Halo"; 7.8 GiB tarball staged off the public CDN URL (§3); pre-encode
-each `reference.wav` with the **PyTorch** NeuCodec encoder because the ONNX codec
-is decoder-only; keep phonemizer + codec on CPU; measure decode tokens/s against
-the 50 Hz budget and peak RSS against the ~14 GiB headroom; expect to write the
-serving shell yourself (AMD ships `model_benchmark`, no TTS demo).
-
-## 9. Loose ends (not verified; I stopped on Main's stand-down)
-
-* Contents of `ryzen_ai-1.8.0.tgz` (7.8 GiB) — whether the Linux tarball contains
-  the AIE-ML compiler (`vaiml`, `aiecompiler_client`) for the *BF16/INT8* flows
-  and which Python wheels it carries. My statement that non-LLM compilation works
-  on Linux rests on AMD's own Linux doc ("quicktest.py … compiles it and runs on
-  … NPU"), not on the package contents. Download was killed mid-flight.
-* Whether AMD's `neutts` artifacts run correctly under the Linux `model_benchmark`
-  runner (the TTS rows are listed without OS qualifiers; no Linux TTS run is
-  documented anywhere in 1.8.0 docs).
-* Exact AIE column count/total `ncols` exposed for Strix Halo (driver reads it
-  from AIE metadata: `ndev->total_col = min(XRS_MAX_COL, ndev->aie.metadata.cols)`).
@@ -1,438 +0,0 @@
-# Qwen3-TTS payload + qwentts.cpp runtime: architecture and portability
-
-Research deliverable (no code). Slice: *what the fleet's TTS payload actually is, and how
-portable it is off the CPU ggml path.* Companion slice: `NpuRuntimeLandscape` (what NPU
-runtimes exist on Dankclaw).
-
-Evidence date: 2026-10-19. Everything below is grounded in:
-
- `github.com/ServeurpersoCom/qwentts.cpp` @ `5916bf7c79dbb4e66fc059cdc507a79a23a28bb4`
-  (2026-10-16, upstream HEAD on 2026-10-19) and its ggml submodule
-  `ServeurpersoCom/ggml` @ `cc5df49a6b878db2b68b9a7b45976e279594627d` (2026-10-16).
- The two GGUF files as they sit on Dankclaw (`/home/dank/models/Qwen3-TTS-GGUF/`), read
-  byte-for-byte from the live host.
- Upstream Qwen3-TTS model cards / release material.
-
-Status: COMPLETE for the assigned slice. ANSWER C was confirmed on the live host the same
-night — the running `frank-tts-engine` container carries `libggml-base.so.0.27.0` +
-`libggml-cpu.so.0.27.0` and nothing else, and our own
-`dankclaw/voiceservice/Containerfile.engine` writes `-DGGML_VULKAN=OFF` out explicitly, so
-the iGPU change is a flag in a file the fleet owns. Verdict, measured contract and the NPU
-exoneration chain: `tts-npu-spike-verdict-2026-10-19.md`.
-
---
-
-## 1. What the payload is
-
-Two files, both read byte-for-byte off the live host (`/home/dank/models/Qwen3-TTS-GGUF/`,
-owned `dank:dank`, mtimes 2026-10-18 18:31), both from
-`https://huggingface.co/Serveurperso/Qwen3-TTS-GGUF` (community converter, not Qwen):
-
-| file | bytes | GGUF | tensors | arch string | file_type |
-|---|---|---|---|---|---|
-| `qwen-talker-1.7b-base-Q4_K_M.gguf` | 1,219,245,248 | v3, kv=64 | 480 | `qwen3-tts` | `Q4_K_M` |
-| `qwen-tokenizer-12hz-Q8_0.gguf` | 291,150,624 | v3, kv=51 | 398 | `qwen3-tts-tokenizer` | `Q8_0` |
-
-`general.name` of the talker file is `Qwen3-TTS-12Hz-1.7B-base`; the header sizes add up
-to the file sizes (sum of tensor bytes + header + alignment padding), so nothing in these
-files is unaccounted for.
-
-### 1.1 Talker file (480 tensors)
-
-Three models in one container, per `docs/ARCHITECTURE.md` §GGUF layout
-(qwentts.cpp @ pinned commit) and confirmed by the tensor census:
-
-| prefix | tensors | bytes | what it is |
-|---|---|---|---|
-| `talker.*` | 316 | 1,061.9 MB | 28-layer Qwen3 decoder (hidden 2048, FFN 6144, 16 Q heads / 8 KV heads, head_dim 128, RoPE θ=1e6, RMS ε=1e-6, ctx 32768), text stream + codec stream summed, `codec_head` → 3072 codebook-0 logits |
-| `code_pred.*` | 88 | 114.8 MB | 5-layer Qwen3 MTP head (hidden 1024, FFN 3072, 15 acoustic codebooks × 2048), own KV cache |
-| `spk_enc.*` | 76 | 36.6 MB | ECAPA-TDNN speaker encoder (x-vector, base checkpoints only; mel → conv0 → 3 SE-Res2Net → MFA → ASP → 2048-d vector) |
-
-Dtype split (exact tensor byte sums): `q4_K` 250 tensors / 864.1 MB, `q6_K` 17 / 312.0 MB,
-`f32` 176 / 25.8 MB, `f16` 37 / 11.4 MB.
-
-Which tensors carry which quant is the load-bearing portability fact:
-
- `q6_K` (17): `talker.text_embd.weight [2048, 151936]` (255.3 MB), `talker.codec_embd.weight
-  [2048, 3072]`, and the 15 `code_pred.codec_embd.<g>.weight [2048, 2048]` embedding tables.
-  These are the llama.cpp `Q4_K_M` "embedding/output rows go to Q6_K" rule, applied to
-  embedding tables rather than to an `lm_head`.
- `q4_K` (250): every attention/FFN projection of the talker and of the code predictor.
-  Distinct matmul shapes: `[2048,2048]`×73, `[2048,1024]`×62, `[2048,6144]`×56,
-  `[6144,2048]`×28, `[1024,2048]`×20, `[1024,3072]`×10, `[1024,1024]`×10, `[3072,1024]`×5,
-  `[2048,3072]`×2, `[2048,151936]`×1. Every one is 2-D and ≥ 1024 on both axes — good
-  GEMM shapes, nothing exotic.
- `f32` (176) / `f16` (37): all of `spk_enc` (conv weights F16 because K=1/3/5 rows are not
-  block-alignable, per the converter's `should_quantize` policy in `tools/quantize.cpp` +
-  `quantize.sh`; every bias/norm/LayerScale vector F32), plus the talker/predictor norms and
-  the `text_proj` biases.
-
-Metadata the runtime actually reads (talker file): `qwen3-tts.num_code_groups=16`,
-`talker.vocab_size=3072`, `text_vocab_size=151936`, `text_hidden_size=2048`,
-`position_id_per_seconds=13`, `rope.mrope_section=[24,20,20]`, `mrope_interleaved=1`,
-the 7 codec special ids (pad 2148 / bos 2149 / eos 2150 / think 2154 / nothink 2155 /
-think_bos 2156 / think_eos 2157), `language_names` (10: chinese, english, german, italian,
-portuguese, spanish, japanese, korean, french, russian) with their codec ids, the text-side
-special ids (im_start 151644, im_end 151645, tts_pad 151671, tts_bos 151672, tts_eos 151673),
-the whole `generation_config.json` (do_sample, top_k 50, top_p 1.0, temperature 0.9,
-repetition_penalty 1.05, `subtalker_*` mirrors, max_new_tokens 8192), and an embedded
-**Qwen2 BPE tokenizer** (`tokenizer.ggml.model=gpt2`, 151,676 tokens, 151,291 merges,
-eos 151643) — that BPE block is 5.95 MB of the header, i.e. 49 % of the header is vocabulary.
-
-Two live-file details that contradict the upstream doc text (worth knowing when reading
-`docs/ARCHITECTURE.md`: the doc example block is stale against the shipped file):
-`mrope_interleaved` is **1** (doc shows `false`), and `spk_enc.sample_rate` is **24000**
-(doc says 16000). Doc also says "404 tensors on the 1.7B voice_design build" — the base
-build shipped here has 480 because it carries `spk_enc.*`.
-
-### 1.2 Tokenizer/codec file (398 tensors)
-
-One neural audio codec, 12.5 frames/s over 24 kHz mono, hop 1920, 16 codebooks × 2048
-entries (1 semantic + 15 acoustic). Dtype split: `f16` 50 / 143.6 MB, `q8_0` 110 / 76.9 MB,
-`f32` 238 / 70.7 MB. `tok_dec.*` 253 tensors / 201.3 MB, `tok_enc.*` 145 / 89.9 MB.
-
-Composition (metadata + tensor census, both files agree):
-
- **Encoder** (used only to make voice references, not on the synthesis hot path): SEANet
-  conv stack (init k=7, 4 stages with ratios 8/6/5/4 → cumulative 960×, 64→512 ch, final
-  k=3), an 8-layer transformer (hidden 512, 8 heads, head_dim 64, FFN 2048, RoPE 1e4,
-  LayerNorm+bias, GELU, LayerScale), a downsample conv k=4 s=2 (25 → 12.5 Hz), then the RVQ
-  argmin loop. Metadata records `encoder.num_quantizers=32` with
-  `encoder_valid_num_quantizers=16` — 16 codebooks are used, 16 are dead weight carried from
-  the upstream config.
- **Decoder** (the hot path: every generated frame): 16 codebook lookups + per-split
-  `output_proj` (256→512), `conv_pre`, an 8-layer transformer (hidden 512, 16/16 heads,
-  head_dim 64, FFN 1024, **sliding window 72**, RMSNorm, SwiGLU, LayerScale, input_proj
-  1024→512 / output_proj 512→1024), a ConvNeXt upampler (`upsampling_ratios=[2,2]`, causal
-  transposed convs + depthwise k=7), then a DAC-style decoder `dec.0..3`
-  (`upsample_rates=[8,5,4,3]`, channels 1536→768→384→192→96, `decoder_dim=1536`) whose blocks
-  are SnakeBeta → causal transposed conv (k = 2·stride) → 3 residual units (dilations 1/3/9)
-  → snake_post → conv_post k=7 → 96→1.
- All 32 codebook tensors are **F32, 67.1 MB total** (16 encode + 16 decode, `[256, 2048]`),
-  deliberately unquantized: the converter's `should_quantize` keeps RVQ codebooks and their
-  wrapping projections at F32 because nearest-neighbour lookup is sensitive to per-row
-  quantisation noise (`docs/ARCHITECTURE.md` "Quantisation policy"; mirrored in
-  `quantize.sh`). 49 of the codec's 55 3-D conv tensors are **F16** with non-alignable K
-  (7, 3, 2, 1, 16, 10, …) — they cannot be block-quantized at all.
- **No flow matching, no diffusion, no vocoder-with-ODE.** It is a pure RVQ codec
-  (Mimi/SoundStream/Mimi-style: SEANet + transformer + DAC decoder). That matters for the
-  NPU question: every component is convolution, attention, or embedding lookup.
-
-### 1.3 Where the payload came from, and what the official weights look like
-
-Official upstream (`huggingface.co/api/models`, retrieved 2026-10-19):
-
-| repo | last modified | license | weights |
-|---|---|---|---|
-| `Qwen/Qwen3-TTS-12Hz-1.7B-Base` | 2026-01-23 | apache-2.0 | `model.safetensors` BF16, 1,928,677,440 params, plus an embedded `speech_tokenizer/model.safetensors` |
-| `Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice` | 2026-01-29 | apache-2.0 | same layout, named-speaker tables |
-| `Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign` | 2026-01-29 | apache-2.0 | 1.7B only |
-| `Qwen/Qwen3-TTS-12Hz-0.6B-Base` / `-0.6B-CustomVoice` | 2026-01-29 | apache-2.0 | narrow variant (hidden 1024) |
-| `Qwen/Qwen3-TTS-Tokenizer-12Hz` | 2026-01-29 | apache-2.0 | codec alone, F32, 170,557,441 params |
-
-So the GGUF is **not** the only public form: the safetensors are the canonical form
-(Apache-2.0, no separate community license risk), and the GGUF is a lossy re-quantization of
-them produced by the same author who wrote the runtime (`Serveurperso`), via
-`checkpoints.sh` → `convert.py` (F32 GGUF) → `quantize.sh` (BF16/Q8_0/Q4_K_M).
-Sizes to remember for any re-export: the talker container is 1.93 B params in BF16
-(that count includes the embedded codec), the codec alone is 170.6 M params in F32.
-
-**Portability consequence #1 — the GGUF is a dead end off-ggml.** Q4_K/Q6_K are ggml
-block formats (256-element super-blocks, F16 sub-block scales, 144 B / 210 B per block). No
-ONNX graph, no Vitis-AI/AIE-ML compiler, and no vendor NPU runtime consumes them; AMD's own
-quantizer works from ONNX (`ryzenai-quantize`, QDQ/QOP power-of-two — Ryzen AI Software
-1.8.0 release notes, https://ryzenai.docs.amd.com/en/latest/relnotes.html). Moving the
-payload to a non-ggml stack therefore means starting from the safetensors again, not
-converting the GGUF.
-
-**Portability consequence #2 — the quantization split is semantic, not cosmetic.** The
-converter keeps RVQ codebooks, `input_proj`/`output_proj` around them, and every 1-D and
-non-alignable-conv tensor out of the k-quants precisely because quantizing them breaks voice
-fidelity. Any re-quantization path (GPTQ/AWQ/QDQ-INT8/INT4 for an NPU compiler) inherits that
-constraint: the talker backbone is requantizable, the codec's codebook lookups and SnakeBeta
-convs are not, at least not without re-measuring voice quality.
-
-## 2. What the runtime is
-
-`qwentts.cpp` @ `5916bf7c` (2026-10-16, MIT; upstream model Apache-2.0) is a C++17 ggml
-application — 43 files under `src/` + `tools/`, three binaries (`qwen-tts`, `qwen-codec`,
-`tts-server`) over one public plain-C99 ABI (`src/qwen.h`, `qt_*`, `QT_ABI_VERSION`
-range-checked; `tests/abi-c.c` is compiled `-std=c99 -Wall -Werror -pedantic` on every
-build). It is not a llama.cpp derivative: it links ggml directly and builds its own graphs.
-
-Five graph families, and their shapes are the whole portability story:
-
-| graph | batch/time axis | notes |
-|---|---|---|
-| talker prefill | T = prompt length (hundreds–thousands) | summed text+codec embedding streams, 28-layer GQA decoder, KV cache |
-| talker decode | **T = 1 per audio frame** | 12.5 frames/s → 80 ms of audio per step; `codec_head` → 3072 logits |
-| code predictor (MTP) | **T = 1, 16 steps per frame** (2 prefill + 14 decode, cache reset every frame) | 5-layer decoder, 15 heads × 2048 |
-| speaker encoder | one pass per voice registration | ECAPA-TDNN with mel fused into the graph |
-| codec decode | offline: T = 300-frame chunks; streaming: **T = 1 frame** with persistent conv/KV-ring state | the streaming graph is built once with fixed topology so the CUDA graph cache stays in pure replay |
-
-Per generated frame the host therefore runs ~1 + 16 + 1 tiny graph computes for 80 ms of
-audio, and `tts-server` serializes synthesis FIFO (`--max-batch` defaults to 1, and the
-fleet runs it at 1). This is the archetypal latency-bound, bandwidth-bound,
-batch-1 workload; the only big-M matmuls in the whole system are the prefill and the
-chunked (buffered) codec decode.
-
-Op contract actually issued to ggml (`src/`, counted): `mul_mat` 77, `cont` 70, `add` 57,
-`mul` 40, `rms_norm` 22, `clamp` 15, `rope_ext` 12, `get_rows` 12, `set_rows` 8,
-`scale_bias` 6, `swiglu_split` 5, `sqr` 5, `mean` 5, `elu` 4, `col2im_1d` 4, `soft_max` 3,
-`conv_1d` 3, `im2col` 2, `cumsum` 2, `sin`, `sigmoid`, `tanh`, `div`, `argsort`, `transpose`,
-`permute`, `reshape/view`, plus the CPU-side Philox multinomial sampler.
-
-**No custom ggml op is required today.** The repo doc still says the fork "provides the two
-custom ops the codec needs: `GGML_OP_SNAKE` and `GGML_OP_COL2IM_1D`"; that is stale at the
-pinned commit. Verified 2026-10-19:
-
- the submodule `ServeurpersoCom/ggml` @ `cc5df49` is upstream `ggml-org/ggml` master
-  (`a05c992`, 2026-10-15) **+2 commits, 0 behind** (GitHub compare API), and
-  `include/ggml.h` is byte-identical to upstream master (`diff` empty);
- the 2 commits are `193e339` (2026-06-19) "cpu: add fused snake activation
-  (mul -> sin -> sqr -> mul -> add)" — a **CPU subgraph fusion**, not an op (the runtime
-  builds snake with `ggml_sin`/`ggml_sqr`/`ggml_mul`/`ggml_add` in `src/dac-decoder-v2.h:222`),
-  and `cc5df49` "sycl: fix get_rows on reordered quantized weights";
- `ggml_col2im_1d` is an **upstream ggml API** (same declaration, line 2097, in both trees);
-  the Vulkan backend fuses the snake chain through ggml's generic subgraph fuser
-  (`ggml-vulkan.cpp:14548 ggml_vk_can_fuse_snake`, `vulkan-shaders/snake.comp`).
-
-Consequence: any stock ggml build (upstream, or llama.cpp's vendored ggml) can execute this
-payload; only the CPU snake fusion is fork-specific, and it is a performance nicety.
-
-Backend/device support: every tool takes `--device <name>` (`CUDA0`, `Vulkan0`, `MTL0`,
-`CPU…`, default best available), resolved through the ggml backend registry; the deployed
-container turns CUDA/HIP/Vulkan/Metal/SYCL/RPC/BLAS all OFF and compiles the native-CPU
-backend with OpenMP, with CPU threads = `hardware_concurrency()/2` (`src/backend.h`).
-The registry in this ggml knows CPU, CUDA, Metal, SYCL, Vulkan, WebGPU, ZDNN, OpenCL,
-**Hexagon**, BLAS, RPC, VirtGPU, **CANN**, ZenDNN, **OpenVINO**, ET — and nothing AMD-NPU
-(no `xdna`/`aie`/`vitis`/`ryzen` string anywhere in the ggml tree).
-
-Voice handling (what the fleet depends on): the registry is **process RAM only**; a restart
-empties it, which is why the unit has a boot warmer. `POST /v1/audio/voices` accepts either
-`{name, ref_text, wav_b64}` (server extracts) or `{name, ref_text, spk_b64, rvq_b64}`
-(pre-extracted latents verbatim); `ref_text` present ⇒ ICL clone mode, absent ⇒ x-vector-only
-mode. `.spk` = the ECAPA x-vector; `.rvq` = 16 codes/frame packed 11 bits LSB-first
-(`src/rvq-file.h`). `wav` and `pcm` `response_format` are different audio at the same seed
-(buffered chunked decode vs stateful streaming decode) — the fleet already encodes that rule.
-
-Parity expectations for any port: the upstream harness matches the PyTorch reference per
-stage (embed cos 0.999988, all 28 hidden taps ≥ 0.999893, logits 0.999824) but explicitly
-reports free-run divergence (`CodesFull exact 4.96 %`, `Audio cos 0.095`) and explains why —
-one argmax tie at FP epsilon diverts the AR trajectory. So a migrated endpoint must be
-accepted by listening + per-stage cosine, never by waveform diff against today's audio.
-
---
-
-## 3. The three answers Main asked for
-
-### ANSWER A — Qwen3-TTS GGUF → ONNX: no, and it does not matter
-
-There is **no GGUF→ONNX path** and there will not be a useful one: GGUF is a weight
-container for ggml, its Q4_K/Q6_K block layouts have no ONNX consumer, and nothing in the
-GGUF carries the *graph* (the talker/predictor/codec structure lives in C++ code, only the
-op-level structure exists in ggml at runtime). But the GGUF is not the payload's only form —
-official Apache-2.0 safetensors are upstream, and **four independent ports of the full
-pipeline already exist**:
-
-| port | form | what it proves |
-|---|---|---|
-| `xkos/Qwen3-TTS-12Hz-1.7B-ONNX` (HF, updated 2026-03-03, apache-2.0) | FP16+FP32 ONNX, fp16/voice_clone + fp16/voice_design splits: `talker_decode.onnx`, `code_predictor.onnx`, `code_predictor_kv.onnx`, `code_predictor_embed_g0..15.onnx`, `text_embedding.onnx`, `codec_embedding.onnx`, shared `speaker_encoder.onnx`, `speech_tokenizer_encoder.onnx`, `speech_tokenizer_decoder.onnx`, `model_cache.npz` | every component of this architecture exports to ONNX, codec decoder included |
-| `zhangluoyang/qwen3-tts-onnxruntime` | export tooling + Python inference for Base/CustomVoice/VoiceDesign | the export is reproducible from the official checkpoints |
-| `Echo9Zulu/Qwen3-TTS-12Hz-CustomVoice-1.7B-INT8-OpenVINO`, `aurora2035/Qwen3-TTS-12Hz-0.6B-Base-OpenVINO-INT8` (2026-03) | OpenVINO IR, **INT8**: `talker`, `code_predictor`, `codec_embedding`, `text_model`, `openvino_speaker_encoder_model`, `speech_tokenizer/speech_decoder` | the talker and predictor INT8-quantize for an NPU-targeting stack |
-| `litert-community/Qwen3-TTS-12Hz-0.6B-Base` (2026-10-17, apache-2.0) | `qwen3_tts_int8.litertlm` / `_int4.litertlm` (LiteRT-LM `OmniEngine`), legacy split graphs `talker_int4.tflite` (prefill_32/prefill_128/decode signatures, KV 1024, blockwise-32 OCTAV INT4), `mtp_folded_int8.tflite` (16 inner steps × 5 layers folded into one graph, in-graph argmax+gather, GPTQ int8, ~5× faster), `codec_partA` (transformer, fp32) + `codec_partB` (conv upsampler, XNNPACK FORCE_FP16, ~2.5×, "ASR-identical") | an NPU/mobile compiler accepts the whole thing — at the price of graph splitting, static prefill buckets, and folding the per-frame inner loop |
-
-What the ports had to concede, and what any port inherits:
-
- the generation structure does not fit existing engine loops: LiteRT-LM's Engine "does not
-  support this generation structure yet", so their first shape was **three graphs + a
-  host-side loop**, and the MTP step was "invoked 17× per frame" until they folded it;
- per-step dispatch cost dominates: even on a Snapdragon 8 Elite Gen 5 they keep the MTP
-  loop, text projection, embedding gathers and the codec on CPU "avoiding 16 synchronous
-  CPU↔GPU round-trips per 80 ms frame", putting only the 28-layer talker on the GPU;
- RTF on that hardware is **2.57–2.97 (Galaxy S26) and 5.65–8.48 (Pixel 10 Pro)** for the
-  0.6 B model, i.e. 2.6–8× slower than real time. Dankclaw's current CPU engine measures
-  **warm RTF 0.55–0.67 with TTFA 94–112 ms** on the 1.7 B model (fleet trial, 2026-10-18).
-  An NPU port is not "faster by default" for this workload; it has to beat that.
- quantize the LM (int8/int4: proven), keep the codec's codebook lookups and SnakeBeta convs
-  in fp16/fp32 (the LiteRT split does exactly this).
-
-So: the answer to "can the fleet's payload be exported" is *yes — from the safetensors, not
-from the GGUF*, and the work is a pipeline rewrite (multi-graph host loop, per-frame fold,
-static prefill buckets), not a conversion. That is the same conclusion NeuTTS-Air's
-precompiled-graph route encodes: on AMD's Linux stack the graph is vendor-curated, so the
-fleet chooses the model that AMD already compiled rather than compiling Qwen3-TTS.
-
-### ANSWER B — is there any ggml backend for an AMD NPU? Not upstream; three out-of-tree, none for TTS
-
-Checked against the trees today (2026-10-19), because this was the crux:
-
- **Nothing in-tree.** `ggml-org/llama.cpp` master (4,237 paths, GitHub tree API) contains no
-  `xdna`/`aie`/`vitis`/`ryzen` file or directory; the ggml backend set is CPU, BLAS, CUDA,
-  HIP, Metal, Vulkan, SYCL, OpenCL, WebGPU, MUSA, RPC, CANN (Ascend), **Hexagon** (Qualcomm),
-  OpenVINO (Intel CPU/GPU/NPU), ZenDNN, ZDNN, VirtGPU, ET. The deployed fork's registry
-  matches. The feature request "Feature Request: XDNA backend" (llama.cpp issue **#21725**,
-  opened 2026-04-10) is still **open**, 34 comments; the older #1499 was closed stale
-  2025-04-07. So Main's premise is confirmed: no ggml XDNA backend exists to flip a flag on.
- **`Cyronius/ggml-xdna`** — a real ggml backend (`XDNA0`) loaded into *stock* llama.cpp via
-  `GGML_BACKEND_PATH`, no fork. It does bfp16 (8-value blocks sharing an exponent) GEMMs from
-  GGUF weights, and is **prefill-only**: it takes only matmuls in prompt chunks ≥ 1024 tokens
-  (`GGML_XDNA_MIN_BATCH` default 1024, `GGML_XDNA_MIN_MFLOP` 256), attention and every decode
-  token stay on the Vulkan device; pinned to llama.cpp release b10944; tested on Strix Point
-  (HX 370) only — Strix Halo "should pass, but nobody has run it on one yet"; Linux support
-  new and untested on hardware by the author; needs `libxrt-npu2`, `/dev/accel/accel0`,
-  unlimited `memlock`. Best measured prefill gain 1.88× on a 27 B model; models under
-  2,048-wide **lose** (0.77× at width 1,536).
- **`BrandedTamarasu-glitch/OllamaAMDNPU`** — a llama.cpp fork with `-DGGML_XDNA=ON`,
-  explicitly targeting **Ryzen AI MAX / XDNA2 on Linux** (`/dev/accel/accel0`, XRT 2.21.75+,
-  amdxdna, unlimited memlock). It implements **only `GGML_OP_MUL_MAT`**, dispatched to
-  mlir-aie-compiled `.xclbin`s built per fixed tile shape (e.g. `M=2048 K=2048 N=64`,
-  int8 in / int32 out, per-row weight scales cached at first use), with a second slot for a
-  different K. Its own numbers on a Ryzen AI MAX 385 with Llama-3.1-8B Q4_K_M: NPU prefill
-  12–19 t/s vs **Vulkan iGPU 600–930 t/s**, NPU decode **0.65–1.40 t/s** ("Decode does not
-  use the NPU … M=1 per token, no xclbin covers it"), 0.6 % AIE utilization, and the stated
-  conclusion "**For LLaMA 8B on Strix Halo: use Vulkan exclusively**". Single-maintainer,
-  AI-assisted, self-measured — treat as a proof of concept, but it is the closest published
-  picture of what XDNA2 graph execution costs.
- Also referenced in the same thread: `Atomic-Germ/OpenFlowLM-Next` (whole-model NPU engine,
-  LLM-only — and slower than the iGPU in the same comparison), `GaTmaNnes/xdna2-`,
-  `hardWorker254/Npu-Embeddings-XDNA1` (adds Whisper ASR ops, NPU1/Linux). None is a TTS path.
-
-Why none of them would carry this workload even if installed: the TTS hot loop is **batch-1
-by construction** (one talker step + 16 predictor steps + one codec frame per 80 ms), which
-sits under every dispatch gate those backends define (≥1024 tokens / ≥32768 M·N·K / tile M
-covering only prefill), and the codec half of the pipeline is conv, col2im, gather and
-elementwise ops — not `MUL_MAT` — so it would fall back to CPU op-by-op. The NPU case for
-this host is therefore *not* "run Qwen3-TTS on the NPU"; it is "run a vendor-compiled AR TTS
-graph on the NPU", which is exactly the NeuTTS-Air spike.
-
-### ANSWER C — would `qwentts.cpp` build for the gfx1151 iGPU? Yes, via **Vulkan**, not HIP
-
- Upstream ships the build scripts and claims the backends: `buildvulkan.sh` ("AMD/Intel GPU
-  (Vulkan)"), `buildcuda.sh`, `buildsycl.sh`, `buildcpu.sh`, `buildall.sh` (all backends,
-  runtime dynamic loading), and README line 6 "runs on CPU, CUDA, Metal, Vulkan". CMake knobs
-  `GGML_VULKAN` / `GGML_HIP` / `GGML_SYCL` are the same ones the fleet's Containerfile turns
-  OFF, so the engine rebuild is a flag change: `-DGGML_VULKAN=ON` plus
-  `--device Vulkan0` on the same command line the unit runs today.
- The codec's exotic ops are covered by the Vulkan backend, which is the thing that decides
-  whether a GPU build is coherent: `vulkan-shaders/col2im_1d.comp`, `snake.comp`,
-  `swiglu*.comp`, `cumsum*.comp` exist and `supports_op` handles `GGML_OP_COL2IM_1D` /
-  `CONV_TRANSPOSE_1D` / `CUMSUM`. By contrast `COL2IM_1D` appears in **no** HIP, OpenCL,
-  WebGPU, Hexagon or CANN source — a **HIP** (ROCm) build would shard the ConvNeXt/DAC
-  transposed convs back to the CPU per op. On an AMD iGPU the portable path is RADV/Vulkan.
- `--device` also accepts a multi-device split (`-dev`-style naming is ggml's), and the
-  runtime's documented FP16 guard (`--clamp-fp16`) exists for sub-Ampere CUDA accumulation;
-  nothing in the tree is NVIDIA-specific.
- Unverified by me: any vendor statement pairing this fork with gfx1151. Supporting evidence
-  that the hardware/driver combination is viable: this very host already runs a Vulkan
-  endpoint on the gfx1151 iGPU (fleet ground truth, 2026-10-19), and Mesa's Vulkan
-  implementation is officially supported on Radeon Software for Linux
-  (AMD release notes RN-AMDGPU-UNIFIED-LINUX-25.20.3). A gfx1151 Vulkan build of this fork is
-  a build-and-measure spike, not an unsupported port.
- The real constraint is the one already in the fleet docs: the iGPU belongs to another
-  endpoint, and GPU+NPU concurrency has hung Strix Halo machines. An iGPU TTS build competes
-  with the text endpoints for the same silicon; the NPU route's only genuine advantage is
-  that it does not (the XDNA backend authors make the same argument: "Use NPU when the iGPU
-  is busy … no GPU contention").
-
---
-
-## 4. What any replacement must reproduce, and what the voices cost
-
-Endpoint contract in production today (from the live unit/env, not the doc):
-`tts-server --model qwen-talker-1.7b-base-Q4_K_M.gguf --codec qwen-tokenizer-12hz-Q8_0.gguf
--alias frank-tts-engine --host 0.0.0.0 --port 8880 --lang auto --max-batch 1`, no published
-port (the façade on :7772 dials the private bridge name), `GET /health`, `GET /v1/models`,
-`GET|POST /v1/audio/voices`, `DELETE /v1/audio/voices/{name}`, `POST /v1/audio/speech` with
-`response_format` wav|pcm (different audio at equal seed) and optional `seed`,
-`max_new_tokens`, `temperature`, `top_k`, `top_p`, `repetition_penalty`, `subtalker_*`.
-Voice registry is RAM-only, so boot warming is mandatory, and `--alias` keeps the quantization
-choice out of every consumer's config.
-
-Live voice census on Dankclaw (2026-10-19, `/home/dank/voices/`): three packs
-(`narrator1..3`), each carrying `reference.wav` (24 kHz mono crop), `reference16k.wav`,
-`manifest.json` (keys: name, kind, generated_utc, pipeline, source, forbidden_region_guard,
-silence_scan, crop, loudnorm, audio, acceptance, acceptance_passed) and ffmpeg analysis
-artifacts. **No `.spk` and no `.rvq` exist on the live host** — the repo's
-`build-voice-pack.sh` treats them as required-but-they-are-not-there, so today the engine
-re-extracts the x-vector (and the ICL codes when a transcript is supplied) from the WAV at
-registration time.
-
-That is the portable part, and it is the same shape AMD's NeuTTS flow wants
-(`encode_reference(reference.wav)` + a reference transcript): the fleet's voice identity is
-*the WAV plus its transcript*, processed at request/registration time by the model's own
-speaker encoder + codec — **never a trained per-voice artifact**. Qwen3-TTS cloning is
-inference-time conditioning (`generate_voice_clone(ref_audio=…, ref_text=…, x_vector_only_mode=…)`;
-`create_voice_clone_prompt` exists purely to avoid recomputing prompt features). So any
-replacement that consumes a WAV (+transcript) keeps all three narrators alive with no
-re-derivation, and if it wants precomputed latents instead, the LiteRT precedent shows what
-that looks like: a 4 KB x-vector `.npy` per voice, extracted offline
-(`extract_speaker_embedding.py`) — for the 1.7 B base checkpoint that vector is 2048-dim
-(`qwen3-tts.spk_enc.embedding_length = 2048` in the shipped GGUF; the LiteRT 0.6 B bundle
-ships a 1024-dim one).
-
-One trap for a mixed-runtime future: `.spk`/`.rvq` files, if they are ever written, are
-engine-specific encodings (ECAPA x-vector layout; 11-bit-packed RVQ stream) and are **not**
-interchangeable between runtimes. Keep `reference.wav` as the pack's source of truth and
-regenerate anything derived per engine — which is what the NeuTTS re-clone plan does anyway.
-
-## 5. Sources
-
-Runtime/payload (all read locally at pinned versions, 2026-10-19):
-
- `github.com/ServeurpersoCom/qwentts.cpp` @ `5916bf7c79dbb4e66fc059cdc507a79a23a28bb4`
-  (commit date 2026-10-16; also upstream HEAD today): `README.md`, `docs/ARCHITECTURE.md`
-  (860 lines), `src/*`, `tools/*`, `docs/DOCKER.md`, `buildvulkan.sh`/`buildcpu.sh`/`buildall.sh`.
- submodule `ServeurpersoCom/ggml` @ `cc5df49a6b878db2b68b9a7b45976e279594627d` (2026-10-16);
-  delta vs `ggml-org/ggml` master `a05c992` (2026-10-15) = ahead 2 / behind 0, per
-  https://github.com/ServeurpersoCom/ggml/compare/ggml-org:master...HEAD ;
-  `include/ggml.h` identical to upstream
-  https://raw.githubusercontent.com/ggml-org/ggml/master/include/ggml.h .
- Live payload bytes: `/home/dank/models/Qwen3-TTS-GGUF/qwen-talker-1.7b-base-Q4_K_M.gguf`
-  (1,219,245,248 B) and `qwen-tokenizer-12hz-Q8_0.gguf` (291,150,624 B), headers parsed from
-  the host; sizes/sha256 match `scripts/download-frank-tts-engine-qwentts.sh`.
- Live units/env/packs: `dankclaw/voiceservice/{Containerfile.engine,tts-engine-start.sh,tts-engine-warm.sh,quadlet/*}`,
-  `scripts/frank-tts-engine.env`, `scripts/frank-tts.env`, `/home/dank/voices/*`.
-
-Model provenance:
-
- `https://huggingface.co/api/models/Qwen/Qwen3-TTS-12Hz-1.7B-Base` (+ CustomVoice,
-  VoiceDesign, 0.6B variants, `Qwen3-TTS-Tokenizer-12Hz`), apache-2.0, last-modified
-  2026-01-23 / 2026-01-29; safetensors param counts 1,928,677,440 (BF16, incl. embedded
-  `speech_tokenizer/`) and 170,557,441 (F32 codec).
- `https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-Base/raw/main/README.md` (clone API:
-  `generate_voice_clone(ref_audio, ref_text, x_vector_only_mode)`,
-  `create_voice_clone_prompt`; reference stack = PyTorch + `device_map="cuda:0"` + bf16 +
-  flash-attn-2, Python 3.12 recommended).
- `https://github.com/QwenLM/Qwen3-TTS`.
-
-Portability precedents:
-
- `https://huggingface.co/xkos/Qwen3-TTS-12Hz-1.7B-ONNX` (2026-03-03) ·
-  `https://github.com/zhangluoyang/qwen3-tts-onnxruntime` ·
-  `https://huggingface.co/Echo9Zulu/Qwen3-TTS-12Hz-CustomVoice-1.7B-INT8-OpenVINO` (2026-03-22) ·
-  `https://huggingface.co/aurora2035/Qwen3-TTS-12Hz-0.6B-Base-OpenVINO-INT8` (2026-03-18) ·
-  `https://huggingface.co/litert-community/Qwen3-TTS-12Hz-0.6B-Base` (2026-10-17) incl.
-  `legacy/README.md` (graph splits, MTP fold, codec partA/partB, RTF table).
-
-AMD NPU / ggml backend status:
-
- `https://api.github.com/repos/ggml-org/llama.cpp/git/trees/master?recursive=1` (no
-  XDNA/AIE/Vitis backend, 2026-10-19) ·
-  `https://github.com/ggml-org/llama.cpp/issues/21725` (open since 2026-04-10) ·
-  `https://github.com/ggml-org/llama.cpp/issues/1499` (closed stale 2025-04-07).
- `https://github.com/Cyronius/ggml-xdna` (prefill-only, gates, Linux requirements, results
-  measured 2026-10-05/06 on HX 370) ·
-  `https://github.com/BrandedTamarasu-glitch/OllamaAMDNPU` (XDNA2 ggml backend on Strix Halo,
-  phases 1–12, "use Vulkan exclusively", measured on Ryzen AI MAX 385).
- `https://github.com/amd/RyzenAI-SW` (main, 669 paths, 2026-08-18): audio content is ASR
-  only (`Demos/ASR/Whisper`, `Demos/ASR/Parakeet-TDT`, `Transformer-examples/ASR/Whisper-AI`);
-  **no TTS example anywhere in the official Ryzen AI Software tree**.
- `https://ryzenai.docs.amd.com/en/latest/relnotes.html` (Ryzen AI SW 1.8.0: ONNX-format
-  models, ONNX Runtime as the loader, power-of-two QDQ/QOP quantization).
- Mesa/RADV on AMD iGPUs: `https://www.amd.com/en/resources/support-articles/release-notes/RN-AMDGPU-UNIFIED-LINUX-25-20.3.html`
-  (Mesa Vulkan driver officially supported).
-
-Status: COMPLETE for the assigned slice (payload architecture, runtime architecture,
-portability). `[INFERENCE]` marks: none above except the Mesa-version caveat in ANSWER C,
-which is flagged there.
@@ -1,191 +0,0 @@
-# TTS on the Strix Halo NPU — spike verdict (2026-10-19)
-
-Question: **how can the fleet's TTS engine run on hardware rather than CPU?** Answer in
-one line: not on the NPU — the loader, not the silicon, refuses it; the door that exists is
-the gfx1151 iGPU behind a build flag we already own.
-
-Spike evidence lives in `report/npu-tts-runtime-landscape-2026-10-19.md` (which runtimes can
-reach XDNA2 at all) and `report/qwen3-tts-payload-portability-2026-10-19.md` (what the
-payload is and where it can be moved). This page owns only the **measured spike result** and
-the **model-side contract** neither slice has. Throwaway work ran in containers under
-`/home/dank/spikes/npu-tts` on Dankclaw; nothing in the fleet was changed — `frank-tts` and
-`frank-tts-engine` still serve `narrator1|2|3` on CPU.
-
-Legend: **[host]** live census 2026-10-19 · **[spike]** measured in the spike containers ·
-**[vendor]** primary vendor source, with the version it holds for.
-
-## 1. Verdict
-
-**NO-GO on the NPU, and the wall is a loader.** The XDNA2 device is healthy by AMD's own
-instruments; the runtime that owns this model family opens the device and then declines the
-graph, silently. One run loaded successfully; byte-identical runs never reproduced it, and a
-path that cannot be reproduced is not a foundation. **[spike]**
-
-**Go on the iGPU, and the lever is in this repo.**
-`dankclaw/voiceservice/Containerfile.engine` builds `qwentts.cpp` (commit `5916bf7`) with
-`-DGGML_VULKAN=OFF` written out deliberately beside CUDA/HIP/METAL/SYCL/RPC — the running
-container carries `libggml-base.so.0.27.0` + `libggml-cpu.so.0.27.0` + `tts-server` and
-nothing else. So: `-DGGML_VULKAN=ON` + `--device Vulkan0`, plus `/dev/dri` passthrough in
-the engine quadlet (the ComfyUI pattern: `AddDevice` + `GroupAdd=keep-groups` + a
-world-writable udev rule). The target shape is **co-tenancy** — the engine resident on the
-gfx1151 *beside* `frank-dragoon`, not a handover. §5 carries the census and both gates:
-the **heap gate is PASSED** — RADV's `DEVICE_LOCAL` heap is 77.33 GiB with 65.16 GiB of
-budget while dragoon is loaded, so memory is not the constraint — which leaves the
-contention gate (dragoon's TTFT with a synthesis in flight) and one unproven piece:
-whether qwentts.cpp's codec ops have Vulkan kernels for every op. **[host]** **[spike]**
-
-Why the iGPU is worth doing at all, given CPU already works: the deployed payload needs
-only **12.5 codes/s** (§2), so this is latency headroom and freed cores, not reviving a
-dead service. The independent Strix Halo measurement in the portability slice ("use Vulkan
-exclusively"; NPU decode under 1.5 t/s for an 8B) points the same way.
-
-## 2. The two arms do not share a real-time bar
-
-|Arm|Talker|Codec|Codes per second of audio|Bar|
-|---|---|---|---|---|
-|**Deployed** (`frank-tts-engine`)|`qwen-talker-1.7b-base-Q4_K_M.gguf`|`qwen-tokenizer-12hz-Q8_0.gguf`|**12.5**|~13 tok/s **[host]**|
-|**Spike** (`amd/neutts-air_rai_1.8.0_npu_4K`)|NuTTS-Air, 24-layer Qwen2|NeuCodec `24000 / hop 480`|**50.00**|~50 tok/s **[spike]**|
-
-Measured from the staged NPU artifact — the contract any future arm must satisfy:
-
-|Fact|Value|
-|---|---|
-Vocab|`genai_config.json` `vocab_size` **217,652**; base BPE 151,643; specials 151,644–151,670 **[spike]**|
-Speech tokens|**65,536** `<|speech_N|>` entries, ids **151,671–217,206** contiguous; the vendor template maps a codec index straight to its id, `"".join(f"<|speech_{i}|>")`. A 10,001-code assumption silently truncates the space — encoded references really do reach the 60k range. **[spike]**|
-Stop token|`<|SPEECH_GENERATION_END|>` = 151,670, but the export's `model.eos_token_id` is `[151645, 151643]` (the *chat turn* stop) and `search.eos_token_ids` is unset — pass 151,670 explicitly or speech ends after one token **[spike]**|
-Input ceiling|4096 tokens **total**, input + output **[vendor]**|
-Frame rate|50.00 codes/s confirmed across three references (49.96 / 49.99 / 50.00) **[spike]**|
-Reference encode|`librosa.load(sr=16000, mono=True)` → `[1,1,T]` float → `codec.encode_code(audio_or_path=…)` (vendor `NeuTTS.encode_reference`); CPU RTF 0.28–0.38 on 8 threads, so codes are a cacheable one-time cost, never a per-request charge. `neucodec` also needs `torchtune`; install both `--no-deps --target <dir>` so a second torch never shadows the venv's. **[spike]**|
-Pre-encoded voices|`/spike/logs/codes-narrator{1,2,3}.npy` — 994 / 1,249 / 1,250 codes for 19.90 / 24.98 / 25.00 s **[spike]**|
-
-**Gap that any arm inherits:** the voice packs carry no transcript. `manifest.json` holds
-provenance and crop geometry (`crop.source_speech_run` 98.739–118.636 s of a 22,569 s
-audiobook) and no text, and Dankclaw has no ASR at all — `whisper`, `whisper.cpp`,
-`faster-whisper` all absent. NuTTS conditioning needs `ref_text`; do not invent it. **[host]**
-
-## 3. Exoneration chain — why the blame is not the hardware
-
-Every row is an instrument, not an inference.
-
-|Probe|Result|
-|---|---|
-`xrt-smi examine`|`[0000:c6:00.1] RyzenAI-npu5 aie2p 6x8`, XRT 2.25.37 (vendor extract), NPU fw 1.1.2.65, `amdxdna 7.0.0-38-generic` **[spike]**|
-`xrt-smi validate`|**PASSED**, including `throughput` **[spike]**|
-`pyxrt.device(0)` in-container|constructs — XRT enumerates the device **[spike]**|
-`strace -f -e trace=openat` on the failing load|`genai_config.json` read → `model.onnx` opened → `libxrt_driver_xdna.so.2` loaded → `/dev/accel/accel0` opened `O_RDWR` → GEM ioctls succeed → **EP declines, before touching `dd_metastate_*` or `cache/txn_bins.zip`** **[spike]**|
-`dmesg`, `fuser /dev/accel/accel0`|silent since boot, no holder — the decline never reaches the kernel **[spike]**|
-Hardlinked mirror dirs of the model tree (`rundir*`)|retired, no change — the "metastate shadowing" theory is dead **[spike]**|
-Every logging lever in that build: `GENAI_ORT_VERBOSE_LOGGING`, `GENAI_LOG_ORT_LIB`, `set_log_options`, `set_log_callback(callable)`, `RYZENAI_EP_TRACING`, `RYZENAI_LLM_PROFILE_INIT`|**zero lines** — the decline path is silent **[spike]**|
-Vendor `quicktest`|`Session successfully initialized` — proves the XRT/driver/device path, **not** NPU residency: that build falls back to CPU silently. Prove residency with `/sys/class/accel/accel0/device/mem_usage` during generation. **[spike] [vendor]**|
-
-No NPU token rate was ever measured, so none is quoted here. The single successful load
-(13:17, `[load] 0.84s`, finite logits, XRT banner at *generate* time) has not been
-reproduced by the byte-identical command since, with the model files pristine by inode and
-mtime. **[spike]**
-
-## 4. Host gates the spike proved (so they are not re-litigated)
-
- **Rootless containers cannot read `/dev/accel/accel0`** (`root:render` 0660): the host
-  gid never enters the userns map. `--group-add keep-groups`, `--userns=keep-id`,
-  `--userns=keep-id:gid=990`, `--group-add render`, `--group-add gid:990` — all fail.
-  Doors: rootful podman (separate storage), or a world-writable udev rule, which the fleet
-  already does for `kfd`/`renderD*`. **[spike]**
- **Never `dpkg -i` the vendor XRT deb** — it installs a DKMS `amdxdna` over the distro
-  driver. `dpkg-deb -x` the debs and bind-mount the tree at `/opt/xilinx/xrt`; the plugin
-  deb is what carries `libxrt_driver_xdna.so`. **[spike]**
- In-tree `amdxdna 0.7.0-38` + `iommu=pt` is enough for context creation; a missing device
-  node is *not* the modern failure mode. **[host]**
- `xrt-smi` inside a container needs `libboost-filesystem1.83.0` +
-  `libboost-program-options1.83.0` (**hyphenated** Ubuntu names, after `apt update`);
-  boost is needed by `xrt-smi` only, never by the OGA/EP path. **[spike]**
- Headroom is tight and it is host RAM, not a separate pool — see §5 for tonight's census.
-  A multi-GB NPU partition plus an AIE compile workspace does not fit beside
-  `frank-dragoon`. Benign-looking `Failed to alloc dev bo memory, ret -28` is that
-  pressure. **[host]**
-
-## 5. Target shape: the TTS engine shares the gfx1151 with `frank-dragoon`
-
-The deployment Frank wants is **both resident at once**, not a handover. Measured 2026-10-19
-(read-only, nothing loaded or unloaded):
-
-|Measure|Value|Reading|
-|---|---|---|
-iGPU GTT in use|**18,523 MiB of 116,733 MiB**|≈96 GiB of aperture free; this is the real budget (fleet law) **[host]**|
-DRM "VRAM" carve-out|2,048 MiB total, 174 MiB used|what KFD/DRM calls dedicated — **not** the budget **[host]**|
-Host RAM|121 GiB total, 17 GiB available, 79 GiB buff/cache|**GTT pages are system pages**: an iGPU allocation is host RAM consumed, so the two tenants share one pool **[host]**|
-TTS engine today|3.84 GiB RSS, CPU|its device-resident replacement lands in that same pool **[host]**|
-
-**Two gates, in this order.**
-
-1. **Heap gate — PASSED.** RADV does **not** hand a container tenant the 2 GiB DRM
-   carve-out. `vulkaninfo` inside the engine's own base image (`ubuntu:24.04` plus
-   `mesa-vulkan-drivers vulkan-tools`, `--device /dev/dri`, rootless) sees
-   `Radeon 8060S Graphics (RADV GFX1151)`, driver `radv`, API 1.4.318, and reports:
-
-   |Heap|size|budget with dragoon loaded|flags|
-   |---|---|---|---|
-   |0|38.67 GiB|32.58 GiB|host-visible, not device-local|
-   |1|**77.33 GiB**|**65.16 GiB**|`DEVICE_LOCAL`|
-
-   The heaps sum to the ~116 GiB aperture in Mesa's ⅓/⅔ APU split — the same RAM twice:
-   a Vulkan allocation is a GTT mapping, so a budget is a ceiling on mappings, not a pile
-   of free silicon memory. ggml's Vulkan backend asks for `DEVICE_LOCAL`, which is heap 1,
-   and the TTS engine needs roughly 4 GiB of it. **Memory stops being the question.** **[spike]**
-   Probe traps: `vulkaninfo --batch` is not a flag in this `vulkan-tools` (it prints the
-   help page and exits 1 — use `vulkaninfo --text`), and the container has **no NVIDIA
-   ICD**, so Vulkan device 0 is the iGPU here. Do not trust that ordering — select by
-   device *name*, or an image that grows an `nvidia_icd.json` silently points TTS at the
-   3090 slot.
-2. **Contention gate — PASSED, measured 2026-10-19 after the cutover.** With a 16.9 s
-   synthesis occupying the device, `frank-dragoon` answered a streamed 124-token request at
-   **17 ms to first token and ~35 tok/s decode**, against 17 ms and ~33 tok/s for the same
-   request with the card idle. Compute-unit sharing showed up as nothing. One request per
-   arm, no ladder: these are the gate's evidence, not a throughput record, and the engine's
-   own shape lives in [[Frank TTS Engine]].
-
-**Change surface — as landed** (PR #34, `6248f02`; the CPU shape stays reachable as the
-rollback):
-
-1. `Containerfile.engine` build stage: add **`libvulkan-dev glslc spirv-headers`**. The
-   prediction above (`libvulkan-dev glslang-dev`) is wrong on two counts and the configure
-   says so: ggml looks for `glslc`, which Ubuntu's `glslang-tools` does **not** ship (it
-   ships `glslangValidator`), and `find_package(SPIRV-Headers CONFIG REQUIRED)` needs the
-   distro's `spirv-headers`, which does carry the CMake config. `glslc` is its own source
-   package on noble, at 2023.8 — old enough that only `GL_KHR_cooperative_matrix` compiles,
-   so the coopmat2 / integer-dot / bf16 / micro-float shader families are built out. Suspect
-   that compiler before the device if a Vulkan prefill ever reads as slow.
-2. `-DGGML_VULKAN=OFF` → `ON`: one word in the written-out flag list.
-3. Runtime stage: `mesa-vulkan-drivers` **and** `libvulkan1` — the ICD and its loader. An
-   image with a loader and no ICD enumerates zero devices.
-4. `frank-tts-engine.container`: `AddDevice=/dev/dri` + `GroupAdd=keep-groups` — the
-   DIRECTORY, never the node: on this hybrid box `renderD128` is the NVIDIA card and
-   `renderD129` the amdgpu iGPU (`readlink -f /sys/class/drm/renderD129/device/driver` →
-   `amdgpu`). On this host the udev rule already makes the nodes world-writable, so the
-   group is belt-and-braces, not the gate.
-5. Engine argv: `--device Vulkan0`, reached through `QWENTTS_DEVICE` in
-   `scripts/frank-tts-engine.env`. **An empty value passes no flag at all** — that is the
-   CPU rollback, and it is why the entrypoint's default is an `-v` test rather than `:=`,
-   which also fires on a set-but-empty value and would silently undo the rollback.
-
-Gate 1 recipe, already run and **PASSED**: `mesa-vulkan-drivers` + `vulkan-tools` in a
-throwaway container off the same base, `--device /dev/dri`, then `vulkaninfo --text` and read
-the `VkPhysicalDeviceMemoryProperties` heap sizes — no model, no load, no compute. The probe
-trap: `vulkaninfo --batch` is not a flag in Ubuntu 24.04's `vulkan-tools` — it prints the help
-page and exits 1, which reads like a driver failure. Every package involved has a candidate on
-the host's apt (`vulkan-tools` 1.4.341, `libvulkan-dev` 1.4.341, `glslang-dev`/`glslang-tools`
-16.2.0); the build stage runs on the digest-pinned Ubuntu 24.04 base, whose versions are older
-but present. Script: `/home/dank/spikes/npu-tts/vk_probe.sh` on Dankclaw.
-
-Constraints that do not relax for a co-tenant: `/dev/dri` passthrough needs the ComfyUI
-pattern (`AddDevice` + `GroupAdd=keep-groups` + the world-writable udev rule) because
-systemd user-unit spawns carry no supplementary groups; the 64 GiB swap at `/swap.img`
-exists to protect co-tenants, and paging a KV pool is worse than the OOM it prevents;
-`pp_dpm_fclk` reads `auto` (unheld), which only becomes a hazard if the NPU ever joins.
-
-## 6. Re-visit condition
-
-Retry the NPU when **a Linux loader claims the graph**: an `onnxruntime-genai-ryzenai`
-build whose RyzenAI adapter takes the NuTTS-Air Full-Fusion graph on `RyzenAI-npu5`, or that
-model appearing in a runtime's own supported catalog with a Linux runner. Then the
-prerequisites are already mapped and priced: the gated NeuCodec encoder for cloned voices,
-the `ref_text` transcripts, the 4096-token prompt budget, and 50 codes/s as the bar.
@@ -0,0 +1,309 @@
+#!/usr/bin/env python3
+"""bench-spec-ab.py — settle "which proposer should this endpoint use?" with data.
+
+Runs one payload through several speculative-decoding configurations of the same
+binary, one at a time, with the unit stopped between runs: on a single-GPU host
+there is no room for two copies of a 29 GiB model, and the benchmarking policy
+applies. Every configuration gets identical prompts at temperature 0, and the
+greedy text is compared across them — speculation may buy speed, never change
+the answer.
+
+    bench-spec-ab.py <env-file> <serve-script> [options]
+    bench-spec-ab.py /home/dank/serverAppConfigs/scripts/frank-medium.env \
+                     /home/dank/serverAppConfigs/scripts/serve-frank-medium.sh \
+                     --configs "dflash=draft-dflash@7;ngram=ngram-mod@64;chain=draft-dflash,ngram-mod@7"
+
+Config syntax: `name=SPEC_TYPE[@n_max][|ENV=VAL|ENV=VAL]`, entries joined by `;`.
+The systemd unit is restored in a finally block, so it comes back on whatever the
+env file says even if a configuration crashes the harness.
+
+THREE SHAPES, because a single one answers the wrong question. A proposer can
+only draft text it has seen, so the shape decides the verdict:
+
+  prose       a 1k prompt asking for one sentence. Nothing to copy and nothing
+              repetitive, so only a model-backed proposer can help: this is the
+              shape that compares drafters.
+  echo        a ~1.5k passage of unique lines, "repeat it back verbatim". The
+              output IS the prompt, which is ngram-mod's home turf.
+  structured  "25 JSON records, one per line" — repetitive OUTPUT with no prompt
+              overlap, which is what an ngram matcher is really for in an
+              agentic fleet (config edits, tables, record lists).
+
+echo and structured disable thinking (`chat_template_kwargs`), otherwise the
+model thinks the answer into `reasoning_content`, `content` comes back empty and
+the cross-configuration identity check passes vacuously — which is exactly how
+this harness first "verified" an echo it never looked at.
+
+Results on Dankserver (frank-medium payload, 2026-09-16, temperature 0). Decode:
+
+  shape            dflash=7   ngram-mod(48/64)   chain        (tok/s)
+  prose 1k/512        43.8        17.7 (0 drafts)   43.7
+  echo 1.5k/400       76.8       102.5             200.6
+  structured/400      91.1        17.8 (0 drafts)   91.2
+  identity            70.4        17.8 (0 drafts)   70.4
+
+`draft-mtp,draft-dflash` fails at load (a chain has ONE draft context and MTP
+builds against whatever model that context holds); ngram-mod has no such
+constraint — it is draftless and joins the chain unconditionally.
+"""
+import argparse
+import json
+import os
+import re
+import subprocess
+import sys
+import time
+import urllib.request
+
+FILLER = ("The quick brown fox jumps over the lazy dog near the river bank. "
+          "A cache is a component that stores data so future requests are faster. ")
+TOKENS_PER_REP = 28
+UNIT = "frank-medium.service"
+MODEL = "frank-medium"
+SPEC_GREP = ("adding speculative implementation|- n_max=|- block_size=|creating MTP|"
+             "draft acceptance")
+
+DEFAULT_CONFIGS = [
+    ("none", {"SPEC_TYPE": "none"}),
+    ("mtp-n3", {"SPEC_TYPE": "draft-mtp", "SPEC_DRAFT_N_MAX": "3"}),
+    ("mtp-n7", {"SPEC_TYPE": "draft-mtp", "SPEC_DRAFT_N_MAX": "7"}),
+    ("dflash-n7", {"SPEC_TYPE": "draft-dflash", "SPEC_DRAFT_N_MAX": "7"}),
+]
+
+WORDS = ["latency", "throughput", "cache", "adapter", "scheduler", "allocator",
+         "queue", "kernel", "buffer", "shard", "replica", "window"]
+
+
+def echo_passage(lines=90):
+    """Unique lines: no long span repeats inside the prompt, so an n-gram can
+    only fire when the OUTPUT copies the INPUT."""
+    return "\n".join(
+        "Record %03d: the %s of subsystem %d reports state %d with checksum %05d."
+        % (i, WORDS[i % len(WORDS)], i % 7, i % 3, (i * 7919) % 100000)
+        for i in range(lines))
+
+
+ECHO = (echo_passage() + "\n\nRepeat the passage above verbatim, every line, exactly as "
+        "written, with no commentary before or after it.")
+STRUCTURED = ("Output exactly 25 JSON records, one per line, no prose and no code fence. "
+              'Each line must look like {"id": N, "state": "active", "region": "us-west-2"} '
+              "with N running from 1 to 25.")
+PROSE = FILLER * (1000 // TOKENS_PER_REP) + "\n\nReply with a single short sentence."
+IDENTITY = "List the first five prime numbers, comma separated, then stop."
+NO_THINK = {"chat_template_kwargs": {"enable_thinking": False}}
+
+
+def log(msg, logfile):
+    line = "[" + time.strftime("%H:%M:%S") + "] " + str(msg)
+    print(line, flush=True)
+    with open(logfile, "a") as fh:
+        fh.write(line + "\n")
+
+
+def sh(cmd):
+    return subprocess.run(cmd, shell=True, capture_output=True, text=True).stdout.strip()
+
+
+def parse_configs(spec):
+    """`name=SPEC_TYPE[@n_max][|KEY=VALUE]...`, entries joined by ';'.
+
+    KEY=VALUE parts override any env file key, which is how an impl's own window
+    (SPEC_NGRAM_N_MIN/MAX, for instance) gets swept without hand-editing env
+    files."""
+    out = []
+    for chunk in spec.split(";"):
+        chunk = chunk.strip()
+        if not chunk:
+            continue
+        name, _, rest = chunk.partition("=")
+        over = {}
+        if "|" in rest:
+            rest, _, extras = rest.partition("|")
+            for item in extras.split("|"):
+                if "=" in item:
+                    k, _, v = item.partition("=")
+                    over[k.strip()] = v.strip()
+        n_max = None
+        if "@" in rest:
+            rest, _, n_max = rest.partition("@")
+        over["SPEC_TYPE"] = rest.strip()
+        if n_max:
+            over["SPEC_DRAFT_N_MAX"] = n_max.strip()
+        out.append((name.strip(), over))
+    return out
+
+
+def write_env(src, name, overrides):
+    """The env file with this configuration's overrides applied."""
+    path = "/tmp/spec-ab-" + name.replace("/", "_") + ".env"
+    with open(src) as fh:
+        lines = fh.read().splitlines()
+    out = []
+    for line in lines:
+        for k, v in overrides.items():
+            if line.startswith(k + "="):
+                line = k + "=" + v
+                break
+        out.append(line)
+    with open(path, "w") as fh:
+        fh.write("\n".join(out) + "\n")
+    return path
+
+
+def post(base, body, timeout=1800):
+    req = urllib.request.Request(
+        base + "/v1/chat/completions", data=json.dumps(body).encode(),
+        headers={"Content-Type": "application/json"})
+    with urllib.request.urlopen(req, timeout=timeout) as r:
+        return json.load(r)
+
+
+def vram():
+    out = sh("rocm-smi --showmeminfo vram 2>/dev/null | grep 'Used Memory'")
+    return [round(int(x) / 1e9, 2) for x in re.findall(r":\s+(\d+)\s*$", out, re.M)][:2]
+
+
+def wait_health(base, timeout=420):
+    t0 = time.time()
+    while time.time() - t0 < timeout:
+        if not sh("pgrep -x llama-server >/dev/null && echo up"):
+            return False
+        try:
+            with urllib.request.urlopen(base + "/health", timeout=5) as r:
+                if r.status == 200 and json.load(r).get("status") == "ok":
+                    return True
+        except Exception:
+            time.sleep(3)
+    return False
+
+
+def stop_unit():
+    sh("sudo systemctl stop " + UNIT)
+    sh("pkill -x llama-server")
+    for _ in range(40):
+        if not sh("pgrep -x llama-server || true"):
+            return
+        time.sleep(1)
+
+
+def restore_unit(base, logfile):
+    """Bring the endpoint back on the deployed configuration. Runs in a
+    finally, because a harness that can leave production down is a hazard."""
+    log("restoring " + UNIT, logfile)
+    sh("sudo systemctl start " + UNIT)
+    for _ in range(120):
+        try:
+            with urllib.request.urlopen(base + "/health", timeout=5) as r:
+                if json.load(r).get("status") == "ok":
+                    log(UNIT + " healthy again", logfile)
+                    return True
+        except Exception:
+            time.sleep(5)
+    log("WARNING: " + UNIT + " did not come back healthy -- check it by hand", logfile)
+    return False
+
+
+def run_shape(base, prompt, max_tokens, label, extra=None):
+    body = {"model": MODEL, "messages": [{"role": "user", "content": prompt}],
+            "max_tokens": max_tokens, "temperature": 0.0}
+    if extra:
+        body.update(extra)
+    d = post(base, body)
+    tm = d["timings"]
+    msg = d["choices"][0]["message"]
+    content = (msg.get("content") or "").strip()
+    row = {"shape": label,
+           "prompt_n": tm["prompt_n"],
+           "prefill_tps": round(1000 * tm["prompt_n"] / tm["prompt_ms"], 1),
+           "predicted_n": tm["predicted_n"],
+           "decode_tps": round(tm["predicted_per_second"], 2),
+           "draft_n": tm.get("draft_n", 0),
+           "accepted": tm.get("draft_n_accepted", 0),
+           "answer": content,
+           "reasoning_len": len(msg.get("reasoning_content") or "")}
+    if label.startswith("echo"):
+        # How much of the answer really is the passage: a configuration that
+        # "wins" by answering something else is not a win.
+        want = {l.strip() for l in echo_passage().splitlines()}
+        got = [l.strip() for l in content.splitlines() if l.strip()]
+        row["copy_ratio"] = round(sum(1 for l in got if l in want) / len(got), 3) if got else 0.0
+    return row
+
+
+def measure(base):
+    run_shape(base, FILLER * (4000 // TOKENS_PER_REP) +
+              "\n\nReply with a single short sentence.", 32, "warmup")
+    return [
+        run_shape(base, PROSE, 512, "prose 1k/512"),
+        run_shape(base, ECHO, 400, "echo 1.5k/400", NO_THINK),
+        run_shape(base, STRUCTURED, 400, "structured/400", NO_THINK),
+        run_shape(base, IDENTITY, 192, "identity"),
+    ]
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("env_file")
+    ap.add_argument("serve_script")
+    ap.add_argument("--base-url", default="http://127.0.0.1:7770")
+    ap.add_argument("--log", default="/tmp/bench-spec-ab.log")
+    ap.add_argument("--configs", default=None,
+                    help="name=SPEC_TYPE[@n_max][|ENV=VAL] joined by ';' "
+                         "(default: the four configs measured on 2026-09-16)")
+    ap.add_argument("--json", default="/tmp/bench-spec-ab.json")
+    args = ap.parse_args()
+
+    configs = parse_configs(args.configs) if args.configs else DEFAULT_CONFIGS
+    # Absolute paths: the launcher must not depend on the caller's cwd, and a
+    # relative path plus a `cd` to its own directory double-counts the prefix.
+    env_file = os.path.abspath(args.env_file)
+    serve_script = os.path.abspath(args.serve_script)
+    base, logfile = args.base_url, args.log
+    results = {}
+    log("spec A/B start: " + ", ".join(n for n, _ in configs) +
+        " (unit " + UNIT + " stopped for the duration)", logfile)
+    try:
+        stop_unit()
+        for name, over in configs:
+            env_path = write_env(env_file, name, over)
+            logf = "/tmp/spec-ab-" + name.replace("/", "_") + ".log"
+            sh("FRANK_MEDIUM_ENV_FILE=" + env_path + " setsid bash " + serve_script +
+               " > " + logf + " 2>&1 < /dev/null &")
+            if not wait_health(base):
+                log(name + ": FAILED to become healthy -- " + sh("tail -3 " + logf), logfile)
+                results[name] = {"error": "not healthy", "tail": sh("tail -6 " + logf)}
+                stop_unit()
+                continue
+            rows = measure(base)
+            results[name] = {"vram_gb": vram(), "rows": rows,
+                             "spec_log": sh("grep -E '" + SPEC_GREP + "' " + logf + " | tail -4")}
+            for r in rows:
+                log(name + " | " + r["shape"] + " prefill " + str(r["prefill_tps"]) +
+                    " tok/s | decode " + str(r["decode_tps"]) + " tok/s | draft " +
+                    str(r["draft_n"]) + "/" + str(r["accepted"]) +
+                    (" (" + str(round(100.0 * r["accepted"] / r["draft_n"])) + " %)"
+                     if r["draft_n"] else "") +
+                    (" | copy " + str(r["copy_ratio"]) if "copy_ratio" in r else "") +
+                    (" | EMPTY ANSWER" if not r["answer"] else ""), logfile)
+            log(name + " | vram " + str(results[name]["vram_gb"]), logfile)
+            stop_unit()
+    finally:
+        restore_unit(base, logfile)
+
+    with open(args.json, "w") as fh:
+        json.dump(results, fh, indent=2)
+    for shape in ("prose 1k/512", "echo 1.5k/400", "structured/400", "identity"):
+        texts = [r["answer"] for v in results.values() if "rows" in v
+                 for r in v["rows"] if r["shape"] == shape]
+        if any(not t for t in texts):
+            log("WARNING: '" + shape + "' produced an empty answer in some configuration -- "
+                "its identity check proves nothing", logfile)
+        elif len(set(texts)) > 1:
+            log("WARNING: configurations disagree on greedy text for '" + shape +
+                "' -- speculation is changing answers", logfile)
+    log("done; results in " + args.json, logfile)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -0,0 +1,318 @@
+# Tunables for frank-medium (orcarouter/Qwen3.8-27B-Uncensored-GGUF Q8_0 +
+# z-lab/Qwen3.8-27B-DFlash2-GGUF Q4_K_M drafter), mainline llama.cpp on
+# Dankserver, 2x AMD Radeon AI PRO R9700 (gfx1201).  Since 2026-09-16 this is
+# the SOLE LLM runtime on the box: the vllm-radiance replica pair and its
+# routing proxy are retired.
+#
+# Hardware: two R9700 (Navi 48, RDNA4, 32,624 MiB each), no XGMI, different
+# IOMMU groups -- every cross-card byte goes over PCIe.  The Raphael iGPU
+# (gfx1036/Radeon 610M) enumerates as HIP device 2 and is hidden below so
+# nothing can drift onto shared system RAM.
+#
+# WHY LAYER SPLIT AND NOT TENSOR (operator call, 2026-09-16): tensor parallel
+# needs a cross-card allreduce per layer, and this host cannot dispatch one.
+# RCCL's kernels carry a hidden hostcall-buffer parameter, and ROCr refuses it
+# on a device without PCIe atomics -- the second card sits behind the B650
+# chipset bridges (AtomicOpsCap: Routing-, read-only), so every collective that
+# touches it dies with hipErrorIllegalState.  Measured on this box 2026-09-15
+# (llama-bench, pp4096 / tg128, f16 KV, fa on):
+#
+#   single R9700    1160.9 / 19.42
+#   layer  2x       1991.2 / 18.76      (859.1 / 17.64 at depth 32768)
+#   tensor 2x        271.8 / 26.26      (no RCCL: butterfly fallback, -7x prefill)
+#
+# Layer split distributes layers in order; only the activation at the boundary
+# crosses the wire, once per forward pass.  That is the "pipeline parallelism"
+# this service wanted, and it needs no collective at all.
+#
+# Runtime: mainline llama.cpp master fb27a525 (2026-09-16), gfx1201, built by
+# scripts/build-llama-rdna4.sh against a private ROCm 7.14.0a20260624 toolchain
+# that is bundled next to the binary (RPATH $ORIGIN/../lib, no LD_LIBRARY_PATH).
+# The ref is pinned by TWO features: DFlash2 (ggml-org/llama.cpp#27342 -- the
+# drafter is detected from its GGUF) and qwen35 MTP.  Do not float it.
+#
+# Model: orcarouter/Qwen3.8-27B-Uncensored-GGUF @ Q8_0, 27.0 GiB, arch qwen35
+# (verified from the GGUF header 2026-09-19: 262144 trained window, 24 heads /
+# 4 KV heads, nextn_predict_layers=1).  The file ALSO carries the model's MTP
+# head (blk.64.nextn.*), so --spec-type draft-mtp is available without a second
+# copy in VRAM -- one SPEC_TYPE change, the drafter below stays on disk.
+#
+# Drafter: z-lab/Qwen3.8-27B-DFlash2-GGUF @ Q4_K_M, 1.06 GiB (operator call
+# 2026-09-19: the Q4 rung -- the drafter's job is throughput, not output
+# quality, and the target verifies every drafted token).  5 layers, hidden
+# 5120, a block-diffusion proposal (block_size 8) under a 2048-token sliding
+# window (dflash.attention.sliding_window=2048, verified from the header).
+# The window is what makes it cheap HERE: its KV is bounded at 2048 positions
+# (~40 MiB) instead of spanning the context (see SIZING).
+#
+# Architecture (GGUF arch qwen35): 64 layers as 16 x (3 x (Gated DeltaNet ->
+# FFN) + 1 x (Gated Attention -> FFN)) -- 48 linear-attention layers whose
+# per-sequence state is CONSTANT-SIZE, 16 full-attention layers at 4 KV heads x
+# 256 head_dim.  A token therefore costs 2 x 4 x 256 x 2 B = 64 KiB of KV at f16
+# in the attention layers only, plus ~0.16 GiB per sequence of f32 DeltaNet
+# state at any depth.
+#
+# SIZING, the deployed shape (32:33 / ctx 327680 / 6144x512 / two slots /
+# DFlash2 Q4_K_M), measured 2026-09-19 after the model swap (rocm-smi):
+#
+#   card 0 (x16)  26.9 GiB used, 5.7 GiB free
+#   card 1 (x1)   27.8 GiB used, 4.8 GiB free
+#
+# The predecessor's measured table (rocm-smi, 2026-09-16, UD-Q8_K_XL target)
+# stays as the calibration history:
+#
+#   CTX     split  batch  slots  drafter  GPU 0 free  GPU 1 free   note (ubatch 1536)
+#   262144   1,1   2048   1      DFlash2   8.64 GiB    4.43 GiB   day's start
+#   262144  39,26  6144   2      DFlash2   2.42 GiB    7.20 GiB
+#   262144  38,27  6144   2      DFlash2   3.85 GiB    5.77 GiB
+#   262144  37,28  6144   2      DFlash2   4.28 GiB    5.34 GiB
+#   262144  36,29  6144   2      DFlash2   4.71 GiB    4.91 GiB
+#   262144  35,30  6144   2      DFlash2   5.14 GiB    4.48 GiB
+#   262144  40,25  6144   2      DSpark    1.87 GiB    1.86 GiB   (retired)
+#
+# ...and the same deployed shape re-measured at ubatch 512, after llama-bench
+# showed 512 is the throughput optimum and costs 3.71 GiB less in total (see the
+# batch block below):
+#
+#   262144  36,29  6144   512    2      DFlash2   6.01 GiB    7.32 GiB   ubatch only
+#   262144  35,30  6144   512    2      DFlash2   6.44 GiB    6.90 GiB
+#
+# ONE -ts STEP IS ONE LAYER, AND LAYERS ARE NOT THE SAME SIZE.  This is a hybrid
+# stack -- 16 x (3 Gated DeltaNet + 1 Gated Attention) blocks -- so moving a
+# DeltaNet layer shifts ~0.43 GiB while an attention layer shifts ~1.43 GiB
+# (0.45 of weights plus ~1.0 GiB of its own f16 KV at 262144).  Both step sizes
+# were measured, which is why the balance is found by measurement rather than by
+# interpolating between splits.
+#
+# WHY THE SPLIT IS 32:33 (operator call, 2026-09-19).  The operator asked for
+# the tensors BALANCED across the two cards, which for 65 layers is 32:33 --
+# the most even split that exists.  The predecessor's 35:30 skew existed to lean
+# work onto the x16 card (ROCm0), because ROCm1's Gen4 x1 chipset path made it
+# the 98 % utilization bottleneck; the measured trade of the balanced point is
+# that the x1 card runs hotter while the x16 card idles a little more, and the
+# operator accepted that for symmetry.  Do not re-skew without asking.
+#
+# THE TWO CARDS' PATHS ARE NOT SYMMETRIC (measured, 2026-09-17): ROCm0
+# (03:00.0) hangs off CPU root port 00:01.1 at Gen4 x16 end to end (~31.5 GB/s);
+# ROCm1 (0b:00.0) rides the B650 chipset through bridges that each negotiate
+# Gen4 x1 (~1.97 GB/s).  The layer-split boundary itself is not the cost
+# (~5.2 MB of activation per 512-token ubatch rides the wire); the x1 card is
+# COMPUTE-bound, not PCIe-bound.
+#
+# A FAULT TO WATCH.  A 39:26 split, at 1x batch / 1 slot, died on its eighth
+# request: the amdgpu driver logged SDMA0 UTCL2 page faults and the engine
+# surfaced `ROCm error: an illegal memory access was encountered` -> ggml_abort,
+# i.e. a fault in the cross-card boundary copy, this layer split's hot path.  It
+# has not reproduced since, but the signature is worth recognising -- it is NOT
+# an OOM, and no amount of headroom prevents it.
+#
+# WARNING: do not name any variable TEMP, HOST, or GPU_MAX_HEAP_SIZE (HIP reads
+# those bare vars; 2026-08-13 incident).  llama.cpp takes --host/--port/--temp
+# as flags only.
+
+# llama-server: mainline fb27a525 built for gfx1201 by scripts/build-llama-rdna4.sh.
+LLAMA_BIN=/home/dank/GitPackages/llama.cpp-dflash2/build-rdna4-rocm714/bin/llama-server
+
+# Payload, fetched with scripts/download-frank-medium-gguf.sh.
+MODEL_DIR=/home/dank/models/Qwen3.8-27B-Uncensored-GGUF
+MODEL_FILE=Qwen3.8-27B-Uncensored-Q8_0.gguf
+# The f16 vision projector, same repo as the target.  llama-server loads it via
+# the explicit --mmproj flag (no autoload in this build); the serve script
+# refuses to start without it when vision is enabled, so a missing file is a
+# deploy error, not a text-only server that pretends to be multimodal.
+MM_PROJ_FILE=mmproj-Qwen3.8-27B-Uncensored-f16.gguf
+DRAFT_MODEL_DIR=/home/dank/models/Qwen3.8-27B-DFlash2-GGUF
+DRAFT_MODEL_FILE=Qwen3.8-27B-DFlash2-Q4_K_M.gguf
+
+# Fleet identity.  7770 is the fleet's dankserver endpoint; the alias is the id
+# at /v1/models and what every client asks for.
+ALIAS=frank-medium
+BIND_ADDR=0.0.0.0
+LISTEN_PORT=7770
+# Device placement.  HIP_VISIBLE_DEVICES hides the Raphael iGPU (HIP device 2)
+# so nothing can drift onto shared system RAM; DEVICES is what llama.cpp splits
+# across, and neither of the two cards has an XGMI link to the other.
+HIP_VISIBLE_DEVICES=0,1
+DEVICES=ROCm0,ROCm1
+SPLIT_MODE=layer
+
+# 32:33 -- balanced (operator call 2026-09-19): 65 layers split as evenly as
+# they can be, ROCm1 (the last-layers card, which also carries the drafter and
+# the output layers) holding the extra one.  See the WHY THE SPLIT IS 32:33
+# block above for the trade against the predecessor's 35:30 x16-lean.
+TENSOR_SPLIT=32,33
+
+NGL=999
+# Auto-fit has no business resizing a context we set explicitly.
+FIT=off
+FLASH_ATTN=on
+
+# 327680 positions over TWO slots, so each conversation gets 163840 --
+# --parallel DIVIDES --ctx-size, so more slots would shrink every
+# conversation rather than grow the cache.  163840 is inside the checkpoint's
+# trained 262144 window, so the whole pool is trained territory (no rope
+# extrapolation, unlike the predecessor's 393216).
+CTX_SIZE=327680
+PARALLEL=2
+
+# MEASURED 2026-09-16 with scripts/bench-batch-ubatch.sh on this box: llama-bench,
+# a 64k prefill plus a 128-token decode with 64k of KV already resident, 3 reps,
+# one invocation per cell, the deployed split and f16 KV.  --ubatch-size is the
+# physical batch the graph is built at -- and the only one of the two knobs that
+# costs VRAM:
+#
+#   batch  ubatch   prefill tok/s        GPU 0 / GPU 1 free
+#   6144   1536    1164.96 +- 2.30      10.28 / 11.71 GiB
+#   6144    768    1135.03 +- 1.09      11.29 / 13.28 GiB
+#   6144    512    1192.93 +- 0.91      11.63 / 13.81 GiB   <- deployed
+#   6144    384    1068.04 +- 0.79      11.80 / 14.07 GiB
+#   6144    256    1045.98 +- 1.69      11.97 / 14.33 GiB
+#   4096   1536    1161.90 +- 0.96      10.28 / 11.71 GiB
+#   2048   1536    1170.23 +- 0.86      10.28 / 11.71 GiB
+#   2048    512    1193.88 +- 1.01      11.63 / 13.81 GiB
+#
+# THE PLATEAU STARTS AT 512.  512 -> 1536 buys ~2 % of prefill -- inside the
+# run-to-run spread, and the 512 cells were the faster pair in BOTH batch
+# columns -- while costing 1.35 GiB on ROCm0 and 2.10 GiB on ROCm1, so the
+# deployment takes the cheap end.  BELOW 512 the SMs are no longer saturated:
+# 384 and 256 fall ~11 % and ~12 %.  The logical batch is free either way --
+# 2048, 4096 and 6144 land within 2 % of each other at a fixed ubatch and use
+# the SAME VRAM, because the compute buffers are sized by ubatch alone.
+#
+# The prefill figure is a property of the SHAPE, not of this knob: the same
+# sweep with no depth in front of it (depth 0) runs 1543 tok/s at ubatch 256 and
+# 1723 at 1536 -- the plateau is in the same place, and the ~32 % between the two
+# tables is what attending over 64k of resident KV costs.  Decode is 15.88 tok/s
+# at depth 65536 on EVERY ubatch (17.80 at depth 0), so this knob cannot touch
+# generation at all.
+#
+# 8192/2048 stays untested under this drafter (the "2020 MiB pp buffer fails on
+# ROCm1" note was measured under DSpark, when it had 0.08 GiB free), and there is
+# no longer a reason to want it: batch buys nothing measurable here.
+BATCH_SIZE=6144
+UBATCH_SIZE=512
+
+# Threads cover tokenization, sampling and HTTP only: every layer is on a GPU.
+NUM_THREADS=8
+THREAD_BATCH=8
+
+# f16 K/V: an agentic checkpoint is the wrong place to spend KV precision.
+CACHE_TYPE_K=f16
+CACHE_TYPE_V=f16
+# Context checkpoints: 0 -- no per-slot KV snapshots at all.  The build's
+# default is 32 (this binary's --help), so the 0 below is a deliberate
+# deviation, not an inherited default: a checkpoint is a VRAM->host-RAM copy
+# of part of a slot's KV, taken so a partially-matching prompt can be restored
+# instead of re-prefilled.  This endpoint trades that away -- its reuse comes
+# from the slot prompt and from the RAM-side prompt cache below.  Stated
+# explicitly so the tunable stays visible rather than merely absent.
+CTX_CHECKPOINTS=0
+# PROMPT CACHE, HOST RAM -- this is the one cache tunable that is NOT VRAM.
+# --cache-ram caps the total bytes of saved prompt states
+# (server_prompt_cache, tools/server/server-task.cpp): update() evicts
+# oldest-first past the cap, and alloc() REFUSES outright any single state
+# bigger than the whole cap.  A state is the serialized KV of a whole sequence
+# plus its DeltaNet state plus the draft context's own KV, so at this shape it
+# costs ~68 KiB per token --
+#   16 full-attention layers x 4 kv heads x 256 dim x 2 B x 2 (K,V) = 64 KiB
+#   plus the draft context's window-bounded KV (a ~40 MiB constant for DFlash2)
+# -- which makes the cap a TOKEN budget in disguise: 16384 MiB holds ~240k
+# tokens, i.e. a FULL 163840-token slot (~10.6 GiB) with ~5 GiB of the budget
+# spare for a second entry.
+#
+# The cost is host RAM, allocated on demand and never pinned (plain heap
+# vectors), so no VRAM number above moves.  dankserver has 62 GiB beside the
+# media-stack co-tenants, so the cache is bounded rather than unlimited (-1):
+# the seam test refuses a budget over half the host's RAM.
+CACHE_RAM=16384
+
+# Sampling defaults for requests that send none -- the values the vLLM
+# predecessor served on this endpoint (`--override-generation-config`), so a
+# client sees no change across the runtime swap.  The model card's own
+# recommendation is temp 1.0 / top_p 0.95; the harness sets its own per model in
+# models.yml, so this is the floor, not the contract.
+TEMPERATURE=0.7
+TOP_P=0.95
+TOP_K=20
+MIN_P=0.0
+PRESENCE_PENALTY=0.0
+
+# Reasoning on, projected into reasoning_content (deepseek format) with no
+# budget: the model decides how long to think.
+REASONING=on
+REASONING_FORMAT=deepseek
+REASONING_BUDGET=-1
+
+# Chat template: empty = the one embedded in the GGUF (Qwen3.8's own).
+CHAT_TEMPLATE_FILE=
+CHAT_TEMPLATE_KWARGS='{"enable_thinking":true,"preserve_thinking":true}'
+
+# Speculative decoding: the DFlash2 sidecar (operator call 2026-09-19).  The
+# target file also carries its own MTP head (nextn_predict_layers=1), so
+# draft-mtp is one SPEC_TYPE change away -- the drafter file stays on disk.
+# n_max 7 is the measured DFlash2 winner (see the table below); the drafter
+# rides ROCm1, the card that holds the last layers.
+SPEC_TYPE=draft-dflash
+SPEC_DRAFT_N_MAX=7
+SPEC_DRAFT_NGL=999
+SPEC_DRAFT_DEVICE=ROCm1
+SPEC_DRAFT_TYPE_K=f16
+SPEC_DRAFT_TYPE_V=f16
+
+# DFlash2 vs the model's own MTP head -- MEASURED 2026-09-16 on the predecessor
+# payload (temperature 0, 1k prompt / 512 out; the autoregressive baseline and
+# both MTP depths were run in one session, same binary, same prompts):
+#
+#   none            17.7 tok/s                     card1 26.07 GB
+#   draft-mtp  n=3  32.4 tok/s  58 % accepted       card1 29.35 GB
+#   draft-mtp  n=7  26.7 tok/s  33 % accepted       card1 29.65 GB
+#   draft-dflash 7  43.5 tok/s  41 % accepted       card1 29.42 GB
+#
+# All four produced IDENTICAL greedy text, so verification is lossless
+# whichever proposer runs.  DFlash2 is the faster proposer (2.46x over
+# autoregressive, 34 % over the best MTP).  The MTP head adds ~3.3 GB to the
+# card that holds it, against 1.06 GB for the Q4_K_M drafter's file -- at the
+# 327680 pool the drafter's window-bounded KV (~40 MiB) leaves headroom on
+# both cards, so the speed wins.
+#
+# THE TWO CANNOT BE CHAINED. `--spec-type draft-mtp,draft-dflash` dies at load:
+#   context type MTP requested but model doesn't contain MTP layers
+#   failed to create MTP context
+# A chain has ONE draft context, and MTP builds against whatever model that
+# context holds -- here the DFlash2 file, which has no MTP layers. MTP's head is
+# only loaded when MTP IS the draft model ("creating MTP draft context against
+# the target model ..."), so a chain of the two is a structural dead end, not a
+# tuning question. See serverAppConfigs-docs/LLM/frank-medium-llamacpp-dflash2.md.
+
+# ngram-mod: the draftless n-gram matcher (CPU hash table in host RAM, no VRAM,
+# no model file, no draft context).  It joins the chain unconditionally --
+# unlike MTP, which needs a draft context and therefore cannot share one with
+# DFlash2.  NOT CURRENTLY ACTIVE: SPEC_TYPE is draft-dflash alone, and the serve
+# script only passes --spec-ngram-mod-* when the type contains "ngram".  The
+# window below is retained because the DFlash2 + ngram chain was measured free
+# on prose and 2.6x on prompt-echo -- one SPEC_TYPE change brings it back.
+#
+# WINDOW MEASURED 2026-09-16 (chain at each window, decode tok/s):
+#
+#   n_min/n_max    prose 1k/512   prompt-echo 1.5k/400
+#   48/64          43.9           200.6
+#   16/32          36.0           161.0
+#    8/16          36.2           113.8
+#
+# The long window is the right one, and not only because it wins on echo: a
+# SHORT window makes the matcher propose on short fragments of ordinary prose,
+# where the proposals are rejected and the verification still costs -- which is
+# the "ngram hurts" result the fleet measured on other payloads twice before.
+# With 48/64 the chain is byte-identical to DFlash2 alone on prose (same 119
+# drafts / 49 accepted) and on the identity prompt (70/53).
+#
+# Do not shorten these without re-running scripts/bench-spec-ab.py: the
+# no-penalty property is what makes the chain safe as a default.
+SPEC_NGRAM_N_MATCH=24
+SPEC_NGRAM_N_MIN=48
+SPEC_NGRAM_N_MAX=64
+
+# Vision enabled: llama-server loads mmproj-Qwen3.8-27B-Uncensored-f16.gguf
+# (931 MiB, lands on ROCm0 -- the fit charges it there) via the explicit
+# --mmproj flag.  This keeps frank-medium multimodal across the model swap.
+NO_MMPROJ=0
				`@@ -1 +0,0 @@`
				`entropy UUID=cd86c728-459e-4f3a-bcc2-41ecf84685ec /etc/entropy nofail,luks`