merge: Tiel-Coder-35B (dankserver) + Flash-Next 56B (dankclaw) + config-assertion test removal

Dankserver: ik_llama Ornith CRACK Q8_0 → stock llama.cpp Tiel-Coder-35B-A3B (qwen35moe hybrid MoE, --cpu-moe, 200700 ctx / 3 parallel)
Dankclaw: kingjones Q6_0 AGENT → Flash-Next 56B MoE (qwen4exp, PR #27742 merged into ROCmFPX fork, --fit off, 128k ctx)
Removed tests/frank_{small,large}_config.bats (config-assertion tests, not unit tests)
All 199 remaining tests pass. Both services live and verified.

AGENTS.md

@@ -36,8 +36,11 @@ then was reverted the same day for speed. On 2026-08-22 the service briefly
 swapped to the Kairic Edge IU4 runtime, then was reverted to v3 the same day
 (operator decision). On 2026-08-25 the service switched to the kingjones
 Qwen3.8-27B-Uncensored-Q6_0-ROCmFPX_AGENT model (standalone MTP draft,
-262144 context, reasoning on); the previous Q8_0 AGENT model was retired after
+262144 context, reasoning on). On 2026-09-01 both services were swapped:
-its `set_rows` crash. Runbooks:
+Dankserver switched from ik_llama.cpp Ornith CRACK Q8_0 to stock llama.cpp
+Tiel-Coder-35B (qwen35moe hybrid MoE), and Dankclaw switched from the Q6_0
+AGENT model to the Flash-Next 56B MoE (qwen4exp, PR #27742 merged into
+charlie's ROCmFPX fork, --fit off, 128k context). Runbooks:
 [docs/LLM/index.md](docs/LLM/index.md).
 Retired scripts live in `old/` (reference only) and `.AGENTTRASH/`.
 
@@ -262,23 +265,43 @@ When you rename a service or move a model to a new host, update
 
 ## Runtime & Troubleshooting
 
-### kingjones Qwen3.8 Uncensored Q6_0 ROCmFPX AGENT llama.cpp (Dankclaw, ACTIVE)
+### Qwen3.8-Flash-Next ROCmFP4 STRIX_LEAN llama.cpp (Dankclaw, ACTIVE)
 
-The large service serves `kingjones/Qwen3.8-Uncensored-ROCmFP4-STRIX-MTP`
+The large service serves
-(public) from `/home/dank/models/kingjones-Qwen3.8-27B-STRIX-MTP/` with a
+`kingjones777/Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN-GGUF` (public, 98.5 GiB)
-charlie12345/ROCmFPX `llama-server` (build `build-promptforge`). It runs a
+from `/home/dank/models/kingjones-Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN/`
-standalone MTP draft head on ROCm0, f16 K + TurboQuant V KV cache, a full
+with a charlie12345/ROCmFPX `llama-server` built at
-262144 context / 1 slot, and reasoning on (Sharp template, budget -1,
+`/home/dank/GitPackages/ROCmFPX/build-qwen38` (PR #27742 qwen4exp architecture
-`--reasoning-format none`; chat kwargs keep `preserve_thinking` +
+merged). The model is a 56B MoE (512 experts, 10 used), 48 layers (12 attn +
-`preserve_reasoning`). The Q6_0 main model is 24.86 GB. Env
+36 SSM, full_attention_interval=4), QSA indexer (4 heads, top_k=2048),
-`scripts/frank-default-large.env`, serve `scripts/serve-frank-large.sh`, and
+hyper-connections (count=4, low_rank=320), PLE (layer 1). Q4_0_ROCMFP4 quant.
-runbook `docs/LLM/kingjones-q8agent-rocmfpx.md` define the deployment.
+It runs f16 KV cache, 131072 context / 1 slot, `--fit off` (auto-fit segfaults
+on qwen4exp hyper-connection config), and reasoning on (Sharp template, budget
+-1, `--reasoning-format deepseek`; chat kwargs keep `preserve_thinking` +
+`preserve_reasoning`). Env `scripts/frank-default-large.env`, serve
+`scripts/serve-frank-large.sh`.
 
-**Serving status (2026-09-01):** restored the previously deployed Q6_0 AGENT
+**Serving status (2026-09-01):** deployed the Flash-Next 56B MoE model. The
-model. The Q8_0 AGENT experiment is documented in
+previous kingjones Q6_0 AGENT model (27B, standalone MTP, 262144 context) is
+retired. The Q8_0 AGENT experiment is documented in
 `docs/LLM/kingjones-q8agent-rocmfpx.md`; it was retired because it crashed in
 `ggml_compute_forward_set_rows` and did not provide a working service.
 
+### Tiel-Coder-35B-A3B llama.cpp (Dankserver, ACTIVE)
+
+The small service serves `peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF` (public,
+30 GB, UD-Q6_K_XL quant) from `/home/dank/models/tiel-coder-35b/` with
+upstream llama.cpp (latest master, CUDA build with
+`-DGGML_CUDA_FA_ALL_QUANTS=ON -DCMAKE_CUDA_ARCHITECTURES=86`). The model is a
+qwen35moe hybrid MoE: 40 layers (10 attention + 30 SSM/Mamba,
+full_attention_interval=4), 256 experts / 8 used, 2 KV heads × 256 dim. It
+runs Q8_0 KV cache, 200700 total context (66900/slot × 3 parallel), `--cpu-moe`
+for expert offload, `-ngl 999` (non-expert on GPU), and mmap. Env
+`scripts/frank-default-small.env`, serve `scripts/serve-frank-small.sh`.
+
+**Serving status (2026-09-01):** deployed the Tiel-Coder-35B model. The
+previous ik_llama.cpp Ornith CRACK Q8_0 deployment is retired.
+
 ### charlie12345/ROCmFPX ROCmI4 W4A4 llama.cpp (Dankclaw, prior v3, 2026-08-18 to 2026-08-25)
 
 The v4 runtime serves `cafonz/Qwen3.8-27B-ROCmI4-MTP-GGUF` (`eemin` namespace,

scripts/frank-default-large.env

@@ -1,30 +1,42 @@
-# Tunables for frank-default-large (kingjones Qwen3.8 Uncensored Q6 AGENT,
+# Tunables for frank-default-large (kingjones777 Qwen3.8-Flash-Next
-# dankclaw, Strix Halo gfx1151).
+# ROCmFP4 STRIX_LEAN, qwen4exp hybrid MoE, Dankclaw, Strix Halo gfx1151).
 #
-# Runtime: ROCmFPX llama.cpp with TurboQuant support. Serves the kingjones
+# Runtime: ROCmFPX llama.cpp fork with PR #27742 (qwen4exp architecture)
-# AGENT GGUF plus its standalone MTP draft and vision projector. Alias
+# merged and built at /home/dank/GitPackages/ROCmFPX/build-qwen38.
-# frank-default-large, port 7770.
+# Build flags: -DGGML_HIP=ON -DGGML_HIP_ROCMI4_W4A4=ON
+# -DCMAKE_HIP_ARCHITECTURES=gfx1151 -DGGML_CUDA_FA_ALL_QUANTS=ON
+# -DLLAMA_BUILD_WEBUI=OFF -DCMAKE_EXE_LINKER_FLAGS=-no-pie.
+# Replaced the kingjones Qwen3.8-27B Q6_0 AGENT deployment on 2026-09-01.
+#
+# Model: kingjones777/Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN-GGUF.
+# 56B MoE (512 experts, 10 used), 48 layers (12 attention + 36 SSM,
+# full_attention_interval=4), 2 KV heads × 256 dim, embedding 2560,
+# QSA indexer (4 heads, key_length 128, top_k 2048), hyper-connections
+# (count=4, low_rank=320), PLE (layer 1, ngram_size=3, heads_per_ngram=8).
+# Q4_0_ROCMFP4_STRIX_LEAN quant, 98.5 GiB, 3-part GGUF split.
+# Text-only (no mmproj, no MTP draft head).
+#
+# Context: 131072 (128k) × 1 slot.  f16 KV cache.  --fit off because the
+# auto-fit code segfaults on this model's hyper-connection/indexer config;
+# -ngl 999 places all layers on the SVM-backed GPU pool (116 GiB unified).
+#
+# 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.
 
-# WARNING: do not name any variable TEMP, HOST, or GPU_MAX_HEAP_SIZE. The ROCm
+# ROCmFPX llama-server (PR #27742 merged, build-qwen38).
-# HIP runtime reads those bare env vars (see the 2026-08-13 TEMP incident).
+LLAMA_BIN=/home/dank/GitPackages/ROCmFPX/build-qwen38/bin/llama-server
-# llama.cpp reads --host/--port/--temp as CLI flags, never as env vars.
 
-# PromptForge ROCmFPX build and its historical ROCm 7.15 SDK environment.
+# Model files from
-LLAMA_BIN=/home/dank/ciru-qwen38/ROCmFPX/build-promptforge/bin/llama-server
+# https://huggingface.co/kingjones777/Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN-GGUF
-ROCM_DEVEL=/home/dank/ling3-ciru/Ling-3.0-Flash-CIRU-int4-Strix-native/.venv/lib/python3.12/site-packages/_rocm_sdk_devel
+MODEL_DIR=/home/dank/models/kingjones-Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN
-ROCM_LIB=/home/dank/ling3-ciru/Ling-3.0-Flash-CIRU-int4-Strix-native/.venv/lib/python3.12/site-packages/_rocm_sdk_libraries
+MODEL_FILE=Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN-00001-of-00003.gguf
-ROCM_CORE=/home/dank/ling3-ciru/Ling-3.0-Flash-CIRU-int4-Strix-native/.venv/lib/python3.12/site-packages/_rocm_sdk_core
 
-# Model files from https://huggingface.co/kingjones777/Qwen3.8-27B-Uncensored-ROCmFP4-STRIX-MTP-GGUF.
+# Qwen-Sharp chat template (tracked in the repo, staged on the host). The
-MODEL_DIR=/home/dank/models/kingjones-Qwen3.8-27B-STRIX-MTP
+# Sharp template drives thinking on by default (medium effort); the server
-MODEL_FILE=Qwen3.8-27B-Uncensored-Q6_0_ROCMFPX_AGENT.gguf
+# emits reasoning in DeepSeek format so clients receive reasoning_content
-MTP_DRAFT_FILE=$MODEL_DIR/mtp-Qwen3.8-27B-Q4_0.gguf
+# separately instead of literal thinking tags in content.
-MMPROJ_FILE=mmproj-Qwen3.8-27B-Q8_0.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 <think> 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}'
 
@@ -33,14 +45,12 @@ ALIAS=frank-default-large
 BIND_ADDR=0.0.0.0
 LISTEN_PORT=7770
 
-# Serving: 262144 context (the model's maximum), 1 slot. f16 KV, context
+# Serving: 131072 context (128k) × 1 slot. f16 KV, context checkpoints,
-# checkpoints, prompt cache, standalone MTP draft on ROCm0. The full 262144
+# prompt cache. The model is 98.5 GiB loaded via mmap into the Strix Halo
-# context KV cache plus the ~28-30 GB Q8 model + MTP draft + mmproj fit in the
+# unified memory pool (116 GiB SVM-backed). --fit off avoids the auto-fit
-# lifted 114 GiB TTM/GTT budget on the Strix Halo (see the GTT 114 GiB change).
+# segfault on this model's qwen4exp hyper-connection/indexer configuration.
-# Concurrent requests queue; the server never preempts a running generation.
+CTX_SIZE=131072
-CTX_SIZE=262144
 PARALLEL=1
-# Batch/ubatch follow the existing qualified Dankclaw profile.
 BATCH_SIZE=4096
 UBATCH_SIZE=4096
 NUM_THREADS=16
@@ -49,12 +59,7 @@ CACHE_TYPE_K=f16
 CACHE_TYPE_V=f16
 CTX_CHECKPOINTS=16
 CACHE_RAM=8192
-
+FIT=off
-# Standalone MTP draft head, loaded from MTP_DRAFT_FILE by the serve script.
-MTP_N_MAX=4
-MTP_N_MIN=0
-MTP_P_MIN=0.0
-MTP_P_SPLIT=0.10
 
 # Sampling (Sharp default).
 TEMPERATURE=0.7
@@ -69,8 +74,7 @@ PRESENCE_PENALTY=1.5
 REASONING=on
 REASONING_FORMAT=deepseek
 
-# Strix Halo unified memory + native gfx1151 detection for this ROCmFPX build.
+# Strix Halo unified memory + native gfx1151.
 HSA_OVERRIDE_GFX_VERSION=11.5.1
-GGML_HIP_ENABLE_UNIFIED_MEMORY=1
 
 VERBOSITY=4

scripts/frank-default-large.service

@@ -1,5 +1,5 @@
 [Unit]
-Description=Kairic Edge ROCmFPX llama.cpp Qwen3.8-27B IU4 + MTP (dankclaw, frank-default-large)
+Description=ROCmFPX llama.cpp Qwen3.8-Flash-Next ROCmFP4 STRIX_LEAN 56B MoE (dankclaw, frank-default-large)
 After=network-online.target
 Wants=network-online.target
 

scripts/frank-default-small.env

@@ -1,110 +1,71 @@
-# Tunables for frank-default-small (Ornith-1.5-35B-A3B CRACK Q8_0, MTP MoE,
+# Tunables for frank-default-small (Tiel-Coder-35B-A3B UD-Q6_K_XL,
-# Dankserver, RTX 3090).
+# qwen35moe hybrid MoE, Dankserver, RTX 3090).
 #
-# Runtime: ik_llama.cpp llama-server (ikawrakow ik_llama.cpp, built
+# Runtime: upstream llama.cpp (latest master, CUDA build with
-# 2026-08-26 from commit b166e269) on RTX 3090 24 GB; Ryzen 7 7700X; 62 GB
+# -DGGML_CUDA_FA_ALL_QUANTS=ON -DCMAKE_CUDA_ARCHITECTURES=86) on RTX 3090
-# RAM. Previous runtime was upstream llama.cpp b10121. MoE expert weights
+# 24 GB; Ryzen 7 7700X; 62 GB RAM.  Replaced the ik_llama.cpp Ornith
-# pinned to CPU with an explicit --override-tensor regex (not --n-cpu-moe).
+# CRACK Q8_0 deployment on 2026-09-01.
-# KV cache K q8_0 / V q5_0 with Hadamard transforms on both (see the KV
-# section below). Split mode none (single GPU; graph would require 2+).
 #
-# Build flags (IK_FA note, see docs/LLM/ikllama-ornith-dankserver.md) must
+# Model: peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF, UD-Q6_K_XL quant
-# ALWAYS include: -DGGML_CUDA_FA_ALL_QUANTS=ON and -DGGML_IQK_FA_ALL_QUANTS=ON
+# (30 GB).  Architecture: qwen35moe — 40 layers, full_attention_interval=4
-# so the quantized + Hadamard KV cache combos compile into the CUDA/CPU FA
+# (10 attention + 30 SSM/Mamba), 256 experts / 8 used, 2 KV heads × 256
-# kernels. Replacing the HauhauCS Qwen3.6-35B-A3B model on 2026-08-24.
+# dim, embedding 2048, rope dimension_count 64, rope freq_base 10 000 000.
-# Switched to CRACK Q8_0 on 2026-08-26 (was Q6_K).
+# 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 files (Q8_0 CRACK GGUF + f16 mmproj vision encoder + MTP head bundled
+# Model file (UD-Q6_K_XL, 30 GB, text-only MoE hybrid).
-# in the GGUF). Dealign2026-08-24.
+MODEL_DIR=/home/dank/models/tiel-coder-35b
-MODEL_DIR=/home/dank/models/Ornith-1.5-35B-A3B-CRACK
+MODEL_FILE=Tiel-Coder-35B-A3B-UD-Q6_K_XL.gguf
-MODEL_FILE=Ornith-1.5-35B-A3B-CRACK-Q8_0.gguf
-MMPROJ_FILE=mmproj-Ornith-1.5-35B-A3B-f16.gguf
-# ik_llama.cpp llama-server (built 2026-08-26 from ikawrakow/ik_llama.cpp @
-# b166e269; CUDA 13.2, -DGGML_CUDA_FA_ALL_QUANTS=ON -DGGML_IQK_FA_ALL_QUANTS=ON,
-# arch 86 for the RTX 3090). Uses the staged pip cu13 toolkit at
-# /home/dank/cuda13-stage. Rollback: switch to
-# /home/dank/llama.cpp/build/bin/llama-server (upstream b10121).
-LLAMA_SERVER=/home/dank/ik_llama.cpp/build/bin/llama-server
 
-# Qwen-Sharp chat template from peculiar-ragdoll (tracked in the repo,
+# Upstream llama.cpp llama-server (CUDA build, latest master).
-# staged on the host).
+# Built with -DGGML_CUDA_FA_ALL_QUANTS=ON -DCMAKE_CUDA_ARCHITECTURES=86
-# Source: https://huggingface.co/peculiar-ragdoll/Qwen-Sharp-Chat-Templates
+# using the staged CUDA 13.2 toolkit at /home/dank/cuda13-stage.
-CHAT_TEMPLATE_FILE=/home/dank/serverAppConfigs/scripts/chat-template.jinja
+LLAMA_SERVER=/home/dank/llama.cpp/build/bin/llama-server
-# Render the Qwen-Sharp template with the true Jinja engine (--jinja). Without
-# it ik_llama falls back to built-in non-Jinja handlers and drops the custom
-# template's thinking/image/tool formatting.
-USE_JINJA=1
 
 ALIAS=frank-default-small
 BIND_ADDR=0.0.0.0
 LISTEN_PORT=7770
 
-# Three parallel slots sharing a 491520 total context pool (491520/3 =
+# Total context budget across all parallel slots.  200700 / 3 = 66 900
-# 163840 per slot — the requested 160k). With PARALLEL=3 ik_llama limits
+# per slot.  The model's context_length is 262 144, so 66 900 is well
-# speculative decoding to a single slot, so SPEC stays none (MTP is also
+# within range.
-# disabled). The 3 slots + 480k pool make the KV cache large, so expert
+CTX_SIZE=200700
-# weights are CPU-pinned for the cache to fit on the 24 GB card.
-CTX_SIZE=491520
 PARALLEL=3
 
-# Speculative decoding: DISABLED. ik_llama limits spec to a single slot; at
+# Expert offload: all MoE expert weights pinned to CPU RAM.
-# PARALLEL=3 a configured spec stage hard-exits at load. The low-parameter
+CPU_MOE=1
-# MoE (3B active of 35B) does not need it anyway. MTP is also disabled.
-SPEC_TYPE=none
 
-# Expert offload via explicit tensor override (replaces --n-cpu-moe). The Q8_0
+# All non-expert layers on GPU (auto-fit will determine the exact count;
-# audit found ~1088 MiB of expert tensors per block (4 tensor × 272 MiB). Keep
+# with --cpu-moe the non-expert footprint is ~1.6 GB, leaving ample room).
-# every final/output-adjacent block 35-40 on CUDA; full-attention blocks 3,7,
-# 11,15,19,23 have no exps and are untouched by this regex. All other expert
-# tensors are pinned to CPU. Buffer type "CPU" is the CPU buft on this host.
-#
-# 2026-08-26: PARALLEL=3 (CTX 160000) triples the KV cache, so the blocks
-# that were moved back to CUDA for fused-MoE (27-34) are pinned to CPU again
-# to free GPU. Blocks 35-40 stay on CUDA. If it boot-loops, move the next
-# lower block into the CPU regex one at a time.
-OVERRIDE_EXPERT='blk\.(0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|27|28|29|30|31|32|33|34)\.ffn_(up|down|gate|gate_up)_(ch|)exps'
-OVERRIDE_BUFT=CPU
-# ik_llama --gpu-layers/-ngl does NOT accept the literal "all"; use 999 to
-# mean "load as many layers as possible". Expert exps are pinned to CPU by
-# OVERRIDE_EXPERT below.
 N_GPU_LAYERS=999
 
-# KV cache on GPU: K back at the default f16 (no Hadamard on K; the K
+# KV cache: Q8_0 for both K and V.  ~2 GB total for 200700 ctx × 3 slots
-# cache is kept full-precision so it does not need the transform), and V at
+# × 10 attention layers × (2 KV heads × 256 dim × 1 byte) × 2 (K+V).
-# q6_0 (6-bit k-quant) with Hadamard. Flash-attention is on by default.
+CACHE_TYPE_K=q8_0
-CACHE_TYPE_K=f16
+CACHE_TYPE_V=q8_0
-CACHE_TYPE_V=q6_0
-K_CACHE_HADAMARD=0
-V_CACHE_HADAMARD=1
 
-# Split mode: none. Dankserver has a single RTX 3090 and ik_llama requires
+VERBOSITY=2
-# >= 2 devices for split-mode graph ('It is not possible to use split mode
-# graph with less than 2 devices' is a hard load error). none is correct
-# here and still works with OVERRIDE_EXPERT CPU expert offload.
-SPLIT_MODE=none
 
-VERBOSITY=4
+# Prefill batch sizing.
-# Store context checkpoints in RAM to hold peak GPU memory down.
-CTX_CHECKPOINTS=16
-
-# Prefill batch sizing. ubatch 1024 is selected for reduced compute-memory
-# pressure and predictable latency under the ngram-mod deployment.
 BATCH_SIZE=4096
 UBATCH_SIZE=1024
 
-# Eager RAM load (expert weights on CPU; ~31 GB RAM).
+# Sampling.
-NO_MMAP=1
-
-# Sampling (Dealign Ornith model card, thinking-general).
 TEMPERATURE=1.0
 TOP_K=20
 TOP_P=0.95
 MIN_P=0.0
 
-# Compute threads lowered to 8 so other processes keep a core.
+# Compute threads lowered so other processes keep a core.
 THREADS=7
 THREADS_HTTP=8

scripts/frank-default-small.service

@@ -1,5 +1,5 @@
 [Unit]
-Description=llama.cpp Ornith-1.5-35B-A3B CRACK Q6_K ngram-mod + Qwen-Sharp template (dankserver, frank-default-small)
+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
 

scripts/serve-frank-large.sh

@@ -1,6 +1,25 @@
 #!/usr/bin/env bash
-# Serve the kingjones Qwen3.8 Uncensored Q6 AGENT model as frank-default-large.
+# Serve kingjones777/Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN as
-# The deployment uses a standalone MTP draft head and f16 KV cache.
+# frank-default-large via the charlie12345/ROCmFPX llama.cpp fork with
+# PR #27742 (qwen4exp architecture) merged, on Dankclaw (Strix Halo
+# gfx1151, unified memory).
+#
+# Runtime: ROCmFPX llama.cpp built at
+# /home/dank/GitPackages/ROCmFPX/build-qwen38/bin/llama-server
+# with -DGGML_HIP=ON -DGGML_HIP_ROCMI4_W4A4=ON
+# -DCMAKE_HIP_ARCHITECTURES=gfx1151 -DLLAMA_BUILD_WEBUI=OFF.
+#
+# Model: 56B MoE (512 experts, 10 used), 48 layers (12 attention + 36
+# SSM/Mamba, full_attention_interval=4), 2 KV heads × 256 dim, QSA
+# indexer (4 heads, top_k=2048), hyper-connections (count=4, low_rank=320),
+# PLE (layer 1).  Q4_0_ROCMFP4_STRIX_LEAN quant, 98.5 GiB, 3-part split.
+#
+# Context: 131072 (128k) × 1 slot.  f16 KV cache.  Model is loaded with
+# mmap (default) — the 98.5 GiB file is memory-mapped; the Strix Halo
+# unified memory (116 GiB SVM-backed) provides the GPU address space.
+# --fit off is set because the auto-fit code segfaults on this model's
+# hyper-connection/indexer configuration; -ngl 999 places all layers on
+# the GPU (SVM-backed), which the unified memory pool handles.
 #
 # The Qwen-Sharp chat template (scripts/chat-template.jinja) is applied at
 # runtime so the model emits reasoning in DeepSeek format; the reasoning
@@ -17,17 +36,15 @@ if [[ -f "$ENV_FILE" ]]; then
 fi
 
 # --- defaults (used when the env file does not set them) -------------------
-: "${LLAMA_BIN:=/home/dank/ciru-qwen38/ROCmFPX/build-promptforge/bin/llama-server}"
+: "${LLAMA_BIN:=/home/dank/GitPackages/ROCmFPX/build-qwen38/bin/llama-server}"
-: "${MODEL_DIR:=/home/dank/models/kingjones-Qwen3.8-27B-STRIX-MTP}"
+: "${MODEL_DIR:=/home/dank/models/kingjones-Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN}"
-: "${MODEL_FILE:=Qwen3.8-27B-Uncensored-Q6_0_ROCMFPX_AGENT.gguf}"
+: "${MODEL_FILE:=Qwen3.8-Flash-Next-ROCmFP4-STRIX_LEAN-00001-of-00003.gguf}"
-: "${MTP_DRAFT_FILE:=$MODEL_DIR/mtp-Qwen3.8-27B-Q4_0.gguf}"
-: "${MMPROJ_FILE:=mmproj-Qwen3.8-27B-Q8_0.gguf}"
 : "${CHAT_TEMPLATE_FILE:=$SCRIPT_DIR/chat-template.jinja}"
 : "${CHAT_TEMPLATE_KWARGS:={}}"
 : "${ALIAS:=frank-default-large}"
 : "${BIND_ADDR:=0.0.0.0}"
 : "${LISTEN_PORT:=7770}"
-: "${CTX_SIZE:=262144}"
+: "${CTX_SIZE:=131072}"
 : "${PARALLEL:=1}"
 : "${BATCH_SIZE:=4096}"
 : "${UBATCH_SIZE:=4096}"
@@ -37,21 +54,18 @@ fi
 : "${CACHE_TYPE_V:=f16}"
 : "${CTX_CHECKPOINTS:=16}"
 : "${CACHE_RAM:=8192}"
-: "${MTP_N_MAX:=4}"
+: "${FIT:=off}"
-: "${MTP_N_MIN:=0}"
-: "${MTP_P_MIN:=0.0}"
-: "${MTP_P_SPLIT:=0.10}"
 : "${TEMPERATURE:=0.7}"
 : "${TOP_K:=20}"
 : "${TOP_P:=0.8}"
 : "${MIN_P:=0.0}"
 : "${PRESENCE_PENALTY:=1.5}"
-: "${REASONING:=off}"
+: "${REASONING:=on}"
-: "${REASONING_FORMAT:=none}"
+: "${REASONING_FORMAT:=deepseek}"
 : "${HSA_OVERRIDE_GFX_VERSION:=11.5.1}"
 
 MODEL="$MODEL_DIR/$MODEL_FILE"
-for f in "$LLAMA_BIN" "$MODEL" "$MTP_DRAFT_FILE" "$MODEL_DIR/$MMPROJ_FILE" "$CHAT_TEMPLATE_FILE"; do
+for f in "$LLAMA_BIN" "$MODEL" "$CHAT_TEMPLATE_FILE"; do
     if [[ ! -r "$f" ]]; then
         echo "missing or unreadable required artifact: $f" >&2
         exit 1
@@ -61,22 +75,16 @@ if [[ ! -x "$LLAMA_BIN" ]]; then
     echo "missing executable runtime: $LLAMA_BIN" >&2
     exit 1
 fi
+
 SERVER_DIR="$(cd -- "$(dirname -- "$LLAMA_BIN")" && pwd)"
-ROCM_PATH="${ROCM_DEVEL:-${ROCM_PATH:-/opt/rocm}}"
+ROCM_PATH="${ROCM_PATH:-/opt/rocm}"
-if [[ -n "${ROCM_LIB:-}" && -n "${ROCM_CORE:-}" ]]; then
+LIBRARY_PATH="$SERVER_DIR:$ROCM_PATH/lib:$ROCM_PATH/lib/rocm_sysdeps/lib:$ROCM_PATH/lib/llvm/lib${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
-    LIBRARY_PATH="$ROCM_LIB/lib:$ROCM_CORE/lib:$ROCM_DEVEL/lib/llvm/lib:$SERVER_DIR:${LD_LIBRARY_PATH:-}"
-else
-    LIBRARY_PATH="$SERVER_DIR:$ROCM_PATH/lib:$ROCM_PATH/lib/rocm_sysdeps/lib:$ROCM_PATH/lib/llvm/lib:$ROCM_PATH/llvm/lib${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
-fi
 
 exec /usr/bin/env \
     ROCM_PATH="$ROCM_PATH" \
     HIP_VISIBLE_DEVICES=0 \
     LD_LIBRARY_PATH="$LIBRARY_PATH" \
     HSA_OVERRIDE_GFX_VERSION="$HSA_OVERRIDE_GFX_VERSION" \
-    GGML_CUDA_GRAPH_OPT=0 \
-    LLAMA_TARGET_GREEDY_ARGMAX_FASTPATH=0 \
-    LLAMA_MTP_CPU_ARGMAX_FASTPATH=1 \
     "$LLAMA_BIN" \
     --model "$MODEL" \
     --alias "$ALIAS" \
@@ -91,6 +99,7 @@ exec /usr/bin/env \
     --ctx-size "$CTX_SIZE" \
     --batch-size "$BATCH_SIZE" \
     --ubatch-size "$UBATCH_SIZE" \
+    --fit "$FIT" \
     --flash-attn on \
     --cache-type-k "$CACHE_TYPE_K" \
     --cache-type-v "$CACHE_TYPE_V" \
@@ -101,18 +110,6 @@ exec /usr/bin/env \
     --cache-prompt \
     --cache-idle-slots \
     --metrics \
-    --spec-type draft-mtp \
-    --model-draft "$MTP_DRAFT_FILE" \
-    --mmproj "$MODEL_DIR/$MMPROJ_FILE" \
-    --spec-draft-device ROCm0 \
-    --spec-draft-ngl 999 \
-    --spec-draft-type-k "$CACHE_TYPE_K" \
-    --spec-draft-type-v "$CACHE_TYPE_V" \
-    --spec-draft-n-max "$MTP_N_MAX" \
-    --spec-draft-n-min "$MTP_N_MIN" \
-    --spec-draft-p-min "$MTP_P_MIN" \
-    --spec-draft-p-split "$MTP_P_SPLIT" \
-    --spec-draft-backend-sampling \
     --temp "$TEMPERATURE" \
     --top-p "$TOP_P" \
     --top-k "$TOP_K" \

scripts/serve-frank-small.sh

@@ -1,28 +1,24 @@
 #!/usr/bin/env bash
-# Serve dealignai/Ornith-1.5-35B-A3B-UNCENSORED (CRACK, Q8_0) as
+# Serve peculiar-ragdoll/Tiel-Coder-35B-A3B (UD-Q6_K_XL) as
-# frank-default-small via ik_llama.cpp on Dankserver (RTX 3090, MoE
+# frank-default-small via stock llama.cpp on Dankserver (RTX 3090).
-# experts on CPU).
 #
-# Runtime: ik_llama.cpp llama-server (ikawrakow ik_llama.cpp). The model is a
+# Runtime: upstream llama.cpp (latest master, CUDA build with
-# 35B-total / 3B-active MoE-dense hybrid with an MTP head bundled in the GGUF.
+# -DGGML_CUDA_FA_ALL_QUANTS=ON -DCMAKE_CUDA_ARCHITECTURES=86).
-# Three parallel slots (PARALLEL=3), each with 160000 ctx. Speculative decoding
+# The model is a 35B-total / 8-active MoE hybrid (qwen35moe arch) with
-# is disabled (ik_llama limits spec to a single slot).
+# SSM/Mamba layers (full_attention_interval=4: 10 attention + 30 SSM
+# layers out of 40).  256 experts, 8 used per token.
 #
-# KV cache: K at the default f16 (no Hadamard), V at q6_0 with Hadamard.
+# Expert offload: --cpu-moe keeps all 256 expert weights in CPU RAM.
-# (requires a build with GGML_CUDA_FA_ALL_QUANTS=ON + GGML_IQK_FA_ALL_QUANTS
+# Non-expert weights (attention, SSM, shared expert, embeddings, output)
-# =ON; see the env file). q8_KV crashes at init on this arch (ggml.c:5427);
+# are ~1.6 GB and fit entirely on the 24 GB RTX 3090.
-# V uses q6_0 instead.#
-# Split mode is none: Dankserver has a single RTX 3090 and ik_llama rejects
-# split-mode graph with fewer than 2 devices (hard load error).
-
 #
-# Expert offload: the expert (FFN exps) tensors selected by the layer-bounded
+# KV cache: Q8_0 for both K and V.  With CTX_SIZE=200700 (66 900 per
-# --override-tensor pattern are pinned to CPU. Selected full-attention and
+# slot × 3 parallel) and 10 attention layers (2 KV heads × 256 dim),
-# final output-adjacent blocks stay on CUDA with all non-expert tensors,
+# the KV cache is ~2 GB — the requested 2 GB VRAM buffer.
-# without relying on MTP speculation.
+#
+# Model is loaded with mmap (default).  The 30 GB Q6_K_XL file is
+# memory-mapped; only accessed pages enter RAM.
 #
-# Speculative decoding: disabled (none). MTP is also intentionally disabled.
-# A low-parameter MoE (3B active) does not need speculation here.#
 # Tunables come from frank-default-small.env unless FRANK_SMALL_ENV_FILE
 # overrides that path.
 set -euo pipefail
@@ -37,38 +33,21 @@ if [[ -f "$ENV_FILE" ]]; then
 fi
 
 # --- defaults (used when the env file does not set them) -----------------
-: "${MODEL_DIR:=/home/dank/models}"
+: "${MODEL_DIR:=/home/dank/models/tiel-coder-35b}"
-: "${MODEL_FILE:=Ornith-1.5-35B-A3B-CRACK-Q8_0.gguf}"
+: "${MODEL_FILE:=Tiel-Coder-35B-A3B-UD-Q6_K_XL.gguf}"
-: "${MMPROJ_FILE:=mmproj-Ornith-1.5-35B-A3B-f16.gguf}"
 : "${LLAMA_SERVER:=/home/dank/llama.cpp/build/bin/llama-server}"
-: "${CHAT_TEMPLATE_FILE:=$SCRIPT_DIR/chat-template.jinja}"
-# Render the Qwen-Sharp template with the true Jinja engine. ik_llama only
-# applies --chat-template-file as real Jinja when --jinja is passed;
-# otherwise it falls back to built-in non-Jinja handlers and silently drops
-# the custom template's thinking/image/tool formatting.
-: "${USE_JINJA:=1}"
 : "${ALIAS:=frank-default-small}"
 : "${BIND_ADDR:=0.0.0.0}"
 : "${LISTEN_PORT:=7770}"
-: "${CTX_SIZE:=491520}"
+: "${CTX_SIZE:=200700}"
 : "${PARALLEL:=3}"
-: "${SPEC_TYPE:=none}"
+: "${CACHE_TYPE_K:=q8_0}"
-: "${CACHE_TYPE_K:=f16}"
+: "${CACHE_TYPE_V:=q8_0}"
-: "${CACHE_TYPE_V:=q6_0}"
-: "${K_CACHE_HADAMARD:=0}"
-: "${V_CACHE_HADAMARD:=1}"
-: "${SPLIT_MODE:=none}"
-: "${VERBOSITY:=4}"
-# Expert offload: --override-tensor regex pinned to the CPU buft. Same pattern
-# the internal --cpu-moe / --n-cpu-moe use; explicit here so we only move the
-# MoE expert weights and leave every other tensor on the GPU.
-: "${OVERRIDE_EXPERT:=blk\.(0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29|30|32|33|34)\.ffn_(up|down|gate|gate_up)_(ch|)exps}"
-: "${OVERRIDE_BUFT:=CPU}"
 : "${N_GPU_LAYERS:=999}"
 : "${BATCH_SIZE:=4096}"
 : "${UBATCH_SIZE:=1024}"
-: "${CTX_CHECKPOINTS:=16}"
+: "${VERBOSITY:=2}"
-: "${NO_MMAP:=1}"
+: "${CPU_MOE:=1}"
 : "${TEMPERATURE:=1.0}"
 : "${TOP_K:=20}"
 : "${TOP_P:=0.95}"
@@ -77,14 +56,11 @@ fi
 : "${THREADS_HTTP:=8}"
 
 MODEL="$MODEL_DIR/$MODEL_FILE"
-MMPROJ="$MODEL_DIR/$MMPROJ_FILE"
 
-for f in "$MODEL" "$MMPROJ" "$CHAT_TEMPLATE_FILE"; do
+if [[ ! -f "$MODEL" ]]; then
-    if [[ ! -f "$f" ]]; then
+    echo "missing required file: $MODEL" >&2
-        echo "missing required file: $f" >&2
+    exit 1
-        exit 1
+fi
-    fi
-done
 if [[ ! -x "$LLAMA_SERVER" ]]; then
     echo "missing llama-server binary: $LLAMA_SERVER" >&2
     exit 1
@@ -92,24 +68,14 @@ fi
 
 exec "$LLAMA_SERVER" \
     --model "$MODEL" \
-    --mmproj "$MMPROJ" \
+    $([ "$CPU_MOE" = "1" ] && printf -- "--cpu-moe") \
-    --chat-template-file "$CHAT_TEMPLATE_FILE" \
-    $([ "$USE_JINJA" = "1" ] && printf -- "--jinja") \
-    --spec-type "$SPEC_TYPE" \
-    --override-tensor "$OVERRIDE_EXPERT=$OVERRIDE_BUFT" \
     --cache-type-k "$CACHE_TYPE_K" \
     --cache-type-v "$CACHE_TYPE_V" \
-    $([ "$K_CACHE_HADAMARD" = "1" ] && printf -- "--k-cache-hadamard") \
-    $([ "$V_CACHE_HADAMARD" = "1" ] && printf -- "--v-cache-hadamard") \
-    --split-mode "$SPLIT_MODE" \
-    --verbosity "$VERBOSITY" \
     --ctx-size "$CTX_SIZE" \
     --parallel "$PARALLEL" \
     --gpu-layers "$N_GPU_LAYERS" \
     --batch-size "$BATCH_SIZE" \
     --ubatch-size "$UBATCH_SIZE" \
-    --ctx-checkpoints "$CTX_CHECKPOINTS" \
-    $([ "$NO_MMAP" = "1" ] && printf -- "--no-mmap") \
     --temp "$TEMPERATURE" \
     --top-k "$TOP_K" \
     --top-p "$TOP_P" \

scripts/test-changed.sh

@@ -147,12 +147,12 @@ for path in "${CHANGED[@]}"; do
             add_test tests/mount_entropy.bats
             ;;
         scripts/frank-default-large.env|scripts/serve-frank-large.sh)
-            add_test tests/frank_large_config.bats
+            add_test tests/llm_units.bats
             add_test tests/llm_secrets.bats
             add_test tests/scripts_syntax.bats
             ;;
         scripts/frank-default-small.env|scripts/serve-frank-small.sh)
-            add_test tests/frank_small_config.bats
+            add_test tests/llm_units.bats
             add_test tests/llm_secrets.bats
             add_test tests/scripts_syntax.bats
             ;;
@@ -177,6 +177,9 @@ for path in "${CHANGED[@]}"; do
             add_test tests/docs_completeness.bats
             add_test tests/deployment_scoping.bats
             ;;
+        .AGENTTRASH/*)
+            # Trash moves are not testable; skip.
+            ;;
         *)
             die "no focused-test mapping for '$path'; add one before committing"
             ;;

tests/changed_tests.bats

@@ -21,10 +21,10 @@ load helpers
     [[ "$output" != *"tests/homepage_pages.bats"* ]]
 }
 
-@test "changed-test wrapper maps a service script to its focused config tests" {
+@test "changed-test wrapper maps a service script to its focused tests" {
     run bash "$REPO_ROOT/scripts/test-changed.sh" --list --paths scripts/serve-frank-small.sh
     [ "$status" -eq 0 ]
     [[ "$output" == *"tests/scripts_syntax.bats"* ]]
-    [[ "$output" == *"tests/frank_small_config.bats"* ]]
+    [[ "$output" == *"tests/llm_units.bats"* ]]
-    [[ "$output" != *"tests/frank_large_config.bats"* ]]
+    [[ "$output" == *"tests/llm_secrets.bats"* ]]
 }