Six statements still claimed the endpoint did not exist: both vault descriptions, the two index.md bullets, the Open Notebook blocker, and the engine page deploy order that opened with "Nothing below has been run". All re-keyed to the deployed state, with the ledger the deploys printed: engine healthy 31 s after the build at 4.0 GiB resident, three voices warm, the ladder measured at 1883 ms then 1226-1283 ms (not the retired 6.2-8.0 s), and two voices rendering the same sentence into different bytes.
Adds a The harness speech role section to the endpoint page with the three wiring laws measured the day of the flip: speech.voice is the forwarded key (tts.localVoice belongs to Kokoro), an explicit speech role has no fallback candidate, and models.yml loads at process start while config.yml is read live - so a session older than the frank-tts row fails exactly like a dead endpoint. Fresh-process proof: omp -p --tools tts saved 180524 bytes with voice=narrator1 while the facade logged the matching speech line.
Stage 2 of the voice stack. Repo state only: nothing deployed, no host touched,
no container started.
- `frank-tts` is now registered everywhere a dankclaw service has to be:
`SERVICE_ORDER` (after `frank-tts-engine`, before the consumers), the planner's
dankclaw set, `deploy-remote.sh`'s allowlist + dispatch + a
`deploy_frank_tts` handler, and `deploy.sh`'s dispatch regex — plus the
`build_compose_image` arm naming `dankclaw/voiceservice/Containerfile`, without
which the build silently returns 0 and the unit's `Pull=never` tag never
exists.
- The unit joins TWO networks, in order: `dankclaw_default` first (default route,
where consumers and open-notebook live), then the voice bridge. `TTS_UPSTREAM`
moves from `host.containers.internal:8880` to `http://frank-tts-engine:8880` —
the actual fix for a shape that could never work, since a bridge container
cannot reach a host-loopback bind and the engine publishes nothing at all.
- `Wants=`/`After=` on the engine, never `Requires=`: the `llama-tts-cli` rollback
backend needs no upstream, so a broken engine may delay this endpoint but must
not take its unit down.
- The handler installs the engine's quadlet pair too, because a referenced
`Network=` file the generator cannot find means no façade unit is produced at
all, and `--only frank-tts` has no other way to get it. It creates
`/home/dank/voices` as the deploy user (rootless podman would auto-create a
missing bind source root-owned, which reads as "no voices"), and gates on
podman's recorded health.
- Voice naming is now written down where it can be found: a voice is a lowercase
directory name matching `^[a-z0-9][a-z0-9._-]*$`, at most 64 chars. That shape
is the intersection of two rules that disagree — the shim would accept mixed
case, the engine lowercases names — so a mixed-case pack resolves on one side
only, and the boot warmer refuses it loudly. `narrator1/2/3` are load-bearing
for open-notebook.
- Homepage card for the pair, `wip: down` and hidden, probing `127.0.0.1:7772`;
the engine gets no card because it publishes nothing and could never be probed.
- **No Prometheus job, deliberately.** No `/metrics` exists on the shim, this
Prometheus 3 refuses HTML/JSON bodies (the existing `comfyui` and `services`
probes are documented down for exactly that), and `LlmEndpointDown` is scoped
`up{job=~"frank-.*"}` — so a `/health` job would page on Signal for a healthy
endpoint. Recorded in the page instead of quietly skipped.
- Vault: new [[Frank TTS]] page linked from the index; [[Frank TTS Engine]] now
points at it for the pair's public signal; log entry for the day.
- Seam grows to 25 tests: the generator's rendering of the pair (both
attachments, first-network order, one publish, Wants-not-Requires, bounded
per-unit blocks), the env file's upstream checked as a network JOIN rather than
a string, planner separation and order, the handler's steps, the card's gating,
routing that selects both seams for either deployable.
Repo state only: nothing deployed, no host touched, no container started.
- `frank-tts-engine` on Dankclaw: `qwentts.cpp` commit
5916bf7c79dbb4e66fc059cdc507a79a23a28bb4, resident `tts-server`, 1.7 B base
talker Q4_K_M + Q8_0 tokenizer, container port 8880. It is deliberately NOT a
fleet endpoint — `frank-tts` on 7772 is the endpoint consumers address.
- Publishes NOTHING. A host-loopback publish would have been useless to the
façade (netavark DNATs published ports only, so a bridge container cannot
reach a host loopback bind — the shape that 502'd hermes) and would have been
the one LAN-reachable surface of an unauthenticated synthesiser. Both voice
containers sit on one committed quadlet network, `frank-tts-engine.network`
pinned to 10.89.5.0/24 (next free /24 after open-notebook's 10.89.4.0/24), and
peers address the engine by container name. Diagnostics and the deployer's
health gate run inside the netns (`podman exec`, `wait_container_healthy`).
- Cold-voice law honoured in the unit: the first request for a registered voice
decodes its whole reference to snapshot codec state (6.2-8.0 s, first audio
byte 5 530 ms), later ones restore in 0.5 ms, and the registry is memory-only.
`ExecStartPost=-/usr/bin/podman exec frank-tts-engine /app/warm-voices.sh`
with `TimeoutStartSec=900`; the `-` prefix keeps a failed warm from
crash-looping the endpoint under `Restart=always`. It logs cold=/warm= per
voice, so the under-a-second second request is proven on every boot.
- Thread law recorded: `tts-server` has no `--threads`; src/backend.h uses
hardware_concurrency()/2, so a cpuset pin silently changes the thread count.
- Image builds from the pinned commit with the proven line (GGML_BLAS=OFF
forced by this host, every GPU backend off, -j 8, no -DGGML_NATIVE because
ggml defaults to -march=native for the building host). Two stages; the runtime
stage COPYs both the entrypoint and the warmer, because a container that
publishes nothing is reached with `podman exec`, not a port.
- `scripts/download-frank-tts-engine-qwentts.sh` fetches the two GGUFs with byte
and LFS-oid checks (talker 1 219 245 248 B, tokenizer 291 150 624 B); the
deployer gates on --verify-only instead of pulling 1.4 GiB mid-deploy.
- Registered in SERVICE_ORDER, DANKCLAW_SERVICES, the deploy-remote.sh dispatch,
its pre-deploy requirement, deploy.sh's regex, the planner's path map (the
whole `dankclaw/voiceservice/` tree mapped to nothing before this), and
test-changed.sh ABOVE the `dankclaw/*.container` case whose `*` crosses `/`.
- New seam tests/frank_tts_engine.bats (19): the Quadlet generator's own output
for network/publish/boot/warm shape, entrypoint argv driven from the tracked
env file, the warmer against a stub tts-server over a real socket
(registration keys, ref_text absent, late engine, hung synthesis, name
refusal), the fetcher's verifier, the pinned cmake line, and both routing
tables. Suite: 422 passed.
- Vault: [[Frank TTS Engine]] linked from the index; log.md entry for the day.
The page a new operator needs: what the three required files are and who
reads each one, how a pack is built (with the extractor's usage text
verbatim, because the flags are not guessable and there is no -o), the
10-second gate and where it fires twice, how to add a fourth voice
including the warm hook that registers a pack without restarting the
engine, and the engine-bump procedure for when the artifact sizes move.
Plus the host census as of 2026-10-18 — the trial packs carry no
artifacts and their manifest is the trial's own schema, so the builder's
first real run is a --force migration — and the three gaps left open on
purpose: no curated ICL transcript, no 16 kHz variant, no loudness pass.
Linked from the Dankclaw services list; log entry appended oldest-first.
Page registered in the vault index (the docs-completeness seam fails an
unregistered page) and logged in the bundle change log.
What the page owes the reader: the provider contract (frank-zealot for scripts
with a /v1 base URL, frank-tts for voices with 1-4 speakers mapped to the
narrator1/2/3 voice packs typed as free text), why Test Connection and Discover
Models fail cosmetically against frank-tts rather than being a fault, the
encryption key as the precondition for saving any credential, and the honest
blocker — frank-tts is not deployed and nothing listens on 7772, so a generation
run writes its script and fails on the first voice.
Episode destination is stated as interim and cites Gitea issue #7 (human-only,
agents must not work it): the app keeps its named volume, one bind mount puts
finished MP3s under /home/dank/media/podcasts, and deleting that one line is the
whole rollback. Rollback section also covers the full undo, including that the
podman network only disappears once the units are gone.
Deployed 2026-10-18 (e2e 33/33). The boot ledger prints
`reserving the KV pool, 401408 positions (about 11.0 GiB)` and
`62.9 GiB of weights locked … 11.0 GiB of KV pool, 5.7 GiB of working memory,
79.6 GiB in all`, so the ~10.8 GiB that every page quoted from the
~28 KiB/position arithmetic is replaced by the reservation the engine actually
makes (11.0) and its preflight bill (11.3). Host MemAvailable fell 26 -> 22 GiB
across the raise: the FLAGS.md collision (pool RAM is n-gram page cache) paid in
full, so the warm prompt-cache ratio is flagged as the number to watch.
All "committed but not yet deployed" clauses retired: AGENTS.md topology row,
KV-pool law and memory ledger, frank-dragoon.env lore, both index pages, the
iommu page, and the runbook chronology (third production boot recorded).
The sweep also caught the LLM index still describing frank-zealot as
Bonsai-2-27B at 2 x 200,704 in its intro and topology diagram - re-keyed to the
live 9B: 8 full-attention + 24 Gated DeltaNet + blk.32 MTP, 3 x 262,144,
tensor type 141.
Swap the 3090 endpoint from Ternary-Bonsai-2-27B PQ2_0 (2 x 200,704,
fork build 10754) to Ternary-Bonsai-2-9B PQ2_0 with the MTP head IN the
file, three conversations at the trained 262,144 window
(--ctx-size 786432 --parallel 3), on vendor fork build 10777 -- the
minimum build that pairs MTP with a quantized V cache. q8_0 K + q8_0 V
is the only KV that fits: 13,056 MiB against a 26,112 MiB f16 pair.
Settled boot 17,876 used / 6,249 free.
MTP is armed and then DECLINED by measurement: the draft context bills
its own f16 KV over the whole pool (3,072 MiB) and at UBATCH_SIZE=2048
that halves decode (167.2 -> 109.2 t/s). The documented MTP arm
(SPEC_TYPE=draft-mtp + UBATCH_SIZE=1024, 213.5 t/s) is recorded in the
env header rather than left to rediscovery.
Two bugs the trial caught: ${SPEC_PRESENCE_PENALTY:=0.0} was a
default-assignment, so the env value always won and the no-presence-penalty
MTP arm was unreachable (now ${VAR=0.0}); and the Q8_0 mmproj is the
publisher's requirement for concurrent image requests, which matters at
--parallel 3.
contextWindow is now 262144 -- the widest in the fleet -- so :7771 is the
compaction hub. 27B runbook re-banned RETIRED with its bench-window
section closed; indexes, homepage card, OpenWebUI doc and the worklog
re-keyed.
frank-dragoon's total in-flight context was HALF frank-zealot's: zealot runs
one 401,408-position pool split into two 200,704 conversations, while dragoon
ran ONE 200,704-position pool behind its two slots, so a single maximum-length
request filled it and two long conversations had to stay under ~100k each.
HALOGEN_KV_POOL_POSITIONS 200704 -> 401408 = slots x ctx, the top of the seam's
legal band (ctx <= pool <= slots * ctx). The PER-REQUEST ceiling does not move:
HALOGEN_CTX stays 200704, because the checkpoint's native window is 262,144 and
the engine's own 400 text refuses to exceed it without HALOGEN_ROPE_YARN. What
this buys is concurrency, not a longer conversation, which is why the harness
contextWindow (155648 = ceiling - 45056) is untouched.
Cost: ~10.8 GiB of KV where 200,704 cost 5.5 (262,144 cost 7.2; the
2026-10-06 shape's 524,288 measured 14.6), taken from the page cache the 51.2 GB
n-gram table reads through — the warm prompt-cache ratio is the number to
re-measure after the boot. Engine-side total goes 74.1 -> ~79.4 GiB of 121.2.
Also carries scripts/download-frank-zealot-bonsai2-ptq1.sh: the PTQ1_0 candidate
payload fetcher (5,946,648,928 B, LFS sha256-verified, mmproj hardlinked), with
the publisher's own throughput table in its header — the packing is smaller but
the card measures it SLOWER on Ampere, so the swap is decided on our card, not
on the card's A100 row.
Window 01:24-01:35. Byte-identical A/B (rebuild proved it: same 1,287,656 B corpus,
same billed token counts 59,024 / 21,568): cold prefill 1,585.1 -> 1,616.0 t/s
(+2.0 %), warm 208,531 -> 214,131, needle hit/hit both boots. The 13-16 % that
justified amd_iommu=off since 2026-09-11 does not exist on this host; the decode
delta (-9.7 %) is instrument noise -- three repeats of one request on the same boot
gave 31.8/46.9/47.8 t/s.
NPU precondition restored: amdxdna bound 0000:c6:00.1 (driver 0.7.0, firmware
amdnpu/17f0_11/npu_7.sbin), group 26 identity, PASID width 10, /dev/accel/accel0
root:render with dank in render. No endpoint deployed, so frank-npu stays retired
-- only its rationale changes. Gates: 0 failed units, 32 groups (31 identity + the
SD reader's DMA quirk), iGPU c5:00.0 identity with ATS, dragoon ledger unchanged
(74.1 GiB, no clamp, listening at 39.8 s).
Deferred on purpose: the pp_dpm_fclk pin (changes iGPU timing for everyone, nothing
co-runs yet) and PerfOpt (7.0.0-38 has zero perfopt symbols, 26.04 caps at 7.0.0-38.38).
Re-keyed: decision record banner/front matter/verification/revert + new execution
section, frank-npu rationale, AGENTS key lesson, both index pages.
Frank's call the same evening as the native-payload cutover: HALOGEN_CTX and
HALOGEN_KV_POOL_POSITIONS both 262144 -> 200704 (a multiple of 4096, the rung
frank-zealot already runs). The checkpoint's native window is 262,144 and the
engine serves anything below it, so this is a RAM decision, not a capability
limit: the KV reservation drops 7.2 GiB -> ~5.4 (the 2026-10-06 pool measured
14.6), and by the FLAGS.md law -- RAM the pool takes is RAM that n-gram table
loses -- the freed positions become page cache for the 47.68 GiB table that
long prompts read. Engine-side total ~74 GiB instead of 75.8.
The 400-zone law is invariant to the value, so the harness moves WITH the
ceiling: models.yml contextWindow 217088 -> 155648 = ceiling - 45056, keeping
the gap that absorbs the chars/4 under-count (worst-case request ~156.2k vs
the 200,704 ceiling, ~44k margin). Compaction now fires at 85% = 132,300 est
(~136k measured); the 180k target of the 262,144 era is unreachable because
the window itself is 155,648. maxTokens 45056 and the coupled
cap/timeout 65536/7200 are untouched.
Cost stated plainly: with pool == ctx, ONE maximum-length request fills the
pool, so two long conversations never coexist (two of <=100k do). Acceptable
because frank-dragoon carries only the slow role and the security-reviewer
subagent, not the interactive workhorse.
Seam re-keyed: the fit-law test now proves the invariant at the new pair, and
the pool law (ctx <= pool <= slots * ctx) still holds with pool == ctx.
New LLM/frank-dragoon-flash-server-v2-native.md carries the shape with the
three deliberate deviations from the card (served id, SPEC_ADAPT=0, engine
pin 0.16.2), the memory ledger with the PIN_TRUNK rejection, the full
measured A/B table, and the companion law (hardlinked stock peonist files,
engine-tool certified). The swift15 page is re-banned RETIRED 2026-10-06 →
10-17 with its rollback terms; both indexes and AGENTS.md (topology table,
service section, runbook index) are re-keyed, and the root index also stops
describing the halogen-27B page as ACTIVE.
New runbook frank-archon-radiance-qwen38-iq4r.md (publisher container,
payload chain, shape-verbatim cutover record with the boot-ledger and
measurement slots left open for the first boot, rollback). LLM index,
root vault index, and AGENTS.md (fleet table row, section, runbook
index: iq4r ACTIVE, w4dn8 RETIRED/on-disk, fp8-native deleted). Full
suite green.
The 2026-10-09 wire aliases become the service names everywhere:
systemd units, serve scripts, env files and FRANK_* variables,
deploy-plan/deploy-remote service keys, container names, Prometheus
job names + alert rules, Grafana dashboard queries, bats seams,
runbook filenames, homepage/AGENTS lore, and historical prose.
Rollback tags renamed in lockstep. Residue in the tree: zero.
Retired frank-medium/small/npu names are history and stay.
Known casualty: scripts/patches/*.patch context lines were rewritten;
those retired-era vLLM-fork patches no longer apply byte-exact.
Capacity trade: the 9B dense-hybrid (qwen35: 32 layers = 8 full-attention
+ 24 Gated DeltaNet, 4 KV heads x 256 dim, native 262,144, MTP-less,
natively multimodal) costs ~7 GiB of weights where the 26B MoE cost
~16.6 GiB. The freed room buys back the second conversation slot and
exact KV: --ctx-size 401408 --parallel 2 (two 200,704 windows), f16 K +
f16 V, MTP off (the file carries no nextn/mtp tensors - flag-only enable
is impossible), vision via the publisher's BF16 mmproj, ubatch 1024
(the CUDA decode-optimal value), verbosity 4, publisher sampler recipe
(0.6/0.95/20/0) in models.yml.
Hosts with the MiMo-V2.6-Distill Q6_K boots at this geometry: 21,276 MiB
used / 2,849 free at ubatch 1024, 40,960 tokens of fp16 KV at ubatch
2048. Compaction stays routed here; window math is unchanged (maxTokens
16,384 keeps the worst-case harness request inside the slot).
Gemma4 payload (Q4_K_M + mmproj + MTP assistant) stays on disk as the
rollback; its runbook is retired in place, keeping the attention_k_eq_v
pair-billing law.
Co-Authored-By: omp <omp@franksdank.cloud>
Payload: HauhauCS/Gemma4-26B-A4B-QAT-Uncensored Q4_K_M (16,796,015,520 B)
+ the publisher's BF16 mmproj, /home/dank/models/Gemma4-26B-A4B/.
Shape: ONE 200,704-token slot (--parallel 1; 2x200,704 does not fit the
24 GiB card), f16 K + q8_0 V under flash attention, batch 4096 / ubatch
1024, --log-verbosity 4, MTP OFF (the gemma4-assistant file is on disk,
not loaded: its draft KV is built against the target's full window,
measured +1.9 GiB -> 23,806/319 free, no margin beside the f16-K raise).
Measured boots (nvidia-smi): q8_0 K + q8_0 V / ubatch 2048 / MTP off =
21,890 used / 2,235 free; with the MTP assistant = 23,806 / 319.
Engine facts recorded: gemma4 hybrid attention (25 sliding-window + 5
full-attention layers) so KV scales with 5 layers only, and
attention_k_eq_v is honoured by build 5fc4f3c (pool bills K-only; the
measured q8_0 pool = 2.05 GiB = the K-only figure).
Serve script gained VERBOSITY (+ --log-verbosity in the argv), the
SPEC_DRAFT_FILE drafter path, and defaults/lore retargeted to Gemma4.
deploy-remote.sh payload gates follow the env. The frank-secondary bats
seam is rewritten for the new contract (argv-based) and fixes a
pre-existing vacuity: serve_argv regenerates the test env, so
mutations made before it were silently undone - tests that mutate now
exec through run_stage. frank-secondary is the fleet compaction hub
(200,704 is the only window that fits a worst-case compaction input).
Execution facts, blocker trail (plugin-tap probe law, weights-slab
formula, KLD_MOUNT), engine verdict, and the post-chain order. Linked
from the docs index per the completeness seam; the cutover runbook
replaces it.
23,120 used / 1,005 free at boot, 763 settled after a vision probe
(encoder buffers resident). Rung cost 746 MiB, not the transferable
616 prediction. Prefill baseline 1,209 t/s @ 3,760 recorded; decode
unchanged as the law says. 2048 = ceiling for this shape.
Cross-cutting rename of the current llama.cpp endpoint on the RTX 3090
(port 7771) from frank-small to frank-secondary:
- Service trio: scripts/{frank-small.env,frank-small.service,
serve-frank-small.sh,download-frank-small.sh} → frank-secondary.*
- Test: tests/frank_small_unit.bats → tests/frank_secondary_unit.bats
- Runbooks: serverAppConfigs-docs/LLM/frank-small-*.md →
frank-secondary-*.md (9 files); the FreeToken runbook stays
freetoken-frank-small.md (it documents the retired CUDA-era service)
- Identifiers: FRANK_SMALL→FRANK_SECONDARY, llm_small→llm_secondary,
LLM_SMALL_*→LLM_SECONDARY_*, deploy_llm_small→deploy_llm_secondary
- Historical references to the OLD frank-small (FreeToken/CUDA on the
3090, and the Dankserver service that became frank-medium) are
preserved verbatim — they describe what the service was called
at the time.
The writer's 31b1ec3 added LLM/frank-large-flash-server-v2-swift15.md
but did not wiki-link it from index.md, which failed the docs-index
seam (test 135). Add the link as the new ACTIVE frank-large runbook
(supersedes halogen 0.1.4).
Runbook reshaped: 2 x 131072 no-spec identity, slot-2 arithmetic (drafter ~0.4 GiB + ubatch ~0.3 GiB bought the second 4,352 MiB KV slot), ngram-mod skipped on the recurrent-rollback law, MTP references kept (44.8-52.1% acceptance, 58.9-66.9 t/s), rollback = one revert of 812d9eb. AGENTS fleet row + section, LLM/index.md intro + entry, root index entry all flipped to the 2026-10-06 shape. 360/360 suite green; e2e gate 33/33 green at deploy.
The parallel session's uncommitted swap, committed at Frank's direction
(commit whatever's in the working tree):
- llama.cpp master on the RTX 3090 now serves ukisai's
Swift-1.5-Qwen3.8-27B GSQ-RCO IQ3_S WITH the in-file MTP head, ONE
131,072-token slot at Frank's 128k cap, q8_0 K + q8_0 V, vision through
the base-model projector; MiMo-V2.6 Q6_K (2x200704 f16 KV) retired to
rollback target.
- env/serve/download + seam twins updated together (deployer payload gates
rode the previous commit's deploy-remote.sh).
- docs: new swift15 runbook; MiMo Q6K runbook retired; AGENTS.md topology
and both vault indexes follow (root index ACTIVE/RETIRED flip, test 135
seam green).
- comfyui: two qwen-image t2i workflow drops under dankclaw/comfyui/
workflows/ (12 KiB).
The parallel session's stash pop left GGUF Compatibility, Multi-Device
Splits, Serving Sizing, PCIe Link and Atomics, RDNA4 Container Serving,
vLLM Porting triage and RDNA4 MoE Budget unindexed and carrying 9
unresolved wikilinks (LLM Bench Windows / Model Swap / AGENTS targets
that never existed, a [[nodiscard]] C++ attribute parsed as a wiki link,
and two multi-line-wrapped link tokens). All repointed to real pages;
six rows added to the vault index; seam green.
Both indexes stated the pre-doubling VRAM as current. Now 2,279 MiB free with
the 4096/2048 pair named, so no index contradicts the runbook it links to.
- new active runbook frank-small-llama-mimo-q6k-2x200704.md: 21,276 used /
2,849 true free (inside the 2-3 GiB floor), decode 94.44 t/s vs 104.84 on
Q5_K_M - the rung costs a measured -10 % (weight-bandwidth-bound card),
512px red-PNG vision verified, and the new probe-size law (>=256 px: a
10x10 image undersupplies image tokens and the model answers 'video'),
nvidia-smi free-field law (driver holds ~450 MiB the subtraction hides)
- frank-small-llama-mimo-q5km.md retired; it keeps the boot-pair table and
becomes the Q5 rollback reference (3,571 free, 104.84 t/s measured)
- AGENTS.md fleet row, endpoint section, --parallel lesson, runbook index
(active row + two retired rows); LLM + root index entries
- Q6_K GGUF-vs-GGUF diff proven: template and all 46 KV keys byte-identical
to Q5_K_M - the vision probe scare was probe size, not payload
- new active runbook frank-large-rocmfpx-q6agent-uncensored-mtp.md: the
Q6_0_ROCMFPX_AGENT file (ftype 114, 339 type-102 + 159 type-103 tensors),
27.6-28.1 t/s at 86.8% draft acceptance, the measured ngram-mod chain
rejection (-15% prose: recurrent-state rollback disabled), the
bench-window hardening laws
- frank-large-rocmfpx-q8agent-mtp.md retired; rollback section corrected
(gufo payloads removed from disk 2026-10-03 - that rollback is now a
re-download; the aligned Q8 file is the direct one, still on disk)
- frank-small-llama-mimo-q5km.md reshaped to 2 x 200704 with fully
unquantized f16 KV: boot-pair table gains the measured row (20,530 used /
4,046 free), 104.84 tok/s, slot-window contract at 200704, rollback to
the 3x180224 + q8_0-V shape documented
- AGENTS.md: fleet table rows, both endpoint sections, the chain law
extended with the recurrent-model measurement, the --parallel law with
the reslice, runbook index (active + retired)
- LLM/index.md + root index entries updated
Runbook, both indexes, AGENTS.md (topology row, endpoint section, three new
Key Lessons, one stale SGLang label in the bench list) and the bundle log now
describe the deployed shape: pool 540,672, 180,224 tokens per conversation,
f16 K + q8_0 V, mmproj vision, 21,684 used / 2,441 free of 24,576 MiB, 106.8
tok/s decode, 3,814 t/s prefill.
The claims that mattered:
- three contexts buy 180,224 tokens per conversation, not the native 262,144,
and a longer prompt is a hard 400 - so models.yml contextWindow is 180224
and frank-primary's 183,501-token compaction request no longer fits here
(forward pointer added to the retired SGLang runbook, which asserted it)
- the vision tower bills ~1.4 GiB for a 918 MB file, established by subtracting
two measured boots, not by a formula
- /v1/chat/completions ignores the legacy top-level 'images' field: the model
invents a color; only the image_url content part reaches it
frank-large: new runbook frank-large-rocmfpx-q8agent-mtp.md — the ftype-115
tensor-type lock-in that forced the fork, the pinned build (set_rows history),
the measured MTP A/B (17.0 vs 7.6 t/s decode, acceptance 0.462), the live
client contract (reasoning_content, effort fold, vision restored) and rollback.
frank-small: new runbook frank-small-llama-mimo-q5km.md; q6k marked retired
(rollback target). gufo runbook marked retired (rollback target). Both indexes,
the fleet topology table, the vision/models.yml notes, the benchmarking law and
three new Key Lessons updated.
The first crash-looping boot (51 restarts): serve received the host path
/home/dank/models/... but the models tree is mounted at /models inside the
container, so transformers failed 'Repo id must be in the form
namespace/repo_name' before the engine even built. Same conversion the spec
draft already used. Pinned by bats. Root docs index links the new runbook
(seam required it) and the work log records the crash-loop plus the 52 G
payload reclaim.
Replaces llama.cpp Gemma4-12B-QAT (payload + binary retained as rollback).
SGLang in a self-contained python3.12 venv with an extracted CUDA 13.0
nvcc toolchain for the sgl-kernel/flashinfer JIT.
Shape: fp8 KV, pinned 606,208-token pool / 4 x 151,552 slots, bf16 mamba
states pinned at 32 slots (the engine autofills the ssm pool to a fixed
~5.1 GB regardless of dtype), eager prefill (no headroom for graph
capture), no spec decode (the checkpoint ships no mtp.* tensors despite
mtp_num_hidden_layers=1 — NEXTN drafts with an uninitialised 4.24 GiB
ghost at 0% acceptance).
Verified: 83.9 tok/s decode, 170k single prompt, 4 concurrent 54k
requests, vision (in-checkpoint tower), reasoning_content field.
Payload: Intel/GLM-5.3-Flash-W4A16-AutoRound (AutoRound symmetric int4 group
128, auto_round:auto_gptq, 34 shards + 2 extras, 113,074 tensors, ~170 GiB).
It is the only W4A16 checkpoint this fork can serve with expert offload: its
quant_method "auto-round" is overridden to "inc", which routes the MoE to
libr4d's grouped W4A16 GEMM (R4dW4a16MoEMethod). A compressed-tensors W4A16
build of the same model resolves to TritonWNA16Experts, which has no offload
recipe, and refuses to serve at all (170 GiB of experts cannot fit 64 GiB of
VRAM, so offload is mandatory).
Shape: tcclaviger RDNA4 vLLM fork (same digest-pinned image as the 2026-09-29
qwen cutover), TP=2, 131,072 ctx x 2 seqs, bf16 KV (fp8 is ILLEGAL here — it
disables the sparse MLA backend's rope-free Triton path, the only MLA decode
gfx12 has), util 0.98, NBATCH 2048, --enable-expert-offload
--expert-precise-memory, prefix caching, glm47 tool + reasoning parsers,
MTP and expert prefetch absent by measurement.
Measured: decode 9.9-10.0 tok/s, prefill 918 tok/s (warm), arena 14.51 GiB =
2,383 slots/rank over 12,096 host-backed experts (73.64 GiB pinned/rank).
THE 35 tok/s TARGET IS NOT MET and the shortfall is measured: live telemetry
(R4D_LRU_TELEMETRY=1) gives a 0.53-0.66 per-layer hit rate, a step moves
0.82 GiB of misses in ~100 ms = 8.2 GB/s effective against a link measured at
22.74 GB/s — the same structural gather-path gap the qwen campaign found.
Growing the arena 13.08 -> 14.51 GiB changed decode by nothing.
Rejected by measurement: MTP4 (head loads, acceptance 3.76-4.90, but decode
unchanged and prefill collapses 918 -> 255 — a verification forward gathers the
union of the batch's experts), NBATCH 8192 (identical prefill).
Also: scripts/patches/gpu_worker-kv-cap.patch (fork-local fix for a KV-reserve
model that under-counts GLM's per-sequence KDA state; not needed by this shape),
runbook, AGENTS.md row + section, both indexes, work log.
Rollback: 2339fae (qwen FP8 + MTP4).
gufo is a vertical inference engine built for exactly this hardware
(Ryzen AI MAX+ 395 / Radeon 8060S, gfx1151), and Dankclaw's frank-large
has swapped runtimes seven times in search of a better one. This one
ships no prebuilt Linux binary, so the runtime is BUILT FROM SOURCE on
the host and the payload, port, alias and client contract stay exactly
where they are.
Runtime: gufo v0.1.1 (b0f8673) at /home/dank/GitPackages/gufo-v0.1.1,
built by the new scripts/build-gufo.sh. It links the host ROCm 7.2.4
through its RPATH — the opposite of the unsloth prebuilt it replaces,
which carried its own — so two host workarounds are encoded in that
script: Ubuntu's librocwmma-dev installs a truncated rocWMMA header set
that shadows AMD's and cannot be overridden with -I in HIP mode, and the
ROCm clang prefers Ubuntu 26.04's GCC 16 headers over the GCC 15 whose
libstdc++ is actually installed.
Shape is unchanged: one session, full 262144 ctx, the same UD-Q8_K_XL
target and the same DFlash2 drafter, greedy, text-only. Speculation
moves from llama.cpp's fixed depth 4 to gufo's adaptive block policy
capped at 7 proposals.
Client-contract deltas, all documented: the chat template is compiled
into the binary (no --chat-template-file, so the Qwen-Sharp 73k-token
failure mode is gone), reasoning_effort now takes the full OpenAI set
(so models.yml's reasoningEffortMap fold is no longer required — kept
anyway, so this commit does not also move every high role from xhigh to
high effort), and there is no KV-quant or prompt-cache-RAM flag.
The test seam now runs the serve script against a stub engine and
asserts on the argv it actually execs, instead of grepping the script's
text. The deploy handler gates on the runtime binary and returns before
it stops the serving unit.
Measured trade, from gufo's own Q8_K_XL tables against llama.cpp
b11069: prefill +37% to +80%, cold load 4.5x faster, decode +7% at zero
depth — but -14% at 65,536 tokens of depth and -24% at 131,072. On the
8-bit target the crossover is ~32k, and frank-large runs one 262,144
conversation. The Q4_K_XL target shows no such regression.
Answers Frank's question: given the 2x R9700 box, which of DeepSeek-V4.1-Flash,
MiMo-V2.6-Flash-MOPD and GLM-5.3-Flash runs best, on what runtime, and can
Affinity / vllm-radiance / FreeToken-AMD be modified to run them fast.
Sizing, measured from the HF API rather than the model cards:
- DS4.1 payload 475.27 GiB, of which 189.13 GiB is Engram (two 94.56 GiB
shards, the layer 1 and 14 lookup tables). Lucebox never maps them, so
they do not evict the expert page cache.
- 240 routed experts x ~11 MB per token at the 37-of-384-per-layer residency
the hot stack can hold = ~2.4 GB across PCIe per token. At 55 GB/s real on
Gen5 x8 that is a ~23 tok/s ceiling, under the 35 tok/s goal before any
attention or dense work.
- The backend refuses layer-major prefill for V4.1, so the only legal
prefill is token-by-token; 1000 t/s is ~20x away.
Affinity (Codeberg StillDeadcode) builds clean here (hipcc 23, gfx1201, zero
errors) and claims 110 tok/s for the 284B V4-Flash, because the 101 GiB
container leaves the cold expert set in a pinned host pool in RAM. V4.1 is
2.5x too big to repeat that. Port scope, from both checkpoints' real
safetensors indices: expert machinery ports unchanged (identical tensor
naming, 15,360 tensors at 40x384), 26/32 per-layer tensors and 34/35 config
keys present. Five gaps - Engram, the disjoint {0,1,2} vs {0,4,128} ratio
vocabulary, CED, the DSpark draft geometry, vision. num_hash_layers is the
only key V4.1 lacks, so model.h:36's default of 3 silently survives and
would route layers 0-2 through a table that does not exist.
Also corrects: Dankserver RAM 60 -> 246 GiB (stale since the 2026-09-20 board
swap) and storage; the top-level index credited V4.1 with V4's geometry
(43 layers / 256 experts vs the actual 40 / 384).
New ACTIVE runbook (setup/doctor PASS=34/FAIL=0, UVA transport verified
in serving workers, wire-verified alias/reasoning/metrics), index +
vault entries flipped: llama.cpp b11030 endpoint RETIRED 2026-09-27 to
rollback target. Topology diagram updated.
The 2026-09-26 same-day cutover that ended the Lucebox DS4.1 detour:
Dankserver's :7770 endpoint is now unsloth llama.cpp (the b11030-mix
prebuilt, ROCm gfx120X, bundled runtime) serving
unsloth/Qwen3.8-Flash-Next-GGUF at Q8_0 — 175.30 GiB across six sha256-
verified split parts — on both Radeon AI PRO R9700s, as a ROOT unit (the
frank-large shape: no container, no generated unit).
The deployed contract:
- Placement law: layer split -ts 1,1 with -ngl 999; the routed experts
and the 51B-param n-gram (PLE) table are COMPUTED in host RAM via
-ot exps=CPU,ngram_embeddings=CPU (the x4-pair measurement: streaming
over PCIe loses at 3.7 t/s, computing in RAM does not care).
per_layer_token_embd.weight=CPU is required by the shared MTP head's
borrow path (a GPU-resident borrow crashes the draft load).
mmap stays on; --cache-ram 16384.
- Spec chain draft-mtp,ngram-mod: the shared Q8_0 MTP head
(mtp-...-shared-Q8_0.gguf, 2,786,568,256 B) at depth 2, with the
draftless ngram-mod matcher chained behind it — the only legal way to
layer two spec implementations in one process.
- Vision explicit: --mmproj names the BF16 projector (this build never
autoloads; a server without it comes up healthy and text-only).
- Reasoning: --reasoning-format deepseek → reasoning_content; the
Qwen3.8 template's {xhigh, medium, low} reasoning_effort gate folded
by reasoningEffortMap in models.yml.
- One stream, the model's full 262,144 context; greedy server default
with Qwen-recommended sampling riding compat.extraBody.
The deployer's deploy_llm_primary is rewritten for the root unit: it
gates every artifact at its declared byte count (stat -c %s, never wc -c
— which read the whole 377 GiB lucebox file for minutes on 2026-09-26),
retires the lucebox user quadlet to dated home trash, waits out port
7770, relinks/enables/restarts the root unit and waits up to 1800 s for
/health. The quadlet-era tree under dankserver/frank-primary/ is
archived (dankserver/.AGENTTRASH/20260926-llamacpp-cutover/ on this
box; git history everywhere else) and its planner mapping replaced by
the scripts/ trio.
Observability flips back to the llamacpp namespace: a frank-primary
Prometheus job at 192.168.0.235:7770 (cross-host LAN — the netavark
published-port quirk that forces host.containers.internal for the two
Dankclaw endpoints is same-host only), the LlmEndpointDown runbook
text, and the dashboard description for three llama.cpp endpoints.
Docs: the new runbook (frank-primary-llamacpp-qwen38-flash-next-q8.md),
RETIRED banners on the lucebox and radiance runbooks and their index
entries, the AGENTS.md fleet section rewritten for the root-unit shape,
and the homepage frank-medium card renamed to frank-primary with the
Q8_0 Flash-Next spec (docs.html #models and the modality note updated:
images on frank-primary and frank-small, audio on frank-small only).
models.yml: the dankserver provider is "llamacpp (dankserver)" serving
frank-primary, contextWindow 262144, maxTokens 65536 (compaction-gap
law: 183,501 + 65,536 <= 262,144), input [text, image].
Suite: 346 green (frank_primary_unit.bats rewritten for the root unit —
16 tests pinning the placement law, the shared-head requirement, the
explicit projector, HIP env hygiene (BIND_HOST, never bare HOST), and
the deployer's byte-count gates; the quadlet-era planner and
changed-tests mappings rewritten for the trio; the observability scrape
test asserts three llamacpp jobs again).
Rollback: the lucebox image (localhost/lucebox-rocm:gfx1201) and the
378 GiB payload stay on the host; the Qwen payload stays fetched, so a
re-swap in either direction is a redeploy, not a refetch.
Replaces frank-medium (vllm-radiance TP=2) on :7770. The 377.2 GiB
43-layer MoE (256 routed experts, top-6) is mmapped: dense work and the
MLA KV cache on the R9700, routed experts computed on CPU from the
page cache (246 GiB RAM). The unit bypasses the image entrypoint
(Qwen/DFlash-only flag set) and execs luce_server with the DS4 launch
shape; --ds4-fused-decode's monolithic-load fallback IS the serving
mode here. deploy_llm_primary gates on image + model byte count,
stops/masks the radiance unit, waits for :7770, installs and starts.
Runbook: frank-primary-lucebox-ds4.md.
Frank
