One fact, one owner — de-stale the technique and era pages (clusters C1-C6)
Docs-only follow-up to #23. After the frank-dragoon v2-abliterated cutover and the frank-zealot Bonsai-2 PQ2_0 cutover, the vault's technique and era pages still carried
fleet facts that the endpoint pages own, and several era pages still read as live state.
This branch fixes both classes, cluster by cluster.
What moved
Cluster
Pages
Change
C1
Quantization Production, Quantization by Hardware Architecture, GGUF Compatibility, Model Swap and Rollback
de-staled: retired-payload claims re-keyed to their era, live shapes replaced by links to the endpoint pages (a518fc7)
C2
RAM CPU Offloading in VLLM + 5 restatement sites
the --swap-space → --kv-offloading-size claim had six holders, two with their own copy of the upstream citation; it now lives only in RAM CPU Offloading in VLLM, which carries today's verification against vllm/config/cache.py (500ac2b)
C3
Running Qwen3.6 27B FP8 on Dankclaw, ling-3.0-flash-vllm-plan, frank-redo-2026-08-17
the 27B FP8 page had no banner and stated "Dankclaw's primary service now runs vLLM-ROCm with Ling CIRU INT4" — false by three fleet generations; the redo banner stopped restating (and mis-stating) live shapes (73fb272)
C4
FastFlowLM NPU Inference
host-RAM figures restated three times → [[Dankclaw System Specs]]/[[Dankserver System Specs]]; the solo-held 14-18 t/s PP rate kept and tagged to the 2026-08 era; "frank-zealot now runs the NPU endpoint" era-tagged (e32df1b)
C5
frank-medium-radiance-mxfp4, frank-medium-model-swap, both index pages
retired endpoints advertised as "ACTIVE" runbook / live swap procedure, in frontmatter and in the hand-copied index rows; both sites re-keyed (6a8cbde)
C6
dankserver-vram-growth
banner disambiguates the 2026-08 frank-zealot (llama.cpp on the 3090 while that card was in Dankserver) from today's same-named endpoint on Dankclaw, so a retired crash signature cannot be read as a live fault on a same-card-class service (500ac2b)
What stayed, and why
Owned ledgers are not duplication, so they were kept deliberately:
Ethernet Bandwidth Math — 457.9 t/s, the ping-RTT table, the per-link cost model: solo-held, it is the page that measured them.
x4-pair-experiment — the cross-device rate table (CUDA+Vulkan split 27.2 t/s / 268.5 prefill, RCCL tensor 19.09, layer split 18.8 / 1991, PP 16.66) and the 6.72 GB/s x4 link figure: no other page restates them (verified by grep across the vault).
Dual-R9700-Cooperation — the R4D-vs-RCCL verdict and board-swap evidence matrix.
Era pages (ling-3.0-flash-trial, Running Qwen3.6 35B A3B FP8 on Dankserver, dankserver-vram-growth, the NPU runbook, Pipeline Parallelism/*) — retired payloads are an era page's content, not staleness. They needed banners and dates, not deletion.
Historical command examples that carry the dead --swap-space flag inside dated era sections: kept as era records, labelled as such.
Verification
./tests/run.sh → ALL TESTS PASSED, 448 assertions, wall 86.49 s (includes the
docs-completeness and vault link-integrity seams).
Independent census: 160 pages, 0 orphan pages (every page reachable from an index),
0 broken wikilinks (the one regex hit is an ANSI-escape sequence inside inline code in vLLM Boot Forensics, not a link).
The vLLM knob claim was re-verified against the primary source today, not inherited from
the pages that asserted it.
Documentation-only: the deployer plans no service for these paths.
Method note
Three of the four research subagents dispatched for this campaign returned unusable or
inverted verdicts — one classified era pages as stale because they document retired
payloads, one reported files that do not exist, one claimed duplications that grep disproved
(457.9 t/s, the RTT table), and one fabricated a line number in a page shorter than the
line it cited. Every change in this branch is parent-written and parent-verified against the
files and the live fleet state; scout output was used only as a lead.
# One fact, one owner — de-stale the technique and era pages (clusters C1-C6)
Docs-only follow-up to #23. After the `frank-dragoon` v2-abliterated cutover and the
`frank-zealot` Bonsai-2 PQ2_0 cutover, the vault's technique and era pages still carried
fleet facts that the endpoint pages own, and several era pages still read as live state.
This branch fixes both classes, cluster by cluster.
## What moved
| Cluster | Pages | Change |
|---|---|---|
| C1 | `Quantization Production`, `Quantization by Hardware Architecture`, `GGUF Compatibility`, `Model Swap and Rollback` | de-staled: retired-payload claims re-keyed to their era, live shapes replaced by links to the endpoint pages (`a518fc7`) |
| C2 | `RAM CPU Offloading in VLLM` + 5 restatement sites | the `--swap-space` → `--kv-offloading-size` claim had **six** holders, two with their own copy of the upstream citation; it now lives only in `RAM CPU Offloading in VLLM`, which carries today's verification against `vllm/config/cache.py` (`500ac2b`) |
| C3 | `Running Qwen3.6 27B FP8 on Dankclaw`, `ling-3.0-flash-vllm-plan`, `frank-redo-2026-08-17` | the 27B FP8 page had **no banner** and stated "Dankclaw's primary service now runs vLLM-ROCm with Ling CIRU INT4" — false by three fleet generations; the redo banner stopped restating (and mis-stating) live shapes (`73fb272`) |
| C4 | `FastFlowLM NPU Inference` | host-RAM figures restated three times → `[[Dankclaw System Specs]]`/`[[Dankserver System Specs]]`; the solo-held 14-18 t/s PP rate kept and tagged to the 2026-08 era; "`frank-zealot` now runs the NPU endpoint" era-tagged (`e32df1b`) |
| C5 | `frank-medium-radiance-mxfp4`, `frank-medium-model-swap`, both index pages | retired endpoints advertised as "ACTIVE" runbook / live swap procedure, in frontmatter **and** in the hand-copied index rows; both sites re-keyed (`6a8cbde`) |
| C6 | `dankserver-vram-growth` | banner disambiguates the 2026-08 `frank-zealot` (llama.cpp on the 3090 **while that card was in Dankserver**) from today's same-named endpoint on Dankclaw, so a retired crash signature cannot be read as a live fault on a same-card-class service (`500ac2b`) |
## What stayed, and why
Owned ledgers are not duplication, so they were kept deliberately:
- `Ethernet Bandwidth Math` — `457.9 t/s`, the ping-RTT table, the per-link cost model: solo-held, it is the page that measured them.
- `x4-pair-experiment` — the cross-device rate table (CUDA+Vulkan split 27.2 t/s / 268.5 prefill, RCCL tensor 19.09, layer split 18.8 / 1991, PP 16.66) and the 6.72 GB/s x4 link figure: **no** other page restates them (verified by grep across the vault).
- `Dual-R9700-Cooperation` — the R4D-vs-RCCL verdict and board-swap evidence matrix.
- Era pages (`ling-3.0-flash-trial`, `Running Qwen3.6 35B A3B FP8 on Dankserver`, `dankserver-vram-growth`, the NPU runbook, `Pipeline Parallelism/*`) — retired payloads are an era page's *content*, not staleness. They needed banners and dates, not deletion.
- Historical command examples that carry the dead `--swap-space` flag inside dated era sections: kept as era records, labelled as such.
## Verification
- `./tests/run.sh` → **ALL TESTS PASSED**, 448 assertions, wall 86.49 s (includes the
docs-completeness and vault link-integrity seams).
- Independent census: 160 pages, **0 orphan pages** (every page reachable from an index),
0 broken wikilinks (the one regex hit is an ANSI-escape sequence inside inline code in
`vLLM Boot Forensics`, not a link).
- The vLLM knob claim was re-verified against the primary source today, not inherited from
the pages that asserted it.
- Documentation-only: the deployer plans no service for these paths.
## Method note
Three of the four research subagents dispatched for this campaign returned unusable or
inverted verdicts — one classified era pages as stale *because* they document retired
payloads, one reported files that do not exist, one claimed duplications that grep disproved
(`457.9 t/s`, the RTT table), and one fabricated a line number in a page shorter than the
line it cited. Every change in this branch is parent-written and parent-verified against the
files and the live fleet state; scout output was used only as a lead.
---
cwd: /home/blank/.omp/wt/wt-20261017-181502-6d1fc1f · session: 01a152fe-c490-702c-8280-c29b0bad4066
Enforces "one fact, one owner": technique pages keep mechanism, laws, traps
and verdicts; per-endpoint shapes, digests and era measurements are cited to
the endpoint page that owns them.
Also repairs the four pointers the pass left empty (each verified against the
cited owner before landing):
- EMULATION dequant doubles host RAM (118 GiB -> ~238 GiB) -> [[frank-archon]]
- 12.11 tok/s flat 304 -> 16,095 tokens -> [[Tcclaviger Fork Ops]]
- bundled-head acceptance ladder 0.739/0.554/0.427/0.340 -> [[Qwen3.8 PP Deployment]]
- llama.cpp PR #23897 dedup provenance -> [[qwen3.8-spec-matrix]]
Replaces the retired-shape illustration in Llama Cpp Serving Sizing with the
owned MiMo Q5_K_M row (6.40 GiB payload -> 20,530 MiB used), and unwraps a
hard-wrapped [[Dankserver System Specs]] link that failed link integrity.
Orphan census over the diff: 0 numbers lost (comma-insensitive match against
the whole vault plus AGENTS.md).
The 3-slot -> 2-slot reslice entry and the contended-rate entry carried the
full measurement ledger; the numbers belong to [[frank-zealot]] (shape block +
Eras rows), so the entry keeps the causal story (pool halving funds exact KV;
rates taken under a live consumer are floors) and links for the figures.
The dedupe passes kept trying to port single-owner experiment figures into
endpoint Eras tables. The law now states the test explicitly: two pages
carrying one figure is the defect; a figure with exactly one holder is fine
wherever it lives.
The page claimed the unit had been failed since 2026-10-18 03:40 with a
hand-run process on the port. Verified live: both units are active under
their repo units (zealot since 21:06, the deployer's 2-slot reslice).
Parent-written (scout summaries were corroborated against the files, not
trusted): every restatement of an endpoint shape now cites the endpoint page,
and four claims contradicted by today's fleet state are corrected.
- Quantization Production: the challenger-contract line quoted the retired
compaction-hub window; now cites [[frank-zealot]]'s Current shape. The 3090
worked example is era-tagged. Sources listed [[frank-dragoon]] three times.
- Quantization by Hardware Architecture: said the 3090 lives on Dankserver
(it moved to Dankclaw 2026-09-16; the Eras table owns the move) and ranked
the retired Ling-3.0 Quark INT4 service as the ACTIVE Dankclaw workload. The
active service is [[frank-dragoon]] on a closed `.hgn` format.
- GGUF Compatibility: Related claimed the ftype-115 ROCmFPX file was "in
production"; that is an era now, so it points at [[kingjones-q8agent-rocmfpx]].
- Model Swap and Rollback: dropped the fragment byte-count ledger (AGENTS.md
owns it), the Ampere FP8 hardware claim ([[Quantization by Hardware
Architecture]] owns it), halogen's HTTP surface ([[frank-dragoon]]'s wire
contract) and three brittle unit-file line refs (accepted loss: line numbers
are derivable, the root-unit law stays). Fixed the per-slot example, which
described the retired 3-slot arithmetic, and a Related bullet that called
[[frank-archon]] retired while linking the same page twice.
Census over the diff: 1 orphan line, the deleted unit line-number refs above.
One fact, one owner (de-dup campaign C4):
- `FastFlowLM NPU Inference.md` carried the host RAM figures three times
(29 GiB budget input kept as era data; the 2026-10-18 `free -g` re-reads
of 121/246 GiB cut -> [[Dankclaw System Specs]]/[[Dankserver System Specs]]).
- The 14-18 t/s PP-split decode is this page's own solo-held era measurement:
kept, tagged to the 2026-08 era with [[frank-redo-2026-08-17]].
- Promotion section: `frank-zealot` "now runs the NPU endpoint" era-tagged
(the name has been reused twice since; the NPU unit is `frank-npu`, masked).
- Repaired the Test Runbook splice introduced by the range edit above.
Refs: vault de-dup campaign, cluster C4.
- `Running Qwen3.6 27B FP8 on Dankclaw`: had no banner at all and opened as a
live how-to. Added the RETIRED banner (iGPU endpoint today is
[[frank-dragoon]]), era-tagged the hardware table, and rewrote the
"primary service now runs vLLM-ROCm with Ling CIRU INT4" production note,
which was false by three fleet generations.
- `ling-3.0-flash-vllm-plan`: "frank-dragoon is the Qwen3.8-27B PP head" ->
past tense with the 2026-08-17 redo as the anchor and the live owner named.
- `frank-redo-2026-08-17` banner: stripped restated live shapes that had
already drifted (radiance 1.0.13, PrismML b10754 - the fleet runs 10777)
and the 3090-provenance clause owned by the PP deployment page; banner is
now locator-only (names + ports + links).
Verified before cutting: 457.9 t/s and the ping-RTT figures are solo-held by
Ethernet Bandwidth Math.md (no duplicate), so both scout-reported "duplicates"
were kept.
Refs: vault de-dup campaign, cluster C3.
The root and LLM index pages still advertised two retired endpoints' pages as
live procedure:
- `frank-medium-radiance-mxfp4`: frontmatter said "ACTIVE frank-medium
runbook" and the body opened "The ACTIVE frank-medium deployment". Both
re-keyed to the era it ran (2026-09-21 -> 10-10), with [[frank-archon]]'s
Eras table named as the owner of that shape.
- `frank-medium-model-swap`: read as live how-to for a retired endpoint.
Re-keyed to historical, pointing at [[Model Swap and Rollback]] for the
current swap law.
- Both index pages' rows re-keyed to match (rows are hand-copied from
frontmatter - no generator - so both sites move together).
Refs: vault de-dup campaign, cluster C5.
The "`--swap-space` is removed; use `--kv-offloading-size`" claim was carried by
six pages, two of them with their own copy of the upstream source citation. It
now lives only in [[RAM CPU Offloading in VLLM]], which carries the 2026-10-19
verification against `vllm/config/cache.py` (no `swap_space`; `kv_offloading_size`
activates only with `kv_offloading_backend`):
- `VLLM Overview` (two argument tables), `VLLM Index` (parameter table),
`DFlash DSpark Speculative Decoding`, `Shared Memory Architecture Dankclaw`,
`Hybrid CPU GPU Processing Dankserver` -> cite the owner.
- `VLLM Index`'s example command used the dead flag; it now shows the live pair.
- `dankserver-vram-growth`: banner disambiguates the 2026-08 `frank-zealot`
(llama.cpp on the 3090 when that card was in Dankserver) from today's endpoint
of the same name on Dankclaw, so a retired crash signature does not read as a
live fault on a same-card-class service.
Refs: vault de-dup campaign, clusters C2 and C6.
The 2026-10-19 vault de-dup campaign sent four read-only scouts over the same
pages and had to parent-verify every claim: one inverted era-page law (flagged
pages stale for documenting the retired payloads they exist to record), one
listed files that do not exist, one reported duplications grep disproved, one
cited a line past the end of the file it judged. Record the rule beside the
other evidence laws so the next agent does not integrate a yield unchecked.
Refs: vault de-dup campaign, PR #24.
dank
merged commit a5fe8db877 into main2026-10-19 02:06:12 -07:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
One fact, one owner — de-stale the technique and era pages (clusters C1-C6)
Docs-only follow-up to #23. After the
frank-dragoonv2-abliterated cutover and thefrank-zealotBonsai-2 PQ2_0 cutover, the vault's technique and era pages still carriedfleet facts that the endpoint pages own, and several era pages still read as live state.
This branch fixes both classes, cluster by cluster.
What moved
Quantization Production,Quantization by Hardware Architecture,GGUF Compatibility,Model Swap and Rollbacka518fc7)RAM CPU Offloading in VLLM+ 5 restatement sites--swap-space→--kv-offloading-sizeclaim had six holders, two with their own copy of the upstream citation; it now lives only inRAM CPU Offloading in VLLM, which carries today's verification againstvllm/config/cache.py(500ac2b)Running Qwen3.6 27B FP8 on Dankclaw,ling-3.0-flash-vllm-plan,frank-redo-2026-08-1773fb272)FastFlowLM NPU Inference[[Dankclaw System Specs]]/[[Dankserver System Specs]]; the solo-held 14-18 t/s PP rate kept and tagged to the 2026-08 era; "frank-zealotnow runs the NPU endpoint" era-tagged (e32df1b)frank-medium-radiance-mxfp4,frank-medium-model-swap, both index pages6a8cbde)dankserver-vram-growthfrank-zealot(llama.cpp on the 3090 while that card was in Dankserver) from today's same-named endpoint on Dankclaw, so a retired crash signature cannot be read as a live fault on a same-card-class service (500ac2b)What stayed, and why
Owned ledgers are not duplication, so they were kept deliberately:
Ethernet Bandwidth Math—457.9 t/s, the ping-RTT table, the per-link cost model: solo-held, it is the page that measured them.x4-pair-experiment— the cross-device rate table (CUDA+Vulkan split 27.2 t/s / 268.5 prefill, RCCL tensor 19.09, layer split 18.8 / 1991, PP 16.66) and the 6.72 GB/s x4 link figure: no other page restates them (verified by grep across the vault).Dual-R9700-Cooperation— the R4D-vs-RCCL verdict and board-swap evidence matrix.ling-3.0-flash-trial,Running Qwen3.6 35B A3B FP8 on Dankserver,dankserver-vram-growth, the NPU runbook,Pipeline Parallelism/*) — retired payloads are an era page's content, not staleness. They needed banners and dates, not deletion.--swap-spaceflag inside dated era sections: kept as era records, labelled as such.Verification
./tests/run.sh→ ALL TESTS PASSED, 448 assertions, wall 86.49 s (includes thedocs-completeness and vault link-integrity seams).
0 broken wikilinks (the one regex hit is an ANSI-escape sequence inside inline code in
vLLM Boot Forensics, not a link).the pages that asserted it.
Method note
Three of the four research subagents dispatched for this campaign returned unusable or
inverted verdicts — one classified era pages as stale because they document retired
payloads, one reported files that do not exist, one claimed duplications that grep disproved
(
457.9 t/s, the RTT table), and one fabricated a line number in a page shorter than theline it cited. Every change in this branch is parent-written and parent-verified against the
files and the live fleet state; scout output was used only as a lead.
cwd: /home/blank/.omp/wt/wt-20261017-181502-6d1fc1f · session: 01a152fe-c490-702c-8280-c29b0bad4066