Merge pull request 'perf(quant): the calibration boot's step and slots are its own knobs — probe at the measured 71 tok/s' (#86) from fix/qwen38-calib-shape into main

Measured from the engine's own counters on the 2026-10-29 drain: prefill 71 tok/s steady state, a
180 s window showing 7 prefill steps of the capped 2,048 tokens and 3 preemptions. One 7,168-token
document is therefore ~100 s, round 1 (2,367 docs, 10.03M tokens) ~39 h, and the 12M-position
floor ~47 h of drained production — not the ~18 h the brief assumed.

The tap is the only phase whose cost is read volume per document: at the inherited 2,048-token step
a document is four steps, and each step re-streams every layer's routed expert union from the
container (43.4 GiB a rank live on disk). One wide step reads each layer's union once.

The 'one step, one slot count' audit law binds kldref and score because their KL numbers are only
comparable at an identical shape. Calib produces layer Hessians and no number any report compares,
so Q38_CALIB_MNBT / Q38_CALIB_SEQS are its own, defaulting to the inherited value — nothing moves
unless a probe asks. The KL modes ignore both.

Tests: 589/0 — the knob reaches calib's argv, does not reach kldref's, and a non-numeric value is
refused by name.

scripts/calib-push.py

@@ -22,6 +22,17 @@ Mechanics notes:
   prompts do not bias the Hessians (the corpus is drawn round-robin over
   source classes, so loss is class-neutral as long as failures are rare;
   a systematic 4xx is loud in the log and must be fixed, not ignored).
+- a round that seals no tap bytes is NORMAL and continues: RADIANCE_CALIB_TOKENS
+  is a pooled floor and this campaign's corpus is 10.03M tokens against a 12M
+  floor, so round 1 seals nothing by construction. A tap that LOSES bytes fails
+  the run, and a run that never sees the engine's completion line exits non-zero.
+  The first version failed on a flat round, which aborted exactly one round too
+  early (2026-10-29).
+- before any prompt, the tap must carry the rig's own marker
+  (`--tap-marker`, written by run-qwen38-bf16.sh calib) naming THIS
+  container, and that container must be running. Measured 2026-10-29:
+  with neither check, four rounds of 10M tokens went into restored
+  production after the rig's boot had died, and recorded nothing.
 
 Usage: calib-push.py [--port 7770] [--parallel 64] [--container arbench]
        [--corpus /home/dank/corpora/out/calib.jsonl] [--max-rounds 4]
@@ -29,11 +40,18 @@ Usage: calib-push.py [--port 7770] [--parallel 64] [--container arbench]
 import argparse
 import concurrent.futures
 import json
+import os
 import subprocess
 import sys
 import time
 import urllib.request
 
+_HERE = os.path.dirname(os.path.abspath(__file__)) if "__file__" in globals() else os.getcwd()
+if _HERE not in sys.path:
+    sys.path.insert(0, _HERE)
+
+import qwen38_common as qc  # noqa: E402 - sibling module, same directory
+
 DONE_MARK = "calibration wrote"
 
 
@@ -49,9 +67,9 @@ def tap_done(container):
     return DONE_MARK in (out.stdout + out.stderr)
 
 
-def push_one(url, prompt):
+def push_one(url, prompt, model):
     body = json.dumps({
-        "model": "frank-archon",
+        "model": model,
         "prompt": prompt,
         "max_tokens": 1,
         "temperature": 0,
@@ -68,38 +86,151 @@ def push_one(url, prompt):
         return False
 
 
-def main():
+def push_round(url, prompts, model, parallel):
+    """Push the whole corpus once; returns how many requests were accepted."""
+    ok = 0
+    with concurrent.futures.ThreadPoolExecutor(parallel) as ex:
+        for good in ex.map(lambda p: push_one(url, p, model), prompts):
+            ok += 1 if good else 0
+    return ok
+
+
+def tap_bytes(path):
+    """Bytes the tap has sealed so far; 0 when the directory is absent.
+
+    A tap whose directory is wrong, full or read-only stops writing while the
+    engine keeps answering 200 — the driver then re-pushes the whole corpus for
+    every round it is allowed and the loss is invisible until the hours are gone.
+    """
+    total = 0
+    try:
+        for root, _, files in os.walk(path):
+            for f in files:
+                try:
+                    total += os.path.getsize(os.path.join(root, f))
+                except OSError:
+                    pass
+    except OSError:
+        return 0
+    return total
+
+
+def round_verdict(before, after, done):
+    """What one round's tap bytes mean.
+
+    'done'   the engine logged that it sealed the tap — the run is over
+    'shrink' the tap LOST bytes: reset, pruned or replaced under us
+    'grew'   sealed bytes this round
+    'flat'   sealed nothing — which is legitimate: RADIANCE_CALIB_TOKENS is a
+             pooled floor, and a round that does not reach it has nothing to
+             write yet (the campaign's corpus is 10.03M tokens against a 12M
+             floor, so round 1 is always flat). Failing on it aborted the run
+             exactly one round too early.
+    """
+    if done:
+        return "done"
+    if after < before:
+        return "shrink"
+    if after == before:
+        return "flat"
+    return "grew"
+
+
+def require_rig(a):
+    """Refuse unless the rig that opened this tap is up and serving right now.
+
+    Three facts, in order: the runner wrote its marker into the tap dir, the
+    marker names the container this pusher targets, and that container is
+    running. A marker without a live container is a boot that died; a live
+    container without a marker is some other engine on the port — on 2026-10-29
+    it was restored production, which took four rounds of 10M tokens before the
+    old driver admitted it had recorded nothing.
+    """
+    marker = os.path.join(a.tap_dir, a.tap_marker)
+    if not os.path.isfile(marker):
+        qc.refuse(f"no rig marker at {marker}: this tap was not opened by "
+                  "run-qwen38-bf16.sh calib, so a push here would feed whatever "
+                  "answers the port and record nothing")
+    try:
+        lines = open(marker, encoding="utf-8", errors="replace").read().splitlines()
+    except OSError as exc:
+        qc.refuse(f"cannot read the rig marker {marker}: {exc}")
+    meta = dict(line.split("=", 1) for line in lines if "=" in line)
+    who = meta.get("container", "")
+    if who != a.container:
+        qc.refuse(f"the rig marker names container {who!r} and this pusher targets "
+                  f"{a.container!r} — wrong tap, or a stale marker")
+    try:
+        p = subprocess.run(["podman", "inspect", "-f", "{{.State.Running}}", a.container],
+                           capture_output=True, text=True, timeout=30)
+    except Exception as exc:
+        qc.refuse(f"cannot inspect the rig container {a.container}: {exc}")
+    if p.returncode != 0 or p.stdout.strip() != "true":
+        qc.refuse(f"the rig container {a.container} is not running — whatever is "
+                  "answering the port, it is not the calibration rig")
+
+
+def parse_args():
     ap = argparse.ArgumentParser()
     ap.add_argument("--port", type=int, default=7770)
     ap.add_argument("--parallel", type=int, default=64)
     ap.add_argument("--container", default="arbench")
+    ap.add_argument("--model", default="frank-archon")
+    ap.add_argument("--tap-dir", default="/home/dank/campaign/qwen38-w5nl64/calib")
+    ap.add_argument("--tap-marker", default=".rig-boot",
+                    help="the file run-qwen38-bf16.sh calib writes into the tap dir; without it the pusher refuses, because an unmarked tap is not a rig")
     ap.add_argument("--corpus", default="/home/dank/corpora/out/calib.jsonl")
     ap.add_argument("--max-rounds", type=int, default=4)
-    a = ap.parse_args()
+    return ap.parse_args()
-    url = f"http://127.0.0.1:{a.port}/v1/completions"
 
+
+def load_prompts(a) -> list[str]:
+    if not os.path.isfile(a.corpus):
+        qc.refuse(f"no corpus at {a.corpus}")
     prompts = []
-    with open(a.corpus) as f:
+    for lineno, rec in qc.iter_jsonl(a.corpus):
-        for line in f:
+        if not isinstance(rec.get("prompt"), str):
-            line = line.strip()
+            qc.refuse(f"{a.corpus}:{lineno} has no string 'prompt' field")
-            if line:
+        prompts.append(rec["prompt"])
-                prompts.append(json.loads(line)["prompt"])
+    if not prompts:
-    print(f"corpus: {len(prompts)} prompts", flush=True)
+        qc.refuse(f"the corpus at {a.corpus} holds no prompts")
+    return prompts
+
+
+def main():
+    a = parse_args()
+    url = f"http://127.0.0.1:{a.port}/v1/completions"
+    prompts = load_prompts(a)
+    require_rig(a)
+    chars = sum(len(p) for p in prompts)
+    print(f"corpus: {len(prompts)} prompts, ~{chars // 4} tokens (chars/4)",
+          flush=True)
 
     for rnd in range(1, a.max_rounds + 1):
         if tap_done(a.container):
             break
+        before = tap_bytes(a.tap_dir)
         t0 = time.time()
-        ok = 0
+        ok = push_round(url, prompts, a.model, a.parallel)
-        with concurrent.futures.ThreadPoolExecutor(a.parallel) as ex:
+        dt = max(time.time() - t0, 1e-9)
-            for good in ex.map(lambda p: push_one(url, p), prompts):
-                ok += 1 if good else 0
-        dt = time.time() - t0
         print(f"round {rnd}: {ok}/{len(prompts)} ok in {dt:.0f}s "
-              f"({sum(len(p) for p in prompts)/4/dt:.0f} chars/s est)",
+              f"({chars / 4 / dt:.0f} tokens/s est)", flush=True)
-              flush=True)
+        if ok == 0:
+            print("FAIL: not one prompt was accepted; fix the engine or the argv "
+                  "before spending another round", flush=True)
+            sys.exit(1)
         if tap_done(a.container):
             break
+        after = tap_bytes(a.tap_dir)
+        verdict = round_verdict(before, after, False)
+        if verdict == "shrink":
+            print(f"FAIL: the tap at {a.tap_dir} LOST bytes this round "
+                  f"({before} B before, {after} B after) — it was reset or replaced "
+                  "under us", flush=True)
+            sys.exit(1)
+        if verdict == "flat":
+            print(f"round {rnd}: the tap sealed nothing yet (byte floor not reached; "
+                  f"{before} B) — continuing", flush=True)
     else:
         print("max rounds exhausted, tap not done", flush=True)
         sys.exit(1)

scripts/convert-qwen38-w5nl64.sh

@@ -71,10 +71,12 @@ PLAN_ONLY=0
 KEEP_RECIPE=0
 while [ "$#" -gt 0 ]; do
     case "$1" in
-      --arm) ARMS+=("$2"); shift 2 ;;
+      --arm) [ "$#" -ge 2 ] || { echo "--arm needs a value (cd3|cd6|ao1)" >&2; exit 2; }
+             ARMS+=("$2"); shift 2 ;;
       --dry-run) DRY=1; shift ;;
       --plan-only) PLAN_ONLY=1; shift ;;
-      --target-count) TARGET_COUNT="$2"; shift 2 ;;
+      --target-count) [ "$#" -ge 2 ] || { echo "--target-count needs a value" >&2; exit 2; }
+             TARGET_COUNT="$2"; shift 2 ;;
       --keep-recipe) KEEP_RECIPE=1; shift ;;
       *) echo "unknown argument: $1" >&2; exit 2 ;;
     esac
@@ -124,6 +126,13 @@ fi
 for a in "${ARMS[@]}"; do
     [ -e "$OUTDIR/qwen38-w5nl64-$a.rad" ] && {
         echo "$OUTDIR/qwen38-w5nl64-$a.rad already exists; move it to trash first" >&2; exit 2; }
+    # A killed run leaves its container behind (--rm only fires when podman itself
+    # exits), and the next arm then dies on "name already in use" AFTER the plan
+    # phase; refuse here, where the fix is one `podman rm -f`.
+    podman container exists "qwen38-convert-$a" && {
+        echo "container qwen38-convert-$a still exists (a killed run?);" \
+             "'podman rm -f qwen38-convert-$a' and re-run" >&2
+        exit 2; }
 done
 
 # --- the protected plane, spliced into a per-run copy of the template --------------------------
@@ -203,6 +212,7 @@ for arm in "${ARMS[@]}"; do
     rc=0; run_convert "$arm" || rc=$?
     [ "$rc" -eq 0 ] || { echo "arm $arm: rad-convert rc=$rc; log: $STATE/convert-$arm-$TS.log" >&2; overall=1; continue; }
     out="$OUTDIR/qwen38-w5nl64-$arm.rad"
+    [ -f "$out" ] || { echo "arm $arm: rad-convert exited 0 but wrote no $out; log: $STATE/convert-$arm-$TS.log" >&2; overall=1; continue; }
     got_mib=$(( $(stat -c %s "$out") / 1048576 ))
     # The receipt: the artifact is what the plan predicted, within the metadata the plan does not
     # count. An arm that came out materially smaller lost a plane somewhere.

scripts/frank-archon.env

@@ -74,12 +74,14 @@ FRANK_ARCHON_PLACEMENT=expert_tiered
 # residency -- ~3.2 GiB a card of experts now lives in host RAM and is read over the
 # x8 link when the router picks it. The ledger's coverage line is the referee, not
 # this file: if vram_kv comes back under 700,000 tokens, raise this 1024 and redeploy.
-# 2026-10-29: 49,152 -> 66,560 (tp x 65 GiB, inside the 246 GiB the box has). The
+# 2026-10-29 A/B, measured on both payloads and folded back the same day: at 66,560 the
-# campaign rig's bf16 container at the shared step reported "kv cache 269.99 MiB a
+# endpoint printed the identical `elastic 26.73 GiB / experts 20.72 / kv cache 5.64 /
-# card: 7436 tokens", so 682 of the 1,294 calibration documents were refused at that
+# vram_weights 24.30` it had printed at 49,152, opening the bigger pool with `0 B used`;
-# pool; the pool, not the step, is what widens it (scripts/run-qwen38-bf16.sh). The
+# the campaign rig (bf16, same step) moved `vram_kv 269.99 -> 271.25 MiB`, a rounding
-# new cache line belongs to the next boot ledger, not to this comment.
+# step — its window is fixed by its own n-gram table (95.37 GiB) and expert plane, not
-FRANK_ARCHON_HOST_POOL_MIB=66560
+# by this pool, so the refused calibration documents were fixed by cutting them to the
+# window (scripts/rechunk-kld-corpus.py), not by widening the pool.
+FRANK_ARCHON_HOST_POOL_MIB=49152
 # VRAM left unclaimed per card (publisher default 96 MiB); the engine's settle
 # floor is built from it.
 FRANK_ARCHON_GPU_HEADROOM_MIB=96
@@ -142,10 +144,10 @@ FRANK_ARCHON_NGRAM_PLACEMENT=ram
 # itself before refusing by name, so recovery is a boot, not a visit.
 # Floor = 2 x HOST_POOL + 48,824 (static weights + arena) + PREFIX_CACHE + MNBT + the
 # auto checkpoint store (MAX_SEQS x 10 slots of 111.98 MiB), per the bats ledger test;
-# at the 66,560 pool with 4 slots that computes to 196,664 MiB (161,848 at the old
+# at the 49,152 pool with 4 slots that computes to 161,848 MiB (196,664 at the
-# 49,152 pool).
+# 66,560 pool the 2026-10-29 A/B rejected).
-FRANK_ARCHON_TTM_MIN_MIB=197000
+FRANK_ARCHON_TTM_MIN_MIB=162000
-FRANK_ARCHON_TTM_TARGET_MIB=224000
+FRANK_ARCHON_TTM_TARGET_MIB=204800
 # Escape hatch for experiments. Unset -> no extra argv; the serve script expands
 # this unquoted on purpose, so keep multi-flag values as one string.
 FRANK_ARCHON_EXTRA_ARGS=

scripts/kld-gate-qwen38.sh

@@ -1,44 +1,53 @@
 #!/usr/bin/env bash
 # kld-gate-qwen38.sh — RUNS ON DANKSERVER (or anywhere the reports are). The campaign's
-# acceptance gate: three KLD reports against ONE reference recording, one verdict.
+# acceptance gate: one verdict from the KLD reports of the containers that ride one reference
+# recording.
 #
 # The reference is the BF16 container built from the vendor checkpoint (run-qwen38-bf16.sh
 # kldref, in the same drained window that records the calibration grams). Every scored
 # container is measured against that one recording, so the KL scale runs from 0 = the
-# checkpoint itself downward in faithfulness. Three containers ride it:
+# checkpoint itself downward in faithfulness.
 #
-#   fp8.json    the published FP8-native release   — the achievable end of the interval
+# TWO MODES, because the disk decides how many anchors can exist at once:
-#   iq4r.json   the payload frank-archon serves    — the bar to beat
-#   cand.json   the arm under test                 — the subject
 #
-# The FP8 number is what makes the interval absolute instead of relative: "half the gap"
+#   gate IQ4R.json CAND.json                    candidate vs the served payload
-# is half of the distance between the 4-bit payload and the near-lossless release, not half
+#   gate FP8.json IQ4R.json CAND.json           the same, inside the absolute interval
-# of an invented distance to the checkpoint. That is also why the FP8 payload is FETCHED and
-# scored, never converted: it is a second opinion about what this checkpoint tolerates, not a
-# step of the campaign. It lands LAST (disk sequencing) — by then the 335 GiB bf16 container
-# is gone, so its 172.8 GiB fits.
 #
-# The law, in order of severity, on the mean KL of one slice (--slice all | a corpus source):
+# iq4r.json is the payload frank-archon serves — the bar to beat. When the FP8-native release
-#   1. KL(fp8) < KL(iq4r)                 the anchors are ordered. If the near-lossless
+# (172.8 GiB, fetched last because of the peak-disk law) is on disk, pass its report first:
-#                                         release does NOT sit closer to the checkpoint than
+# "half the gap" is then half of the distance between the 4-bit payload and the near-lossless
-#                                         the 4-bit payload, the corpus, the tokenizer or the
+# release, not half of an invented distance to the checkpoint. Without it the gate still
-#                                         boot shape is wrong and no arm number means anything.
+# answers the campaign's question — does the candidate beat what is served — and says so.
-#   2. KL(fp8) < KL(cand) < KL(iq4r)      the candidate lies strictly inside the interval —
+#
-#                                         outside it, something is mis-measured, not better
+# The law, on the summary of one slice (--slice all | a corpus source):
-#   3. KL(cand) <= KL(iq4r) - R x gap     it closes at least R (--recover, default 0.5) of the
+#
-#                                         gap = KL(iq4r) - KL(fp8): at least half the way from
+#   candidate vs served (2 reports)
-#                                         what is served toward the release (agreed 2026-10-27)
+#   1. mean KL(cand) < mean KL(iq4r)         the served payload is beaten on the corpus
-#   4. top-1(cand) >= top-1(iq4r)         the agreement the harness keys on does not fall
+#   2. p99.9(cand) <= p99.9(iq4r)            and the tail is no worse — a mean can hide a few
+#                                            positions taking the whole hit
+#   3. top-1(cand) >= top-1(iq4r)            the agreement the harness keys on does not fall
+#   4. every source slice passes 1-3         a win bought by regressing one distribution
+#                                            (code, prose, tools) is not a win
+#
+#   inside the interval (3 reports), laws 1-4 above (1-3 renumbered to the interval and the
+#   recovery fraction, as agreed 2026-10-27) plus:
+#   5. p99.9(cand) <= p99.9(iq4r)            the tail law
+#   6. every source slice passes            the slice law
 #
 # Booting the containers is NOT this script's job: `run-qwen38-bf16.sh score RAD OUT.json`
 # writes the reports and owns the single engine argv (same pin, --tp 2, --kv-cache-dtype fp8,
-# same step size, --deterministic, no speculation) that makes the three numbers comparable.
+# same step size, --deterministic, no speculation) that makes the numbers comparable.
 # This script is arithmetic, so it re-runs against old reports and is testable without a GPU.
 #
-# Usage: kld-gate-qwen38.sh FP8.json IQ4R.json CAND.json [--slice NAME] [--recover F]
+# Usage: kld-gate-qwen38.sh IQ4R.json CAND.json [--slice NAME]
+#        kld-gate-qwen38.sh FP8.json IQ4R.json CAND.json [--slice NAME] [--recover F]
 set -uo pipefail
 
-[ "$#" -ge 3 ] || { echo "usage: kld-gate-qwen38.sh FP8.json IQ4R.json CAND.json [--slice NAME] [--recover F]" >&2; exit 2; }
+if [ "$#" -lt 2 ]; then
+    echo "usage: kld-gate-qwen38.sh IQ4R.json CAND.json [--slice NAME]" >&2
+    echo "       kld-gate-qwen38.sh FP8.json IQ4R.json CAND.json [--slice NAME] [--recover F]" >&2
+    exit 2
+fi
 python3 - "$@" <<'PY'
 import json, sys
 
@@ -52,13 +61,15 @@ while i < len(a):
     else:
         files.append(a[i])
         i += 1
-if len(files) != 3:
+if len(files) not in (2, 3):
-    sys.exit("usage: kld-gate-qwen38.sh FP8.json IQ4R.json CAND.json [--slice NAME] [--recover F]")
+    sys.exit("usage: kld-gate-qwen38.sh IQ4R.json CAND.json [--slice NAME]\n"
+             "       kld-gate-qwen38.sh FP8.json IQ4R.json CAND.json [--slice NAME] [--recover F]")
 o = dict(zip(opts[0::2], opts[1::2]))
 slice_ = o.get("--slice", "all")
 recover = float(o.get("--recover", 0.5))
 if not 0.0 < recover <= 1.0:
     sys.exit(f"--recover must be in (0,1], got {recover}")
+interval = len(files) == 3
 
 
 def die(msg):
@@ -66,11 +77,15 @@ def die(msg):
     raise SystemExit(2)
 
 
-def read(path):
+def load(path):
     try:
-        d = json.load(open(path))
+        return json.load(open(path))
     except Exception as e:
         die(f"cannot read the KLD report {path}: {e}")
+
+
+def summary(path, d):
+    """The slice this gate grades: 'all', or one corpus source."""
     if slice_ == "all":
         g = d.get("all")
     else:
@@ -80,23 +95,25 @@ def read(path):
                 f"{', '.join(sorted(d.get('by_source') or {})) or '(none)'}")
     if g is None or "kld" not in g or "top1_agreement" not in g:
         die(f"{path} has no {slice_!r} summary — wrong report version?")
-    return d, g
+    return g
 
 
+docs = {label: load(path) for label, path in
+        zip(("fp8", "iq4r", "cand") if interval else ("iq4r", "cand"), files)}
 rows = []
-for label, path in zip(("fp8", "iq4r", "cand"), files):
+for label, path in zip(("fp8", "iq4r", "cand") if interval else ("iq4r", "cand"), files):
-    d, g = read(path)
+    g = summary(path, docs[label])
-    rows.append((label, path, d.get("model", "?"), g["kld"]["mean"], g["kld"]["p99_9"],
+    rows.append((label, path, docs[label].get("model", "?"), g["kld"]["mean"],
-                 g["top1_agreement"], g["positions"]))
+                 g["kld"]["p99_9"], g["top1_agreement"], g["positions"]))
 
-print(f"reference: BF16 container; slice: {slice_}")
+print(f"reference: BF16 container; slice: {slice_}; "
+      + ("inside the FP8 interval" if interval else "candidate vs served (no FP8 anchor)"))
 print(f"{'':6} {'model':38} {'KL mean':>10} {'KL p99.9':>10} {'top-1':>8}  positions")
 for label, path, model, kl, p999, top1, n in rows:
     print(f"{label:6} {str(model)[:38]:38} {kl:10.6f} {p999:10.4f} {100*top1:7.2f}%  {n}")
-r = {label: (kl, top1) for label, _p, _m, kl, _q, top1, _n in rows}
+r = {label: (kl, p999, top1) for label, _p, _m, kl, p999, top1, _n in rows}
-kl_fp8, kl_iq4r, kl_cand = r["fp8"][0], r["iq4r"][0], r["cand"][0]
+kl_cand, p_cand, t_cand = r["cand"]
-t_iq4r, t_cand = r["iq4r"][1], r["cand"][1]
+kl_iq4r, p_iq4r, t_iq4r = r["iq4r"]
-gap = kl_iq4r - kl_fp8
 ok = True
 
 
@@ -106,6 +123,43 @@ def law(n, text, passed):
     print(f"  [{'PASS' if passed else 'FAIL'}] {n}. {text}")
 
 
+def tail_law(n):
+    law(n, f"candidate KL p99.9 {p_cand:.4f} <= served iq4r {p_iq4r:.4f}",
+        p_cand <= p_iq4r)
+
+
+# --- the slice law: every corpus source must pass the metric laws -------------
+def slice_law(n):
+    """A win that regresses one source class is not a win: check them all."""
+    mine = docs["cand"].get("by_source") or {}
+    theirs = docs["iq4r"].get("by_source") or {}
+    if not mine or not theirs:
+        law(n, "no per-source summaries to check", True)
+        return
+    missing = sorted(set(theirs) - set(mine))
+    extra = sorted(set(mine) - set(theirs))
+    if missing or extra:
+        law(n, "same source slices in both reports"
+            + (f" (missing from cand: {', '.join(missing)})" if missing else "")
+            + (f" (only in cand: {', '.join(extra)})" if extra else ""), False)
+        return
+    bad = []
+    for src in sorted(theirs):
+        c, q = mine[src], theirs[src]
+        worse_mean = c["kld"]["mean"] >= q["kld"]["mean"]
+        worse_tail = c["kld"]["p99_9"] > q["kld"]["p99_9"]
+        worse_top1 = c["top1_agreement"] < q["top1_agreement"]
+        if worse_mean or worse_tail or worse_top1:
+            why = ",".join(m for m, w in (("mean", worse_mean), ("p99.9", worse_tail),
+                                          ("top-1", worse_top1)) if w)
+            bad.append(f"{src} ({why})")
+    law(n, "every source slice passes (mean, p99.9, top-1)"
+        + (f" — regressed: {'; '.join(bad)}" if bad else ""), not bad)
+
+
+if interval:
+    kl_fp8 = r["fp8"][0]
+    gap = kl_iq4r - kl_fp8
     law(1, f"anchors ordered: fp8 {kl_fp8:.6f} < iq4r {kl_iq4r:.6f} (gap {gap:.6f})",
         gap > 0)
     law(2, f"candidate inside the interval: {kl_fp8:.6f} < {kl_cand:.6f} < {kl_iq4r:.6f}",
@@ -117,6 +171,16 @@ law(3, f"candidate closes {100*max((kl_iq4r - kl_cand)/gap, 0):.1f}% of the "
         gap > 0 and kl_cand <= need)
     law(4, f"candidate top-1 {100*t_cand:.2f}% >= served iq4r top-1 {100*t_iq4r:.2f}%",
         t_cand >= t_iq4r)
+    tail_law(5)
+    slice_law(6)
+else:
+    law(1, f"candidate beats the served payload: KL {kl_cand:.6f} < iq4r {kl_iq4r:.6f}",
+        kl_cand < kl_iq4r)
+    tail_law(2)
+    law(3, f"candidate top-1 {100*t_cand:.2f}% >= served iq4r top-1 {100*t_iq4r:.2f}%",
+        t_cand >= t_iq4r)
+    slice_law(4)
+
 print("KLD GATE " + ("PASS" if ok else "FAIL"))
 sys.exit(0 if ok else 1)
 PY

scripts/make-calib-corpus.py

@@ -34,14 +34,25 @@ import json
 import os
 import random
 import re
+import sys
 
 from jinja2 import Environment, FileSystemLoader
-from tokenizers import Tokenizer
+
+_HERE = os.path.dirname(os.path.abspath(__file__)) if "__file__" in globals() else os.getcwd()
+if _HERE not in sys.path:
+    sys.path.insert(0, _HERE)
+
+import qwen38_common as qc  # noqa: E402 - sibling module, same directory
 
 CODE_EXT = (".sh", ".py", ".jinja", ".recipe", ".env", ".example",
             ".yaml", ".yml", ".bats", ".service", ".container", ".toml")
 JSONL_EXTS = (".jsonl", ".json")
 DOC_DIR = "serverAppConfigs-docs"
+FRAGMENT_TOKENS = 2400   # what one fragment may contribute
+CLASS_MIX = {"code": .45, "prose": .30, "tools": .25}
+HF_CAP = {"code": 40000, "prose": 60000, "tools": 0}
+POOL_CYCLES = 40         # how many times a class pool may repeat before giving up
+KLD_GUARD = 5000         # rendered conversations tried before declaring the pool short
 
 FLEET_CMDS = [
     "systemctl status frank-archon --no-pager",
@@ -142,7 +153,7 @@ def load_parquet_texts(path: str, kind: str, rng: random.Random,
     return out[:cap]
 
 
-def main() -> None:
+def parse_args() -> argparse.Namespace:
     ap = argparse.ArgumentParser(description=__doc__)
     ap.add_argument("--repo", required=True)
     ap.add_argument("--tokenizer-json", required=True)
@@ -156,22 +167,26 @@ def main() -> None:
     ap.add_argument("--hf-code-parquet", default="")
     ap.add_argument("--hf-prose-parquet", default="")
     ap.add_argument("--seed", type=int, default=20261006)
-    a = ap.parse_args()
+    return ap.parse_args()
 
-    rng = random.Random(a.seed)
+
-    tokenizer = Tokenizer.from_file(a.tokenizer_json)
+def load_template(template: str) -> tuple[Environment, str]:
-    env = Environment(loader=FileSystemLoader(os.path.dirname(a.template)),
+    """The chat template as the engine will render it, plus its short name."""
+    env = Environment(loader=FileSystemLoader(os.path.dirname(template)),
                       trim_blocks=True, lstrip_blocks=True)
     env.filters["tojson"] = lambda v: json.dumps(v, ensure_ascii=False)
     env.filters["raise_exception"] = lambda m: (_ for _ in ()).throw(
         ValueError(m))
-    tname = os.path.basename(a.template)
+    return env, os.path.basename(template)
 
+
+def classify(repo: str) -> dict[str, list[str]]:
+    """Walk the checkout into the three source classes the mix draws from."""
     code, prose, jsonl = [], [], []
-    for root, dirs, files in os.walk(a.repo):
+    for root, dirs, files in os.walk(repo):
         dirs[:] = [d for d in dirs if not d.startswith(".")]
         for f in sorted(files):
-            p = os.path.relpath(os.path.join(root, f), a.repo)
+            p = os.path.relpath(os.path.join(root, f), repo)
             if DOC_DIR in p and f.endswith(".md"):
                 prose.append(p)
             elif p.startswith(DOC_DIR):
@@ -180,64 +195,92 @@ def main() -> None:
                 code.append(p)
             elif f.endswith(JSONL_EXTS) and f != "package-lock.json":
                 jsonl.append(p)
-    # file-level split: bucket 0 -> KLD pool, 1..4 -> calibration pool
+    return {"code": code, "prose": prose, "tools": jsonl}
-    pools: dict[str, tuple[list, list]] = {}
+
-    for cls, lst in (("code", code), ("prose", prose), ("tools", jsonl)):
+
-        pools[cls] = ([p for p in lst if fnv(p) % 5 == 0],
+def split_pools(classes: dict[str, list[str]]) -> dict[str, tuple[list, list]]:
+    """File-level split: bucket 0 -> KLD pool, everything else -> calibration."""
+    return {cls: ([p for p in lst if fnv(p) % 5 == 0],
                   [p for p in lst if fnv(p) % 5 != 0])
+            for cls, lst in classes.items()}
 
-    def text_of(p: str) -> str:
-        path = os.path.join(a.repo, p)
-        raw = open(path, encoding="utf-8", errors="replace").read()
-        if p.endswith(JSONL_EXTS):
-            lines = [json.loads(l)["content"] for l in raw.splitlines()
-                     if l.strip() and "content" in l]
-            raw = "\n".join(lines)
-        toks = tokenizer.encode(raw, add_special_tokens=False).ids
-        return tokenizer.decode(toks[:2400], skip_special_tokens=True)
 
+def jsonl_content(raw: str) -> str:
+    """The content fields of a content-keyed JSONL, or "" for anything else.
+
+    A .json in this repo is not necessarily a content-keyed JSONL: the substring
+    test passes on any line that merely mentions "content", and a list-shaped
+    document raises TypeError — both used to crash the whole build. Skip what
+    does not fit instead.
+    """
+    lines = []
+    for line in raw.splitlines():
+        if not line.strip() or "content" not in line:
+            continue
+        try:
+            rec = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+        if isinstance(rec, dict) and isinstance(rec.get("content"), str):
+            lines.append(rec["content"])
+    return "\n".join(lines)
+
+
+def fragment_text(repo: str, rel: str, tok) -> str:
+    """One file as a fragment: its first FRAGMENT_TOKENS tokens, decoded."""
+    raw = open(os.path.join(repo, rel), encoding="utf-8",
+               errors="replace").read()
+    if rel.endswith(JSONL_EXTS):
+        raw = jsonl_content(raw)
+    ids = tok.encode(raw, add_special_tokens=False).ids
+    return tok.decode(ids[:FRAGMENT_TOKENS], skip_special_tokens=True)
+
+
+def build_frags(a: argparse.Namespace, tok,
+                rng: random.Random) -> dict[str, dict[str, list[str]]]:
+    """Both fragment pools per class, seeded so the same seed rebuilds the corpus."""
+    pools = split_pools(classify(a.repo))
     cmds = harvest_cmds(a.repo, rng)
-    # fragment pools per class, corpus-split first
     frags: dict[str, dict[str, list[str]]] = {}
-    for cls in pools:
+    for cls, (kld_files, cal_files) in pools.items():
-        kld_files, cal_files = pools[cls]
-        fr = {}
-        src = {"kld": kld_files, "cal": cal_files}
         if cls == "tools":
-            fr["kld"] = [tool_call(rng, cmds) for _ in range(600)]
+            frags[cls] = {"kld": [tool_call(rng, cmds) for _ in range(600)],
-            fr["cal"] = [tool_call(rng, cmds) for _ in range(40000)]
+                          "cal": [tool_call(rng, cmds) for _ in range(40000)]}
         else:
-            fr["kld"] = [text_of(p) for p in src["kld"]]
+            frags[cls] = {"kld": [fragment_text(a.repo, p, tok) for p in kld_files],
-            fr["cal"] = [text_of(p) for p in src["cal"]]
+                          "cal": [fragment_text(a.repo, p, tok) for p in cal_files]}
-        frags[cls] = fr
     # per-class fragment caps: prose rows are short (~66 tok) so need many;
     # codeparrot files are long (~2.6k tok) so a small cap suffices
-    hfcap = {"code": 40000, "prose": 60000, "tools": 0}
     for cls, spec in (("code", a.hf_code_parquet),
                       ("prose", a.hf_prose_parquet)):
         for path in [p for p in spec.split(",") if p]:
             frags[cls]["cal"].extend(
-                load_parquet_texts(path, cls, rng, cap=hfcap[cls]))
+                load_parquet_texts(path, cls, rng, cap=HF_CAP[cls]))
     for cls in frags:
         rng.shuffle(frags[cls]["cal"])
+    return frags
 
-    def toks(s: str) -> int:
-        return len(tokenizer.encode(s, add_special_tokens=False).ids)
 
-    def emit_calib() -> tuple[int, list[dict]]:
+def pack_class(cls: str, pool: list[str], want: int, tok,
-        targets = {"code": .45, "prose": .30, "tools": .25}
+               chunk_tokens: int) -> list[dict]:
-        rows: list[dict] = []
+    """Greedily pack `pool` into chunk_tokens-sized calibration prompts."""
-        for cls, frac in targets.items():
-            want = int(a.calib_tokens * frac)
-            pool = frags[cls]["cal"]
     if not pool:
-                continue
+        return []
-            got, i, buf, nbuf = 0, 0, [], 0
+    rows: list[dict] = []
+    # Count every fragment once: the loop cycles the pool up to POOL_CYCLES
+    # times, and re-tokenizing a 2.4k-token fragment on each pass dominated the
+    # build's wall time.
+    counts = [qc.n_tokens(tok, f) for f in pool]
+    got = 0
+    i = 0
+    buf: list[str] = []
+    nbuf = 0
     while got < want:
-                frag = pool[i % len(pool)]
+        j = i % len(pool)
+        frag = pool[j]
         i += 1
-                n = toks(frag)
+        n = counts[j]
-                if nbuf + n > a.chunk_tokens and buf:
+        if nbuf + n > chunk_tokens and buf:
             rows.append({"prompt": "\n".join(buf), "cls": cls})
             got += nbuf
             buf, nbuf = [], 0
@@ -245,22 +288,36 @@ def main() -> None:
                 break
         buf.append(frag)
         nbuf += n
-                if i > len(pool) * 40:
+        if i > len(pool) * POOL_CYCLES:
             break  # pool exhausted with repeats; stop
     if buf and got < want:
         rows.append({"prompt": "\n".join(buf), "cls": cls})
         got += nbuf
     print(f"calib {cls}: {got} tokens, {i} fragments")
-        rng.shuffle(rows)
+    return rows
-        return len(rows), [{k: v for k, v in r.items() if k == "prompt"}
-                           for r in rows]
 
-    def emit_kld() -> tuple[int, list[dict]]:
+
-        rows, total = [], 0
+def emit_calib(a: argparse.Namespace, frags, tok,
+               rng: random.Random) -> tuple[int, list[dict]]:
+    """calib.jsonl rows: the class mix, shuffled, prompts only."""
+    rows: list[dict] = []
+    for cls, frac in CLASS_MIX.items():
+        rows.extend(pack_class(cls, frags[cls]["cal"], int(a.calib_tokens * frac),
+                               tok, a.chunk_tokens))
+    rng.shuffle(rows)
+    return len(rows), [{"prompt": r["prompt"]} for r in rows]
+
+
+def emit_kld(a: argparse.Namespace, frags, tok, rng: random.Random,
+             env: Environment, tname: str) -> tuple[int, list[dict], int]:
+    """kld.jsonl rows: multi-turn conversations, each with a scored tail."""
+    template = env.get_template(tname)
+    rows: list[dict] = []
+    total = scored_total = 0
     classes = ["code", "prose", "tools"]
-        i = {c: 0 for c in classes}
+    cursor = {c: 0 for c in classes}
     guard = 0
-        while total < a.kld_tokens and guard < 5000:
+    while total < a.kld_tokens and guard < KLD_GUARD:
         guard += 1
         msgs, srcs = [], []
         for t in range(rng.choice([2, 3, 4])):
@@ -268,49 +325,55 @@ def main() -> None:
             pool = frags[cls]["kld"]
             if not pool:
                 continue
-                frag = pool[i[cls] % len(pool)]
+            frag = pool[cursor[cls] % len(pool)]
-                i[cls] += 2
+            cursor[cls] += 1
             msgs.append({"role": "user",
                          "content": "Summarize and explain this "
                                     f"engineering material:\n{frag}"})
             msgs.append({"role": "assistant",
                          "content": frag[:4000]})
             srcs.append(cls)
-            rendered = env.get_template(tname).render(
+        rendered = template.render(messages=msgs, add_generation_prompt=False)
-                messages=msgs, add_generation_prompt=False)
+        n = qc.n_tokens(tok, rendered)
-            n = toks(rendered)
         if n < a.kld_min_prompt:
             continue
         rows.append({"prompt": rendered,
                      "score_from": n - a.score_tail,
                      "source": "+".join(sorted(set(srcs)))})
         total += n
-        return len(rows), rows
+        scored_total += a.score_tail
+    return len(rows), rows, scored_total
 
-    os.makedirs(a.out_dir, exist_ok=True)
+
-    nc, rows = emit_calib()
+def manifest(a: argparse.Namespace, tname: str, nc: int, nk: int,
-    with open(os.path.join(a.out_dir, "calib.jsonl"), "w") as f:
+             scored: int) -> dict:
-        for r in rows:
+    return {
-            f.write(json.dumps(r) + "\n")
-    nk, krows = emit_kld()
-    with open(os.path.join(a.out_dir, "kld.jsonl"), "w") as f:
-        for r in krows:
-            f.write(json.dumps(r) + "\n")
-    scored = sum(len(tokenizer.encode(r["prompt"], add_special_tokens=False).ids)
-                 - r["score_from"] for r in krows)
-    manifest = {
         "seed": a.seed, "template": tname,
         "tokenizer": os.path.basename(a.tokenizer_json),
         "calib_prompts": nc, "kld_prompts": nk,
         "kld_scored_positions": scored,
         "kld_score_tail": a.score_tail,
         "disjoint_file_pools": "local KLD material only; HF code feeds calib",
-        "calib_mix": {"code": .45, "prose": .30, "tools": .25},
+        "calib_mix": dict(CLASS_MIX),
         "hf_code": a.hf_code_parquet, "hf_prose": a.hf_prose_parquet,
     }
-    with open(os.path.join(a.out_dir, "manifest.json"), "w") as f:
+
-        json.dump(manifest, f, indent=1)
+
-    print(json.dumps(manifest, indent=1))
+def main() -> None:
+    a = parse_args()
+    rng = random.Random(a.seed)
+    tok = qc.load_tokenizer(a.tokenizer_json)
+    env, tname = load_template(a.template)
+
+    frags = build_frags(a, tok, rng)
+    os.makedirs(a.out_dir, exist_ok=True)
+    nc, calib_rows = emit_calib(a, frags, tok, rng)
+    qc.write_jsonl(os.path.join(a.out_dir, "calib.jsonl"), calib_rows)
+    nk, kld_rows, scored = emit_kld(a, frags, tok, rng, env, tname)
+    qc.write_jsonl(os.path.join(a.out_dir, "kld.jsonl"), kld_rows)
+    man = manifest(a, tname, nc, nk, scored)
+    qc.write_json(os.path.join(a.out_dir, "manifest.json"), man)
+    print(json.dumps(man, indent=1))
 
 
 if __name__ == "__main__":

scripts/qwen38-calib-window.sh

@@ -0,0 +1,122 @@
+#!/usr/bin/env bash
+# qwen38-calib-window.sh — RUNS ON DANKSERVER. The calibration window, and nothing else:
+# boot the BF16 rig, wait until it is actually serving, push the corpus until the engine says
+# the tap is full, stop the rig, restore frank-archon, and bank the reference container.
+#
+# The reference recording is already taken (kld-ref/kld.json exists), so this window is the
+# calibration tap alone — the one job that needs both cards after it. The kldref phase is NOT
+# re-run here: the chained launcher that used to do that moved the banked kld-ref into trash
+# first, which is how a 16-hour phase gets paid for twice.
+#
+# Every phase is gated on the previous one, because the last time this ran unguarded
+# (2026-10-29) the rig's boot died of an OOM, the health loop quietly fell through, and the
+# pusher fed four rounds of 10M tokens into RESTORED PRODUCTION with nothing recorded:
+#
+#   1. the runner must still be alive and its container must answer /health, or this script
+#      kills it, restores the service and exits non-zero — no push, ever;
+#   2. the pusher refuses on its own unless the rig wrote its marker into the tap dir, the
+#      marker names this container, and the container is running (scripts/calib-push.py
+#      --tap-marker), so a second lock survives a mistake in this file;
+#   3. the reference container is banked into trash only when BOTH the serve and the push
+#      exited 0 and the tap dir holds grams. The checkpoint that reproduces it stays.
+#
+# knobs: Q38_WINDOW_CALIB_CORPUS (calib-7168.jsonl — the re-chunked file the rig's pool can
+#        hold), Q38_WINDOW_HEALTH_TIMEOUT (seconds to wait for /health), Q38_WINDOW_ROUNDS
+#        (pusher rounds; one is 10M tokens), Q38_WINDOW_KEEP_RAD=1 to leave the .rad in place.
+# Usage: scripts/qwen38-calib-window.sh [--dry-run]
+set -uo pipefail
+cd "$(dirname "${BASH_SOURCE[0]}")/.." || exit 9
+
+DRY=0
+[ "${1:-}" = "--dry-run" ] && DRY=1
+say() { printf '%s %s\n' "$(date -Is)" "$*"; }
+
+MODELS="${FRANK_QWEN38_MODELS:-/home/dank/models}"
+CAMPAIGN="${FRANK_QWEN38_CAMPAIGN:-/home/dank/campaign/qwen38-w5nl64}"
+RAD="${FRANK_QWEN38_RAD:-$MODELS/rad/qwen38-bf16.rad}"
+CORPUS="${Q38_WINDOW_CALIB_CORPUS:-/home/dank/corpora/out/calib-7168.jsonl}"
+KLDREF="$CAMPAIGN/kld-ref/kld.json"
+CALIB="$CAMPAIGN/calib"
+HEALTH_TIMEOUT="${Q38_WINDOW_HEALTH_TIMEOUT:-3600}"
+ROUNDS="${Q38_WINDOW_ROUNDS:-6}"
+RIG=qwen38-bf16
+SERVICE=frank-archon
+PY="${PY:-python3}"
+
+# ---- preflight: everything this window needs, before a card is touched -------
+[ -f "$KLDREF" ] || { echo "REFUSED: no reference recording at $KLDREF — record it first (scripts/run-qwen38-bf16.sh kldref); this window only fills the calibration tap" >&2; exit 2; }
+[ -f "$RAD" ]    || { echo "REFUSED: no BF16 reference container at $RAD — nothing to serve; scripts/build-qwen38-bf16-rad.sh" >&2; exit 2; }
+[ -f "$CORPUS" ] || { echo "REFUSED: no calibration corpus at $CORPUS" >&2; exit 2; }
+[ -d "$CALIB" ]  || { echo "REFUSED: no tap dir at $CALIB" >&2; exit 2; }
+[ -w "$CALIB" ]  || { echo "REFUSED: tap dir $CALIB is not writable by $(id -un)" >&2; exit 2; }
+! [[ "$HEALTH_TIMEOUT" =~ ^[0-9]+$ ]] && { echo "REFUSED: Q38_WINDOW_HEALTH_TIMEOUT must be seconds (got: $HEALTH_TIMEOUT)" >&2; exit 2; }
+! [[ "$ROUNDS" =~ ^[0-9]+$ ]] && { echo "REFUSED: Q38_WINDOW_ROUNDS must be a round count (got: $ROUNDS)" >&2; exit 2; }
+
+if [ "$DRY" = 1 ]; then
+    echo "dry run: calib window — the service is left running, nothing is mounted or written"
+    echo "would run: bash scripts/run-qwen38-bf16.sh calib          (rig boots, writes $CALIB/.rig-boot)"
+    echo "would wait: /health up within ${HEALTH_TIMEOUT}s, else kill the rig, restore $SERVICE, exit 1"
+    echo "would run: $PY scripts/calib-push.py --container $RIG --corpus $CORPUS --max-rounds $ROUNDS"
+    echo "would bank: $RAD -> ~/.AGENTTRASH/$(date +%Y%m%d)-qwen38-bf16-rad/  (only if both phases exit 0)"
+    exit 0
+fi
+
+say "calib window start pid=$$ corpus=$CORPUS rounds=$ROUNDS"
+[ -n "$(ls -A "$CALIB" 2>/dev/null)" ] && say "tap dir $CALIB is not empty; the engine's token floor is pooled, so an earlier partial tap still counts"
+
+# ---- phase 1: the rig -------------------------------------------------------
+say "phase=calib serve begin"
+bash scripts/run-qwen38-bf16.sh calib & spid=$!
+
+rig_up() {
+    podman inspect -f '{{.State.Running}}' "$RIG" 2>/dev/null | grep -qx true \
+        && curl -sf --max-time 10 -o /dev/null http://127.0.0.1:7770/health
+}
+
+waited=0
+until rig_up; do
+    if ! kill -0 "$spid" 2>/dev/null; then
+        wait "$spid"; crc=$?
+        say "phase=calib serve end rc=$crc (the rig exited before it ever answered /health)"
+        echo "FAIL: the rig boot died (rc=$crc) — nothing was pushed; read /home/dank/logs/qwen38_bf16_calib.log" >&2
+        exit 1
+    fi
+    if [ "$waited" -ge "$HEALTH_TIMEOUT" ]; then
+        say "phase=calib health never came after ${HEALTH_TIMEOUT}s; killing the rig"
+        podman stop "$RIG" >/dev/null 2>&1 || true
+        wait "$spid"
+        echo "FAIL: the rig never answered /health within ${HEALTH_TIMEOUT}s — nothing was pushed" >&2
+        exit 1
+    fi
+    sleep 15
+    waited=$((waited + 15))
+done
+say "phase=calib health ok waited=${waited}s"
+
+# ---- phase 2: the tap -------------------------------------------------------
+say "phase=calib push begin"
+"$PY" scripts/calib-push.py --port 7770 --parallel 8 --container "$RIG" \
+    --tap-dir "$CALIB" --corpus "$CORPUS" --max-rounds "$ROUNDS"
+prc=$?
+say "phase=calib push end rc=$prc"
+
+podman stop "$RIG" >/dev/null 2>&1 || true
+wait "$spid"; crc=$?
+say "phase=calib serve end rc=$crc"
+
+if [ "$prc" -ne 0 ] || [ "$crc" -ne 0 ]; then
+    echo "FAIL: serve rc=$crc push rc=$prc — the tap is NOT banked and $RAD stays: $(basename "$RAD")" >&2
+    exit 1
+fi
+[ -n "$(ls -A "$CALIB" 2>/dev/null)" ] || { echo "FAIL: both phases exited 0 but the tap dir $CALIB is empty" >&2; exit 1; }
+say "TAP BANKED: $(ls -1 "$CALIB" | wc -l) file(s) in $CALIB"
+
+# ---- phase 3: bank the reference container ----------------------------------
+if [ "${Q38_WINDOW_KEEP_RAD:-0}" = "1" ]; then
+    say "reference .rad kept (Q38_WINDOW_KEEP_RAD=1)"
+else
+    T="$HOME/.AGENTTRASH/$(date +%Y%m%d)-qwen38-bf16-rad"
+    mkdir -p "$T" || { echo "FAIL: cannot create trash dir $T" >&2; exit 1; }
+    mv "$RAD" "$T/" && say "reference .rad banked in $T (the checkpoint that rebuilds it stays)"
+fi
+say "calib window end"

scripts/qwen38_common.py

@@ -0,0 +1,94 @@
+#!/usr/bin/env python3
+"""Shared plumbing for the qwen38 campaign corpus tools.
+
+The three tools move the same two shapes — a JSONL of prompt records and a
+one-artifact JSON manifest — and they all tokenize with the SAME tokenizer. That
+last part is not a detail: a KLD reference is only meaningful against the exact
+token sequence the corpus recorded, so `n_tokens` exists once, here, and every
+tool counts the same way. A window that decode/encode does not round-trip is a
+refusal, never a warning.
+
+Loaded as a sibling module. Each tool puts its own directory on `sys.path`
+first, so `python3 scripts/<tool>.py` and the driver's container mount
+(`runpy.run_path('/work/<tool>.py')`, `-v scripts:/work:ro`) both resolve it —
+run the tools by path, not piped through stdin.
+"""
+import contextlib
+import json
+import os
+import sys
+import tempfile
+
+
+def refuse(msg: str) -> None:
+    """Exit 2 with a REFUSED line — the code argparse uses for unusable input."""
+    print(f"REFUSED: {msg}", file=sys.stderr)
+    sys.exit(2)
+
+
+def load_tokenizer(path: str):
+    try:
+        from tokenizers import Tokenizer
+    except ImportError as exc:  # the corpus tools carry their own venv
+        refuse(f"tokenizers is not importable by {sys.executable} — run the "
+               f"corpus venv's python (on dankserver: "
+               f"/home/dank/venvs/radcorpus/bin/python) or set "
+               f"FRANK_QWEN38_TOKENIZER: {exc}")
+    if not os.path.isfile(path):
+        refuse(f"no tokenizer at {path}")
+    return Tokenizer.from_file(path)
+
+
+def n_tokens(tok, text: str) -> int:
+    """The corpus's own count: no special tokens added, prompt side only."""
+    return len(tok.encode(text, add_special_tokens=False).ids)
+
+
+def iter_jsonl(path: str):
+    """Yield `(lineno, record)` for every non-blank line; refuse on a bad line.
+
+    Field validation stays with the caller: the three corpora carry different
+    fields (`prompt` alone, `prompt`+`cls`, `prompt`+`score_from`+`source`), and
+    a record missing the one the tool needs is that tool's refusal to make.
+    """
+    with open(path, encoding="utf-8", errors="replace") as fh:
+        for lineno, line in enumerate(fh, 1):
+            if not line.strip():
+                continue
+            try:
+                rec = json.loads(line)
+            except json.JSONDecodeError as exc:
+                refuse(f"{path}:{lineno} is not JSON: {exc}")
+            yield lineno, rec
+
+
+def write_jsonl(path: str, rows) -> None:
+    with open(path, "w", encoding="utf-8") as fh:
+        for row in rows:
+            fh.write(json.dumps(row, ensure_ascii=False) + "\n")
+
+
+@contextlib.contextmanager
+def atomic_writer(path: str):
+    """Write `path` through a temporary sibling and rename on success.
+
+    An interrupted run then leaves no half corpus behind, and the temporary name
+    keeps the one-artifact-one-name law the driver applies to a score report.
+    """
+    d = os.path.dirname(os.path.abspath(path)) or "."
+    fd, tmp = tempfile.mkstemp(dir=d, prefix="." + os.path.basename(path) + "-")
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as fh:
+            yield fh
+        os.replace(tmp, path)
+    except BaseException:
+        if os.path.exists(tmp):
+            os.unlink(tmp)
+        raise
+    os.chmod(path, 0o644)
+
+
+def write_json(path: str, obj) -> None:
+    with atomic_writer(path) as fh:
+        json.dump(obj, fh, indent=1)
+        fh.write("\n")

scripts/rechunk-kld-corpus.py

@@ -38,12 +38,19 @@ one-artifact-one-name law run-qwen38-bf16.sh applies to a score report. The file
 is written through a temporary name and renamed, so an interrupted run leaves no
 half corpus behind. A JSON summary of what was written goes to stdout.
 """
-
 import argparse
+import dataclasses
 import json
 import os
 import sys
-import tempfile
+
+_HERE = os.path.dirname(os.path.abspath(__file__)) if "__file__" in globals() else os.getcwd()
+if _HERE not in sys.path:
+    sys.path.insert(0, _HERE)
+
+import qwen38_common as qc
+
+MISSING = object()  # "the parent record carried no source field"
 
 
 def parse_args() -> argparse.Namespace:
@@ -69,95 +76,96 @@ def parse_args() -> argparse.Namespace:
     return ap.parse_args()
 
 
-def load_tokenizer(path: str):
+@dataclasses.dataclass(frozen=True)
-    try:
+class Window:
-        from tokenizers import Tokenizer
+    """One scoring document cut out of a parent record."""
-    except ImportError as exc:  # the corpus scripts carry their own venv
+    prompt: str
-        raise SystemExit(
+    score_from: int
-            f"REFUSED: tokenizers is not importable by {sys.executable} — run the "
+    n_tokens: int
-            f"corpus venv's python (on dankserver: /home/dank/venvs/radcorpus/bin/python) "
+    clamped: bool
-            f"or set FRANK_QWEN38_TOKENIZER: {exc}") from None
+    source: object = MISSING
-    if not os.path.isfile(path):
+
-        raise SystemExit(f"REFUSED: no tokenizer at {path}")
+    def record(self) -> dict:
-    return Tokenizer.from_file(path)
+        rec = {"prompt": self.prompt, "score_from": self.score_from}
+        if self.source is not MISSING:
+            rec["source"] = self.source
+        return rec
 
 
-def refuse(msg: str) -> None:
+def validate(a: argparse.Namespace) -> None:
-    print(f"REFUSED: {msg}", file=sys.stderr)
+    """Everything that makes this run unusable, named before a token is cut."""
-    sys.exit(2)  # 2 is the code argparse uses for unusable arguments
-
-
-def main() -> None:
-    a = parse_args()
     if a.max_doc_tokens <= 0:
-        refuse(f"--max-doc-tokens must be positive (got {a.max_doc_tokens})")
+        qc.refuse(f"--max-doc-tokens must be positive (got {a.max_doc_tokens})")
     if a.score_tail <= 0:
-        refuse(f"--score-tail must be positive (got {a.score_tail})")
+        qc.refuse(f"--score-tail must be positive (got {a.score_tail})")
     if a.max_doc_tokens <= a.score_tail:
-        refuse(f"--max-doc-tokens {a.max_doc_tokens} cannot hold a {a.score_tail}-position "
+        qc.refuse(f"--max-doc-tokens {a.max_doc_tokens} cannot hold a "
-               "score tail: every window would be all tail and no context")
+                  f"{a.score_tail}-position score tail: every window would be all "
+                  "tail and no context")
     if not os.path.isfile(a.corpus):
-        refuse(f"no corpus at {a.corpus}")
+        qc.refuse(f"no corpus at {a.corpus}")
     if os.path.realpath(a.out) == os.path.realpath(a.corpus):
-        refuse(f"--out {a.out} is the corpus itself; move it to trash first — the trash law "
+        qc.refuse(f"--out {a.out} is the corpus itself; move it to trash first — the "
-               "means the old segmentation stays recoverable")
+                  "trash law means the old segmentation stays recoverable")
     if os.path.exists(a.out) and os.path.getsize(a.out) > 0 and not a.force:
-        refuse(f"{a.out} exists and is not empty; pass --force after moving it to trash")
+        qc.refuse(f"{a.out} exists and is not empty; pass --force after moving it to trash")
 
-    tok = load_tokenizer(a.tokenizer_json)
 
-    docs_in = docs_out = tokens_in = tokens_out = 0
+def cut(rec: dict, tok, a: argparse.Namespace,
-    longest_child = 0
+        where: str) -> tuple[list[Window], int]:
-    clamped = scored = 0
+    """Cut one record's token ids into windows, proving each one round-trips.
-    sources: dict[str, int] = {}
+
-    fd, tmp = tempfile.mkstemp(dir=os.path.dirname(os.path.abspath(a.out)) or ".",
+    The round-trip encode is the whole reason this tool may claim the reference
-                              prefix=".rechunk-")
+    scores the corpus's own token sequence: decoding a window to text and
-    try:
+    encoding it back must return the identical ids, or the score is of a
-        with os.fdopen(fd, "w", encoding="utf-8") as out:
+    different document than the one on disk.
-            with open(a.corpus, encoding="utf-8") as src:
+    """
-                for lineno, line in enumerate(src, 1):
+    ids = tok.encode(rec["prompt"], add_special_tokens=False).ids
-                    if not line.strip():
+    windows: list[Window] = []
-                        continue
-                    try:
-                        rec = json.loads(line)
-                    except json.JSONDecodeError as exc:
-                        refuse(f"{a.corpus}:{lineno} is not JSON: {exc}")
-                    prompt = rec.get("prompt")
-                    if not isinstance(prompt, str):
-                        refuse(f"{a.corpus}:{lineno} has no string 'prompt' field")
-                    ids = tok.encode(prompt, add_special_tokens=False).ids
-                    docs_in += 1
-                    tokens_in += len(ids)
     for start in range(0, max(len(ids), 1), a.max_doc_tokens):
         window = ids[start:start + a.max_doc_tokens]
         if not window:
             continue
-                        score_from = max(len(window) - a.score_tail, 0)
+        text = tok.decode(window)
-                        if len(window) < a.score_tail:
+        if tok.encode(text, add_special_tokens=False).ids != window:
-                            clamped += 1
+            qc.refuse(f"{where} window at token {start} does not round-trip through "
-                        scored += len(window) - score_from
+                      "text — the reference would score a different token sequence "
-                        child = {"prompt": tok.decode(window),
+                      "than the corpus holds")
-                                 "score_from": score_from}
+        windows.append(Window(prompt=text,
-                        if "source" in rec:
+                              score_from=max(len(window) - a.score_tail, 0),
-                            child["source"] = rec["source"]
+                              n_tokens=len(window),
-                        out.write(json.dumps(child, ensure_ascii=False) + "\n")
+                              clamped=len(window) < a.score_tail,
+                              source=rec["source"] if "source" in rec else MISSING))
+    return windows, len(ids)
+
+
+def rechunk(a: argparse.Namespace, tok) -> dict:
+    """Write the re-chunked corpus and return the summary it printed."""
+    docs_in = docs_out = tokens_in = tokens_out = 0
+    longest_child = clamped = scored = 0
+    sources: dict[str, int] = {}
+    with qc.atomic_writer(a.out) as out:
+        for lineno, rec in qc.iter_jsonl(a.corpus):
+            where = f"{a.corpus}:{lineno}"
+            if not isinstance(rec.get("prompt"), str):
+                qc.refuse(f"{where} has no string 'prompt' field")
+            windows, n_parent = cut(rec, tok, a, where)
+            docs_in += 1
+            tokens_in += n_parent
+            for w in windows:
+                out.write(json.dumps(w.record(), ensure_ascii=False) + "\n")
                 docs_out += 1
-                        tokens_out += len(window)
+                tokens_out += w.n_tokens
-                        longest_child = max(longest_child, len(window))
+                longest_child = max(longest_child, w.n_tokens)
+                clamped += 1 if w.clamped else 0
+                scored += w.n_tokens - w.score_from
                 if "source" in rec:
                     for cls in str(rec["source"]).split("+"):
                         sources[cls] = sources.get(cls, 0) + 1
     if tokens_out != tokens_in:
         raise SystemExit(f"REFUSED: the windows hold {tokens_out} tokens and the corpus "
                          f"has {tokens_in} — re-chunking lost or invented tokens")
-        os.replace(tmp, a.out)
+    return {
-    except BaseException:
-        if os.path.exists(tmp):
-            os.unlink(tmp)
-        raise
-    os.chmod(a.out, 0o644)
-
-    print(json.dumps({
         "corpus": a.corpus,
         "out": a.out,
         "tokenizer": os.path.basename(a.tokenizer_json),
@@ -171,7 +179,13 @@ def main() -> None:
         "scored_positions": scored,
         "clamped_score_from": clamped,
         "source_documents": sources,
-    }, indent=1))
+    }
+
+
+def main() -> None:
+    a = parse_args()
+    validate(a)
+    print(json.dumps(rechunk(a, qc.load_tokenizer(a.tokenizer_json)), indent=1))
 
 
 if __name__ == "__main__":

scripts/run-qwen38-bf16.sh

@@ -92,7 +92,8 @@
 #        the container reads through /models), FRANK_QWEN38_LOG_DIR, and the step
 #        and slot counts inherited as FRANK_ARCHON_MAX_BATCHED_TOKENS /
 #        FRANK_ARCHON_MAX_SEQS — no per-mode literals (the audit law: a mode that
-#        carries its own step measures a shape nothing serves) —
+#        carries its own step measures a shape nothing serves; calib is exempt, see
+#        Q38_CALIB_MNBT / Q38_CALIB_SEQS below — its output is never compared) —
 #        and FRANK_QWEN38_MAX_MODEL_LEN — the window every mode boots at (32,768;
 #        see the corpus census above, and raise it together with the corpus).
 # kldref refuses (exit 2) when the longest record of the corpus it is about to
@@ -120,7 +121,8 @@ while [ $# -gt 0 ]; do
     case "$1" in
         kldref|calib|score) MODE="$1" ;;
         --dry-run)          DRY=1 ;;
-        --image)            IMAGE="$2"; shift ;;
+        --image)            [ "$#" -ge 2 ] || { echo "--image needs a reference" >&2; exit 2; }
+                            IMAGE="$2"; shift ;;
         *)
             # score's two positional arguments, and nothing else may be bare.
             if [ "$MODE" = score ] && [ "${#POS[@]}" -lt 2 ]; then
@@ -170,19 +172,38 @@ fi
 
 # ONE step size for every mode, quoted from the file that owns it — the audit
 # law for a radiance rig is to inherit the proven number, not to carry a
-# campaign-private value: production's FRANK_ARCHON_MAX_BATCHED_TOKENS is 2,048,
+# campaign-private value: production's FRANK_ARCHON_MAX_BATCHED_TOKENS. The
-# the widest step the quantised expert FILE tiers can stream (above it the
+# 1,024 default this script carried was PR #80's answer to a starved cache;
-# engine charges a 10.65 GiB arena to the card and the load refuses), and the
+# production moved to its widest legal step (2,048) on 2026-10-29 and the rig
-# arena it charges comes out of the cache — so the corpora are cut to the window
+# follows. The pool raise that shipped with it (66,560 a rank) bought this
-# the boot ledger reports, never the other way round. The 1,024 default this
+# container 1.26 MiB of cache (`vram_kv 269.99 -> 271.25 MiB`, window still
-# script carried was PR #80's answer to a starved cache; the pool raise
+# 7,436 tokens — the 95.37 GiB n-gram table's claim decides the rig's cache,
-# (frank-archon.env: 66,560 MiB a rank) is the fix, not a narrower step.
+# not the pool) and was folded back the same day. The corpora stay cut to the
+# window the boot ledger reports, never the other way round.
 MNBT="${FRANK_ARCHON_MAX_BATCHED_TOKENS:?FRANK_ARCHON_MAX_BATCHED_TOKENS must be set}"
 if ! [[ "$MNBT" =~ ^[0-9]+$ ]]; then
     echo "REFUSED: FRANK_ARCHON_MAX_BATCHED_TOKENS must be a token count (got: $MNBT)" >&2
     exit 2
 fi
 
+# Calib's shape is NOT a measuring shape. kldref and score produce KL numbers that are only
+# comparable across boots at an identical shape, which is why they inherit production's step and
+# slots to the letter. The tap's output is layer Hessians, never a number any report compares, so
+# its step and slot count are its own — and they are the tap's dominant cost: a 7,168-token
+# document cut into four 2,048-token steps re-streams each layer's routed expert union four
+# times from the container, so one wide step is worth up to ~4x the read volume. Measured
+# 2026-10-29 at the inherited 2,048/4: prefill 71 tok/s and a 47-hour tap; the probe exists to
+# find out what the wide step is worth before the campaign pays for two days of drained serving.
+# Both default to the inherited value, so nothing moves unless a probe asks it to.
+CALIB_MNBT="${Q38_CALIB_MNBT:-$MNBT}"
+CALIB_SEQS="${Q38_CALIB_SEQS:-$MAX_SEQS}"
+for pair in "Q38_CALIB_MNBT=$CALIB_MNBT" "Q38_CALIB_SEQS=$CALIB_SEQS"; do
+    if ! [[ "${pair#*=}" =~ ^[0-9]+$ ]]; then
+        echo "REFUSED: ${pair%%=*} must be a number (got: ${pair#*=})" >&2
+        exit 2
+    fi
+done
+
 # The window every mode boots at. Over-length is a REFUSAL, not a truncation (the
 # env file says so for production), so this is the number the corpus must fit in.
 MAX_MODEL_LEN="${FRANK_QWEN38_MAX_MODEL_LEN:-32768}"
@@ -191,6 +212,25 @@ if ! [[ "$MAX_MODEL_LEN" =~ ^[0-9]+$ ]]; then
     exit 2
 fi
 
+# Where the n-gram table lives. On the bf16 payload it is 95.37 GiB and `ram`
+# PINS it: the calibration boot died at "startup stopped: out of memory" with
+# 230.22 GiB pinned of 246.07 GiB and 8.06 GiB available (measured 2026-10-29),
+# while the reference recording and every score boot had already succeeded with
+# the same pool and slots. The difference that matters is the job: calib only
+# needs the routed weights prefilled, so it reads the table from the container's
+# own file tier and the pin never happens; kldref and score keep production's
+# placement, because their numbers are only comparable when every other term of
+# the shape is identical. FRANK_QWEN38_NGRAM_PLACEMENT overrides either default.
+case "$MODE" in
+    calib) NGRAM_PLACEMENT="${FRANK_QWEN38_NGRAM_PLACEMENT:-disk}" ;;
+    *)     NGRAM_PLACEMENT="${FRANK_QWEN38_NGRAM_PLACEMENT:-ram}" ;;
+esac
+case "$NGRAM_PLACEMENT" in
+    auto|vram|ram|disk) ;;
+    *) echo "REFUSED: FRANK_QWEN38_NGRAM_PLACEMENT must be auto|vram|ram|disk (got: $NGRAM_PLACEMENT)" >&2
+       exit 2 ;;
+esac
+
 # Every container this campaign boots is read through the ro /models mount, so
 # the host path must live under it; the container-side spelling is what goes
 # into the argv. Same rule for the container a score pass opens.
@@ -228,7 +268,7 @@ ARGS=(--tp 2 --placement expert_tiered --weights-disk-tier
       --host-pool-mib "$HOST_POOL_MIB" --gpu-headroom-mib "$GPU_HEADROOM_MIB"
       --deterministic --num-speculative-tokens 0
       --kv-cache-dtype fp8 --max-model-len "$MAX_MODEL_LEN"
-      --ngram-placement ram
+      --ngram-placement "$NGRAM_PLACEMENT"
       --override-chat-template /models/tpl.jinja)
 MOUNTS=(-v "$MODELS:/models:ro")
 TAPENV=()
@@ -249,7 +289,8 @@ case "$MODE" in
     # time the engine reads a recording and writes a report instead of the
     # reverse. The reference dir mounts READ-ONLY — a scoring pass must not be
     # able to touch the zero point — and the report gets its own writable bind.
-    if [ "$OUT_DIR" = "$KLD_REF_DIR" ] || same_dir "$OUT_DIR" "$KLD_REF_DIR"; then
+    if [ "$OUT_DIR" = "$KLD_REF_DIR" ] || same_dir "$OUT_DIR" "$KLD_REF_DIR" \
+       || [ "$(realpath -m "$OUT_DIR")" = "$(realpath -m "$KLD_REF_DIR")" ]; then
         echo "REFUSED: score output $SCORE_OUT is inside the reference dir $KLD_REF_DIR, which this mode mounts read-only" >&2
         exit 2
     fi
@@ -274,10 +315,29 @@ case "$MODE" in
     # VRAM knob. The slots follow frank-archon.env's FRANK_ARCHON_MAX_SEQS
     # through MAX_SEQS, exactly as kldref and score do.
     ARGS+=(--model "$MODEL_IN" --port 7770
-           --max-num-seqs "$MAX_SEQS" --max-num-batched-tokens "$MNBT")
+           --max-num-seqs "$CALIB_SEQS" --max-num-batched-tokens "$CALIB_MNBT")
+    # The tap is only real if this directory exists and the engine can write it:
+    # without the check a typo'd CALIB_DIR auto-creates a root-owned empty dir in
+    # whatever path was given, and the run burns hours serving prompts into a tap
+    # nobody reads. RADIANCE_CALIB_TOKENS is the whole corpus' floor, not a round's.
+    [ "$DRY" = 1 ] || [ -w "$CALIB_DIR" ] || {
+        echo "REFUSED: calib tap dir $CALIB_DIR is not writable (mkdir it, chown it, then re-run)" >&2
+        exit 2
+    }
     MOUNTS+=(-v "$CALIB_DIR:/data/calib")
     TAPENV=(-e RADIANCE_CALIB_DIR=/data/calib
-            -e RADIANCE_CALIB_TOKENS=12000000)
+            -e "RADIANCE_CALIB_TOKENS=${FRANK_QWEN38_CALIB_TOKENS:-12000000}")
+    # The tap's own authorization. A pusher must see this marker before it sends
+    # a prompt (calib-push.py --tap-marker), it names the container so a push
+    # aimed anywhere else refuses, and a boot that fails removes it again below.
+    # Measured 2026-10-29: with no marker, the chain pushed 4 rounds x 10M tokens
+    # into RESTORED PRODUCTION after the rig's boot died, and nothing recorded.
+    if [ "$DRY" = 1 ]; then
+        :
+    else
+        printf 'container=%s\nimage=%s\nmode=calib\nstarted=%s\npid=%s\n' \
+            "$C" "$IMAGE" "$(date -Is)" "$$" > "$CALIB_DIR/.rig-boot"
+    fi
     ;;
 esac
 
@@ -351,7 +411,14 @@ cp -f "$TEMPLATE" "$TPL" || {
 [ "$MODE" = kldref ] && mkdir -p "$KLD_REF_DIR"
 [ "$MODE" = calib ]  && mkdir -p "$CALIB_DIR"
 
-restore() { sudo systemctl start "$SERVICE" >/dev/null 2>&1 || true; }
+# Stop the campaign container BEFORE handing the cards back: a killed run leaves
+# its conmon alive (--rm only fires when podman itself exits), and production
+# then boots beside a container still holding the GPUs — measured 2026-10-29,
+# where the service reported `active` while its engine starved on the rig's VRAM.
+restore() {
+    podman rm -f "$C" >/dev/null 2>&1 || true
+    sudo systemctl start "$SERVICE" >/dev/null 2>&1 || true
+}
 trap restore EXIT
 echo "run-qwen38-bf16 $MODE: stopping $SERVICE ($(date -Is)); log: $LOG"
 sudo systemctl stop "$SERVICE" >/dev/null 2>&1 || true
@@ -359,5 +426,12 @@ podman rm -f "$C" >/dev/null 2>&1 || true
 
 "${CMD[@]}" 2>&1 | tee "$LOG"
 rc=${PIPESTATUS[0]}
+# A boot that failed must not leave the tap marked as rig-authorized: the marker
+# is what calib-push.py trusts, and the point of it is that a dead rig can never
+# be pushed into again.
+if [ "$MODE" = calib ] && [ "$rc" -ne 0 ] && [ -f "$CALIB_DIR/.rig-boot" ]; then
+    mv -f "$CALIB_DIR/.rig-boot" "$CALIB_DIR/.rig-boot.failed-$(date +%Y%m%d-%H%M%S)" 2>/dev/null || true
+    echo "the rig boot failed (rc=$rc): the tap marker is retired, so no pusher can take this tap for a live rig"
+fi
 echo "run-qwen38-bf16 $MODE exited rc=$rc; $SERVICE restored; log: $LOG"
 exit "$rc"

scripts/stack-qwen38-official-fp8.py

@@ -337,8 +337,12 @@ def main():
     if any(meta[shard_ids[i]][4][1] != hd for i in range(n_shards)):
         log("ERROR: PLE shards disagree on HD")
         return 2
-    if hd % 128 != 0 and hd != 160:
+    if hd != 160 and hd % 128 != 0:
-        log(f"WARNING: PLE HD={hd} is neither 160 nor a multiple of 128")
+        # The pack below writes lanes of 128 codes; any other HD would produce a
+        # table that decodes to the wrong rows. The old warning let it continue.
+        log(f"ERROR: PLE HD={hd} is neither 160 nor a multiple of 128, and the "
+            "128-lane pack cannot express it")
+        return 2
 
     # ---- idempotence gate -------------------------------------------------
     stacked_names = [f"{p}.mlp.experts.gate_up_proj" for p in prefixes]
@@ -521,12 +525,16 @@ def verify(stage, prefixes, n_e, shapes, meta, shard_ids, rows, hd,
         }
         for name, (dt, shp) in want.items():
             e = hdr[name]
-            assert e["dtype"] == dt and list(e["shape"]) == shp, name
+            if e["dtype"] != dt or list(e["shape"]) != shp:
-            assert e["data_offsets"][1] - e["data_offsets"][0] == \
+                raise RuntimeError(f"{name}: header is {e['dtype']}{e['shape']}, expected {dt}{shp}")
-                int(np.prod(shp)) * (1 if dt == F8 else 2), name
+            if e["data_offsets"][1] - e["data_offsets"][0] != int(np.prod(shp)) * (1 if dt == F8 else 2):
+                raise RuntimeError(f"{name}: data span is {e['data_offsets']}, shape {shp} needs "
+                                   f"{int(np.prod(shp)) * (1 if dt == F8 else 2)} B")
     _, phdr = parse_header(out_path)
     e = phdr[next(iter(phdr))]
-    assert e["dtype"] == BF16 and list(e["shape"]) == [sum(rows), hd]
+    if e["dtype"] != BF16 or list(e["shape"]) != [sum(rows), hd]:
+        raise RuntimeError(f"PLE table header is {e['dtype']}{e['shape']}, "
+                           f"expected {BF16}[{sum(rows)}, {hd}]")
 
     # sampled PLE rows: recompute and compare bytes exactly
     lut = e4m3_lut()
@@ -546,7 +554,8 @@ def verify(stage, prefixes, n_e, shapes, meta, shard_ids, rows, hd,
                 with open(out_path, "rb") as ofh:
                     ofh.seek(out_off + got_row * hd * 2)
                     got = ofh.read(hd * 2)
-                assert got == want, f"PLE row {got_row} mismatch"
+                if got != want:
+                    raise RuntimeError(f"PLE row {got_row} does not match the decoded source")
             mm.close()
     log("verify: new-file headers, sizes and sampled PLE rows all exact; "
         "expert slabs were byte-checked in-stack")

serverAppConfigs-docs/LLM/frank-archon.md

@@ -91,27 +91,36 @@ on stop. Boot budget 1800 s (the publisher's compose allows a 15-minute `start_p
 | Slots / window | `MAX_SEQS=4`, `MAX_MODEL_LEN=200000` (native 262,144; longer is refused, never truncated) — this pair is the **cache bound**: the resolver sizes the KV carve against `slots × window`, and no slot count funds itself (Laws) |
 | KV dtype | `KV_DTYPE=fp8` — the only dtype that funds the slot count (Laws) |
 | VRAM split | **derived at boot** from `GPU_HEADROOM_MIB=96` + the cache bound + `HOST_POOL_MIB`; naming `VRAM_WEIGHTS_MIB`/`VRAM_KV_MIB` is a boot refusal (removed at 1.3.0) |
-| Expert tier | `PLACEMENT=expert_tiered`, `HOST_POOL_MIB=66560` **per rank** (raised 2026-10-29 from 49152 — see *Eras*) — the pinned host pool is also the **KV lever**: the expert floor is `plane + 2 units − host room`, so every MiB pinned there lowers what the experts claim on the card one-for-one and the cache takes it (Laws) |
+| Expert tier | `PLACEMENT=expert_tiered`, `HOST_POOL_MIB=49152` **per rank** (raised to 66560 on 2026-10-29, folded back the same day — the A/B under *Laws* shows the raise bought no cache) — the pinned host pool is also the **KV lever**: the expert floor is `plane + 2 units − host room`, so every MiB pinned there lowers what the experts claim on the card one-for-one and the cache takes it (Laws) |
 | PLE | `NGRAM_PLACEMENT=ram` (47.7 GiB pinned; refuses boot on a short memlock) |
 | Spec / batch | `SPEC_TOKENS=2` — MTP depth 2, the `.rad` carries the model's own `mtp.*` head (`mtp.fc_hidden`/`fc_embedding`/`draft_head`); `MAX_BATCHED_TOKENS=2048` (MNBT — 2560 collapses decode, Laws) |
 | Prefix cache | `PREFIX_CACHE_MIB=8192`, pinned **host tier only** — the 64 GiB disk tier is gone, so an evicted prefix is re-prefilled; `STATE_DIR=/home/dank/radiance-state` now exists only to hold the rad-tune cache |
 | Template | `TEMPLATE=scripts/qwen38-froggeric-v22.5.jinja` (28,234 B) via `--override-chat-template` — froggeric's fixed Qwen3.8 template (the repo root's `chat_template.jinja`, vendored); the checkpoint's own template is baked into the `.rad` by rad-convert and this overrides it, and radiance derives the reasoning and tool-call reply format from it |
 | Step sharing | `BUSY_PREFILL_CHUNK=auto` — prefill-speed first: it reserves 3 tokens per waiting decoder and ⅛ of the budget per waiting request, and does **not** shrink the chunk for decoders alone · `PREFILL_INTERLEAVE=1` — the decoder dial, **armed 2026-10-20**: one decode-only step after each chunk step taken beside decoders; meter `engine_steps_held_total`, price is the prefilling prompt's TTFT |
-| TTM gate | `TTM_MIN_MIB=197000`, `TTM_TARGET_MIB=224000` |
+| TTM gate | `TTM_MIN_MIB=162000`, `TTM_TARGET_MIB=204800` |
 | Boot | `RAD_MIN_MIB=110000` deployer floor; the listener binds only after the full load (health ~100 s, budget 1800 s) |
 
-Boot ledger for the live shape (2026-10-20, pin 1.8.1, `MAX_SEQS=4`, `HOST_POOL_MIB=49152`):
+Boot ledger for the live payload. Last boot: 2026-10-29 at `MAX_SEQS=4`,
-the resolver prints `elastic 26.73 GiB` → **experts 20.72 GiB (78 %) / kv cache 5.64 GiB (21 %)**,
+`HOST_POOL_MIB=66560`, `MAX_BATCHED_TOKENS=2048`; the live pool is `49152` again, and the
-`prefill staging 373.65 MiB` out of the cache, `vram_weights 24.30 GiB`, `host_pool 48.59 GiB`.
+two boots are **identical above `host_pool`**, so one ledger serves both shapes:
-The referee line: `kv: the cache can grow to 5.64 GiB a card — 747,492 tokens of context, 3.7
+`elastic 26.73 GiB` → **experts 20.72 GiB (78 %) / kv cache 5.64 GiB (21 %)**,
-sessions at 200,000 — with every block it lends the experts taken back; 440.04 MiB of that is
+`activation arena 990.36 MiB` (the price of the 2,048-token step: `256 tokens use
-per-sequence state, and up to 5.20 GiB is lent to the experts while no request holds it`.
+155.60 MiB, 1024 tokens use 472.86 MiB of 896.20 MiB`), `prefill staging 373.65 MiB`,
-`--checkpoint-slots 40` is auto (`max_num_seqs × 10` slots of 111.98 MiB = 4,479 MiB of pinned
+`vram_weights 24.30 GiB`, `host_pool 65.59 GiB` (opened with `0 B used`; `48.59 GiB` at
-host), `/health` reports `queue_cap 32` (`max(8, seqs × 8)`), sampler **12 rows**
+the 49,152 boot), `unclaimed 96.89 MiB`; referee line `kv: the pool backs 93% of the worst
-(= `seqs × (1 + speculative depth)`), 4 hardware queues a card, template 28,234 B, vision live
+case (4 sequences all at 200000 tokens), 3.7 full-length sessions`; `device: the 2 cards
-(≤ 4,194,304 px, 16,384 patches/pass), P2P confirmed (`the 2 cards reach each other's memory
+reach each other's memory directly (peer to peer)`; sampler **12 rows**
-directly`). Experts living in the pinned host pool: **8.99 GiB a rank** (was 5.99) — the price of
+(= `seqs × (1 + speculative depth)`), 4 hardware queues a card (133.45 MiB), `already held
-the bigger cache, paid over PCIe when those experts are hit.
+256.67 MiB`, mover `17519 slab slots, 7624 pinned pool slots, 3.58 GiB static vram`,
+`load: rank 0, mover took 24.30 GiB`.
+
+The A/B that ended the raise: the 2026-10-20 boot at `HOST_POOL_MIB=49152` printed the
+**same** `elastic 26.73 / experts 20.72 / kv cache 5.64 / vram_weights 24.30` (Laws: the
+lever saturates at the expert floor), and the campaign rig — the other payload that
+inherits this env — moved `vram_kv 269.99 → 271.25 MiB`, a rounding step, because its
+window is set by its own n-gram table (95.37 GiB) and expert plane rather than by the
+pool. The refused calibration documents are fixed by cutting them to the window, not by
+widening the pool.
 
 The boot **ledger**, not this arithmetic, is the referee: radiance states every budget before
 anything is allocated (weights vs KV, the host pool's coverage of the worst case, the PLE pin,
@@ -150,7 +159,7 @@ marked "re-fetch" is gone from `/home/dank/models/`, which today holds only `Sti
 
 | Window | Payload | Runtime | Shape | Replaced because | Rollback |
 |---|---|---|---|---|---|
-| 2026-10-20 → 10-29 | same `qwen3.8-next-flash-fp8-iq4r-moe.rad` (payload unchanged) | radiance **1.8.1**, `sha256:d590e300…` (same digest) | 4 × 200,000 fp8 KV — resolver cache 5.64 GiB = 747,492 tokens, `prefill staging 373.65 MiB`, `vram_weights 24.30 GiB`, `host_pool 48.59 GiB`; host pool 49,152 MiB a rank, `TTM_MIN_MIB=162000` / `TTM_TARGET_MIB=204800`, `queue_cap 32`, sampler 12 rows, `--prefill-interleave 1` | the `qwen38-w5nl64` campaign rig booted this pool at the shared 1,024 step and reported `kv cache 269.99 MiB a card: 7436 tokens`, so 682 of its 1,294 calibration documents were refused; Frank ordered the pool **and** the step raised to production's widest legal step (2,048) as one shape | revert the `feat/frank-archon-pool-step` commit and redeploy `scripts/deploy.sh dankserver --service frank-archon` — trio back to `HOST_POOL_MIB=49152`, `TTM_MIN_MIB=162000`, `TTM_TARGET_MIB=204800` |
+| 2026-10-20 → 10-29 | same `qwen3.8-next-flash-fp8-iq4r-moe.rad` (payload unchanged) | radiance **1.8.1**, `sha256:d590e300…` (same digest) | 4 × 200,000 fp8 KV — resolver cache 5.64 GiB = 747,492 tokens, `prefill staging 373.65 MiB`, `vram_weights 24.30 GiB`, `host_pool 48.59 GiB`; host pool 49,152 MiB a rank, `TTM_MIN_MIB=162000` / `TTM_TARGET_MIB=204800`, `queue_cap 32`, sampler 12 rows, `--prefill-interleave 1` | Frank ordered the campaign's step (2,048) onto production; the pool half of that order (66,560) booted, came back byte-identical to this row's cache, and was folded back the same day — the calibration documents it was meant to save were cut to the window instead (`scripts/rechunk-kld-corpus.py`) | the live shape is this row's pool with `MAX_BATCHED_TOKENS=2048`: the pair is the rollback unit, not the merge commit (`scripts/frank-archon.env`) |
 | 2026-10-19 → 10-20 | same `qwen3.8-next-flash-fp8-iq4r-moe.rad` (payload unchanged) | radiance **1.8.1**, `sha256:d590e300…` (same digest) | 2 × 200,000 fp8 KV — resolver cache 2.76 GiB = 365,204 tokens (91 % backing); host pool 44,928 MiB a rank, `TTM_MIN_MIB=150000`, `queue_cap 16`, sampler 9 rows; `--prefill-interleave 1` armed on 10-20 | Frank ordered **four** concurrent conversations over a ≥ 700,000-token pool; the resolver funds that only by lowering the expert floor with more pinned host room (Laws) | revert `1bb6399` and redeploy `scripts/deploy.sh dankserver --service frank-archon` — trio back to `MAX_SEQS=2` / `HOST_POOL_MIB=44928` / `TTM_MIN_MIB=150000` |
 | 2026-10-15 → 10-19 | same `qwen3.8-next-flash-fp8-iq4r-moe.rad` (payload unchanged) | radiance **1.0.13**, `sha256:4c3f4845…` | 3 × 200,000 fp8 KV bought with the **explicit** pair `--vram-weights-mib 25600 --vram-kv-mib 5120`; host pool 44,928 → 45,568 MiB a rank; prefix cache 8 GiB host + 64 GiB disk | decode turned to sludge beside any long prompt — step contention, 18.7 ms decode-only steps against 317 ms steps carrying a prefill chunk (Laws) — and the only two knobs that fix it, `--busy-prefill-chunk` and `--prefill-interleave`, exist only after this pin | `FRANK_ARCHON_IMAGE` back to `sha256:4c3f4845…` **plus a trio rewrite**: restore the pair and the host+disk prefix knobs, delete `BUSY_PREFILL_CHUNK`/`PREFILL_INTERLEAVE`. A digest swap alone is a boot refusal (Runtime) |
 | 2026-10-08 → 10-15 | in-house `qwen4exp-swift15-w4dn8-i8-hc8m.rad`, 116,416 MiB (Swift1.5 abliterated BF16, GPTQ w4 experts + 24 bf16 protected, KLD 0.0131 / top-1 97.92 %) | radiance 1.0.13, same digest | 3 × 200,000 fp8 KV, slab 25,600 / pool 5,120 — the shape the live file still carries | the engine author's own container is the supported path and his quality ceiling; the finetune's refusal-bending left the endpoint | `.rad` purged 2026-10-17 (trash-staged then emptied, Frank's approval). Re-download `frank8465/Swift-1.5-Qwen3.8-Flash-Next-GPTQ-W4A8-Radiance-AMD-R9700` → `qwen4exp-w4nl64-i8-hc8m.rad`, 122,071,973,888 B, LFS sha256 `09dd9306…`, via the iq4r fetch script's hfget/verify pattern retargeted at that repo; point `MODEL_FILE` at it, redeploy. Re-**convert** is dead: the BF16 source was `rm`'d 2026-10-15 (`kld-ref`, grams, corpora and `/home/dank/logs/swift15_*.log` survive) |
@@ -244,7 +253,8 @@ vllm-radiance era (this table opens on 09-26; the era page claims 10-10).
   | 3 × 200,000, host pool 44,928 | 600,000 tok | 4.20 GiB | 93 % |
   | 3 × 200,000, host pool 45,568 | 600,000 tok | **4.20 GiB — byte-identical** | 93 % |
   | 2 × 200,000, host pool 44,928 | 400,000 tok | 2.76 GiB | 91 % |
-  | 4 × 200,000, host pool 49,152 (live) | 800,000 tok | **5.64 GiB = 747,492 tok** | 93 % |
+  | 4 × 200,000, host pool 49,152 (live) | 800,000 tok | 5.64 GiB = 747,492 tok | 93 % |
+  | 4 × 200,000, host pool 66,560 (2026-10-29 A/B, folded back) | 800,000 tok | **5.64 GiB — byte-identical to 49,152** | 93 % |
 
   The 45,568 row is the ceiling case, not a mystery: at 3 slots the cache already held its whole
   `ceiling`, so extra host room could not grow it and shows up as `unclaimed`. **The referee is the
@@ -552,6 +562,29 @@ vllm-radiance era (this table opens on 09-26; the era page claims 10-10).
   fork bomb inside the container. Kill with `podman rm -f` (the cgroup owns every generation)
   never pattern-kill an interpreter name on this host; use `TORCH_LOGS` for dynamo questions.
 
+### The quantization pipeline
+
+- **A calibration tap is only a tap when the rig that opened it is running.** The runner writes
+  `.rig-boot` into the tap dir (container, image, pid) and retires it when its boot exits non-zero;
+  `calib-push.py` refuses without that marker, refuses when it names another container, and refuses
+  when the named container is not running; `qwen38-calib-window.sh` stops on a failed phase instead
+  of falling through its health wait. All three exist because on 2026-10-29 the unguarded chain's
+  rig boot died of an OOM, the health loop expired without failing, and the pusher fed four rounds
+  of 10M tokens into **restored production** — 86 minutes of live prefill that recorded nothing.
+- **The n-gram table's pin is what starves a rig boot, and the tier is a knob.** On the BF16 payload
+  the table is 95.37 GiB, `--ngram-placement ram` pins it, and the calibration boot died at
+  `startup stopped: out of memory` with 230.22 GiB pinned of 246.07 GiB and 8.06 GiB available —
+  while the reference recording and every score boot succeeded at the same pool and slot count.
+  Calibration only needs the routed weights prefilled, so it reads the table from the container's
+  file tier (`--ngram-placement disk`); kldref and score keep production's placement, because their
+  numbers are only comparable at an identical shape.
+- **Acceptance is three metrics and every slice, not a mean.** `scripts/kld-gate-qwen38.sh` grades
+  mean KL, the p99.9 tail, top-1 agreement and every `by_source` slice (code, prose, tools):
+  a mean can hide a few positions taking the whole hit, and an overall win can be bought by
+  regressing one distribution. The FP8-native anchor is an interval, not a requirement — without
+  the 172.8 GiB fetch the gate runs in two-report mode (candidate vs the served payload) and says
+  so on its first line.
+
 ## Contract (the wire)
 
 - **Live surface:** `/health` (== `/ping`), `/v1/models`, `/v1/chat/completions`
@@ -628,11 +661,19 @@ vllm-radiance era (this table opens on 09-26; the era page claims 10-10).
   rebuild), `scripts/download-frank-archon-swift15.sh` (the BF16 source a w4dn8 re-convert
   needs), `scripts/hfget.sh` / `scripts/hf-fetch.sh`; template
   `scripts/qwen38-froggeric-v22.5.jinja`.
-- Convert campaign (dormant, kept for a future quant): `scripts/convert-swift15-w4dn8.sh`
+- The campaign (qwen38-w5nl64, live): `scripts/run-qwen38-bf16.sh` (kldref / calib / score — the
-  `scripts/recipes/qwen4exp-swift15-w4dn8-i8-hc8m.recipe`, `scripts/run-swift15-bf16.sh`
+  ONE measuring shape), `scripts/qwen38-calib-window.sh` (the gated calibration window),
-  `scripts/kld-gate-swift15.sh`, `scripts/kld-then-calib.sh`, `scripts/calib-push.py`
+  `scripts/convert-qwen38-w5nl64.sh` with
-  `scripts/make-calib-corpus.py`, `scripts/derive-protected-experts.py`
+  `scripts/recipes/qwen4exp-qwen38-w5nl64-bf16trunk.recipe` (the arm driver and its template),
-  `scripts/apply-protected-cast.sh`, `scripts/verify-swift15-ple.py`.
+  `scripts/kld-gate-qwen38.sh` (acceptance), `scripts/calib-push.py`,
+  `scripts/make-calib-corpus.py`, `scripts/rechunk-kld-corpus.py`,
+  `scripts/qwen38_common.py` (the tokenizer, JSONL and refusal plumbing those three
+  corpus tools share — one `n_tokens`, so every tool counts the same way),
+  `scripts/derive-protected-experts.py`, `scripts/apply-protected-cast.sh`.
+- Retired swift15-era campaign scripts, superseded by the above and kept only as reference:
+  `scripts/convert-swift15-w4dn8.sh`, `scripts/recipes/qwen4exp-swift15-w4dn8-i8-hc8m.recipe`,
+  `scripts/run-swift15-bf16.sh`, `scripts/kld-gate-swift15.sh`, `scripts/kld-then-calib.sh`,
+  `scripts/verify-swift15-ple.py`.
 - Probes: `scripts/probe-frank-archon-kv.py` (reads slots/window/dtype from the deployed env so
   it cannot drift; leg D = decode), `scripts/frank-archon-expert-activity.py`
   `scripts/autoresearch-frank-archon.sh` + `scripts/autoresearch/frank-archon-measure.py`.

tests/frank_archon_unit.bats

@@ -158,7 +158,7 @@ nosudo_bin() {
     # so the pool's token count is a boot-ledger line — verified live by the
     # capacity leg of scripts/probe-frank-archon-kv.py at every cutover.
     grep -q '^FRANK_ARCHON_PLACEMENT=expert_tiered$' "$ENV_FILE"
-    grep -q '^FRANK_ARCHON_HOST_POOL_MIB=66560$' "$ENV_FILE"
+    grep -q '^FRANK_ARCHON_HOST_POOL_MIB=49152$' "$ENV_FILE"
     grep -q '^FRANK_ARCHON_KV_DTYPE=fp8$' "$ENV_FILE"
     grep -q '^FRANK_ARCHON_NGRAM_PLACEMENT=ram$' "$ENV_FILE"
     # Prefix cache: pinned host RAM only (the disk tier was removed), and the

tests/qwen38_campaign.bats

@@ -200,6 +200,9 @@ DERIVE="$REPO_ROOT/scripts/derive-protected-experts.py"
 APPLY="$REPO_ROOT/scripts/apply-protected-cast.sh"
 CONVERT="$REPO_ROOT/scripts/convert-qwen38-w5nl64.sh"
 GATE="$REPO_ROOT/scripts/kld-gate-qwen38.sh"
+RUNNER="$REPO_ROOT/scripts/run-qwen38-bf16.sh"
+PUSHER="$REPO_ROOT/scripts/calib-push.py"
+WINDOW="$REPO_ROOT/scripts/qwen38-calib-window.sh"
 TEMPLATE="$REPO_ROOT/scripts/recipes/qwen4exp-qwen38-w5nl64-bf16trunk.recipe"
 
 # make_gram <dir> <layer> <expert> <energy> <rows> — a RADGRAM2 down-Gram whose
@@ -525,3 +528,206 @@ gate_case() { # gate_case <kl fp8> <kl iq4r> <kl cand> <top1 cand> [extra flags.
     [[ "$output" == *"has no slice 'code'"* ]]
     [[ "$output" == *"it has: ops"* ]]
 }
+
+# --- the n-gram tier: the pin that starved the calibration boot ---------------
+
+@test "the calibration boot reads the n-gram table from the file tier" {
+    # 2026-10-29: 95.37 GiB pinned beside a 65.59 GiB/rank pool left 8.06 GiB of
+    # RAM and the boot died "startup stopped: out of memory". calib only needs the
+    # routed weights prefilled, so it reads the table off the container's file tier.
+    run bash "$RUNNER" calib --dry-run
+    [ "$status" -eq 0 ]
+    grep -A1 -x -- '--ngram-placement' <<<"$output" | grep -qx 'disk'
+}
+
+@test "the recording and every score pass keep production's n-gram placement" {
+    # Those numbers are only comparable when the shape is identical; the tier is
+    # part of the shape.
+    run bash "$RUNNER" kldref --dry-run
+    [ "$status" -eq 0 ]
+    grep -A1 -x -- '--ngram-placement' <<<"$output" | grep -qx 'ram'
+}
+
+@test "an unknown n-gram placement is refused by name before the boot" {
+    run env FRANK_QWEN38_NGRAM_PLACEMENT=ssd bash "$RUNNER" calib --dry-run
+    [ "$status" -eq 2 ]
+    [[ "$output" == *"auto|vram|ram|disk"* ]]
+}
+
+# --- the pusher: a tap without a live rig is not a tap ------------------------
+
+# stub_podman <dir> <running|absent> — a podman that answers the one question the
+# pusher asks: is the rig container running?
+stub_podman() {
+    mkdir -p "$1/bin"
+    if [ "$2" = running ]; then
+        printf '#!/usr/bin/env bash\necho true\n' > "$1/bin/podman"
+    elif [ "$2" = stopped ]; then
+        printf '#!/usr/bin/env bash\necho false\n' > "$1/bin/podman"
+    else
+        printf '#!/usr/bin/env bash\nexit 125\n' > "$1/bin/podman"
+    fi
+    chmod +x "$1/bin/podman"
+}
+
+pusher_fixture() { # pusher_fixture <marker-container or empty>
+    D="$BATS_TEST_TMPDIR/push"
+    mkdir -p "$D/tap" "$D/bin"
+    printf '{"prompt":"hello"}\n' > "$D/corpus.jsonl"
+    if [ -n "$1" ]; then
+        printf 'container=%s\nimage=radiance@sha256:test\nmode=calib\npid=1\n' "$1" \
+            > "$D/tap/.rig-boot"
+    fi
+    echo "$D"
+}
+
+@test "the pusher refuses a tap that carries no rig marker: production is not a rig" {
+    # With no marker, the old driver fed four rounds of 10M tokens into restored
+    # production after the rig's boot had died, and recorded nothing.
+    D="$(pusher_fixture '')"
+    run python3 "$PUSHER" --corpus "$D/corpus.jsonl" --tap-dir "$D/tap" --port 1
+    [ "$status" -eq 2 ]
+    [[ "$output" == *"no rig marker"* ]]
+}
+
+@test "the pusher refuses a marker that names another container" {
+    D="$(pusher_fixture arbench)"
+    run python3 "$PUSHER" --corpus "$D/corpus.jsonl" --tap-dir "$D/tap" \
+        --container qwen38-bf16 --port 1
+    [ "$status" -eq 2 ]
+    [[ "$output" == *"names container 'arbench'"* ]]
+    [[ "$output" == *"targets 'qwen38-bf16'"* ]]
+}
+
+@test "the pusher refuses when the rig container is not running" {
+    D="$(pusher_fixture qwen38-bf16)"
+    stub_podman "$D" stopped
+    run env PATH="$D/bin:$PATH" python3 "$PUSHER" --corpus "$D/corpus.jsonl" \
+        --tap-dir "$D/tap" --container qwen38-bf16 --port 1
+    [ "$status" -eq 2 ]
+    [[ "$output" == *"is not running"* ]]
+}
+
+# --- the acceptance gate without an FP8 anchor -------------------------------
+
+# kld_report_slices <path> <model> <kl all> <kl code> <top1 all> <top1 code>
+kld_report_slices() {
+    python3 - "$@" <<'PY'
+import json, sys
+path, model, kl_all, kl_code, t_all, t_code = sys.argv[1:7]
+def g(kl, t, n=25000):
+    return {"kld": {"mean": float(kl), "p99_9": float(kl) * 4},
+            "top1_agreement": float(t), "positions": n}
+json.dump({"model": model, "all": g(kl_all, t_all),
+           "by_source": {"code": g(kl_code, t_code), "ops": g(kl_all, t_all)}},
+          open(path, "w"))
+PY
+}
+
+gate2() { # gate2 <kl iq4r> <kl cand> <top1 iq4r> <top1 cand> [extra]
+    D="$BATS_TEST_TMPDIR/kl2"
+    mkdir -p "$D"
+    kld_report_slices "$D/iq4r.json" iq4r-moe "$1" "$1" "$3" "$3"
+    kld_report_slices "$D/cand.json" w5nl64 "$2" "$2" "$4" "$4"
+    run bash "$GATE" "$D/iq4r.json" "$D/cand.json" "${@:5}"
+}
+
+@test "a candidate that beats the served payload on mean, tail and top-1 passes without an anchor" {
+    gate2 0.0190 0.0090 0.962 0.971
+    [ "$status" -eq 0 ]
+    [[ "$output" == *"candidate vs served (no FP8 anchor)"* ]]
+    [[ "$output" == *"KLD GATE PASS"* ]]
+}
+
+@test "a better mean with a worse tail fails: the tail is where a mean hides its damage" {
+    D="$BATS_TEST_TMPDIR/kl3"
+    mkdir -p "$D"
+    kld_report_slices "$D/iq4r.json" iq4r-moe 0.0190 0.0190 0.962 0.962
+    python3 - "$D/cand.json" <<'PY'
+import json, sys
+# mean better than the served payload, p99.9 worse: the shape a mean-only gate waves through
+g = {"kld": {"mean": 0.0090, "p99_9": 0.4000}, "top1_agreement": 0.971, "positions": 25000}
+json.dump({"model": "w5nl64", "all": g, "by_source": {"code": dict(g), "ops": dict(g)}},
+          open(sys.argv[1], "w"))
+PY
+    run bash "$GATE" "$D/iq4r.json" "$D/cand.json"
+    [ "$status" -eq 1 ]
+    [[ "$output" == *"[FAIL] 2."* ]]
+    [[ "$output" == *"KL p99.9"* ]]
+}
+
+@test "a candidate that wins overall but regresses one source slice fails" {
+    D="$BATS_TEST_TMPDIR/kl4"
+    mkdir -p "$D"
+    kld_report_slices "$D/iq4r.json" iq4r-moe 0.0190 0.0190 0.962 0.962
+    # 'all' is better, 'code' is worse: exactly the trade a mean-only gate cannot see
+    kld_report_slices "$D/cand.json" w5nl64 0.0090 0.0300 0.971 0.950
+    run bash "$GATE" "$D/iq4r.json" "$D/cand.json"
+    [ "$status" -eq 1 ]
+    [[ "$output" == *"[PASS] 1."* ]]
+    [[ "$output" == *"[FAIL] 4."* ]]
+    [[ "$output" == *"regressed: code"* ]]
+}
+
+# --- the calibration window: gated, and it says so ---------------------------
+
+window_fixture() {
+    D="$BATS_TEST_TMPDIR/win"
+    mkdir -p "$D/campaign/kld-ref" "$D/campaign/calib" "$D/rad" "$D/corpora"
+    echo '{}' > "$D/campaign/kld-ref/kld.json"
+    : > "$D/rad/qwen38-bf16.rad"
+    : > "$D/corpora/calib.jsonl"
+    echo "$D"
+}
+
+@test "the calibration window refuses before touching a card without a reference recording" {
+    D="$BATS_TEST_TMPDIR/win-less"
+    mkdir -p "$D/campaign/calib"
+    run env FRANK_QWEN38_CAMPAIGN="$D/campaign" bash "$WINDOW"
+    [ "$status" -eq 2 ]
+    [[ "$output" == *"no reference recording"* ]]
+}
+
+@test "the calibration window's dry run prints the gated plan, including what it would bank" {
+    D="$(window_fixture)"
+    run env FRANK_QWEN38_CAMPAIGN="$D/campaign" \
+        FRANK_QWEN38_RAD="$D/rad/qwen38-bf16.rad" \
+        Q38_WINDOW_CALIB_CORPUS="$D/corpora/calib.jsonl" bash "$WINDOW" --dry-run
+    [ "$status" -eq 0 ]
+    [[ "$output" == *"run-qwen38-bf16.sh calib"* ]]
+    [[ "$output" == *"else kill the rig"* ]]
+    [[ "$output" == *"calib-push.py"* ]]
+    [[ "$output" == *"AGENTTRASH"* ]]
+}
+
+@test "a flat calibration round is not a failure: the tap's floor is pooled" {
+    # round 1 of a 10.03M-token corpus against a 12M floor seals nothing, and the
+    # first version of this gate aborted the run there — one round too early.
+    run python3 - "$PUSHER" <<'PY'
+import importlib.util, sys
+spec = importlib.util.spec_from_file_location("cp", sys.argv[1])
+m = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(m)
+print(m.round_verdict(8192, 8192, False),   # flat: continue
+      m.round_verdict(8192, 4096, False),   # shrink: fail
+      m.round_verdict(8192, 20000, False),  # grew: continue
+      m.round_verdict(0, 0, True))          # done: stop
+PY
+    [ "$status" -eq 0 ]
+    [ "$output" = "flat shrink grew done" ]
+}
+
+@test "the calibration boot's step is its own knob; the KL modes keep production's" {
+    # Calib is the only phase whose cost is read volume per document: four 2,048-token steps
+    # re-stream each layer's routed experts four times. Measured 2026-10-29: 71 tok/s, 47 h tap.
+    run env Q38_CALIB_MNBT=8192 bash "$RUNNER" calib --dry-run
+    [ "$status" -eq 0 ]
+    grep -A1 -x -- '--max-num-batched-tokens' <<<"$output" | grep -qx '8192'
+    # and the same knob must NOT reach the measuring shape those numbers are compared in
+    run env Q38_CALIB_MNBT=8192 bash "$RUNNER" kldref --dry-run
+    [ "$status" -eq 0 ]
+    ! grep -qx '8192' <<<"$output"
+    run env Q38_CALIB_MNBT=wide bash "$RUNNER" calib --dry-run
+    [ "$status" -eq 2 ]
+    [[ "$output" == *"Q38_CALIB_MNBT must be a number"* ]]
+}

tests/qwen38_reference.bats

@@ -29,7 +29,7 @@ RUNNER="$REPO_ROOT/scripts/run-qwen38-bf16.sh"
 # The reference shape, spelled exactly as run-qwen38-bf16.sh must pass it: this
 # is the contract the recorded reference is measured at, so a silent edit here
 # is a silent edit to every KLD number in the campaign.
-SHAPE="--tp 2 --placement expert_tiered --weights-disk-tier --host-pool-mib 66560 --gpu-headroom-mib 96 --deterministic --num-speculative-tokens 0 --kv-cache-dtype fp8 --max-model-len 32768 --ngram-placement ram --override-chat-template /models/tpl.jinja"
+SHAPE="--tp 2 --placement expert_tiered --weights-disk-tier --host-pool-mib 49152 --gpu-headroom-mib 96 --deterministic --num-speculative-tokens 0 --kv-cache-dtype fp8 --max-model-len 32768 --ngram-placement ram --override-chat-template /models/tpl.jinja"
 
 # oneline — both dry runs print one shell-quoted argument a line; joining them
 # lets an option and its value be asserted as the pair they are.