Files
BFM-decomp/tools/campaign_status.py
T
Drew T 094dfebef4 fix(campaign): measure what the campaign can see — banked-today from the INCLUDE_ASM invariant, gate_min beside wall_min, a lock-aware fleet R22 window (P31 S61 T10+/S61-8)
campaign_status: 'today: N banked' summed '— N banked' commit subjects and missed every bank that
rode in a chore/maint commit (S60: 2,185 reported vs 2,644 net stubs removed); it now derives the
number from INCLUDE_ASM stub counts at last-commit-before-midnight / HEAD / working tree (R33), and
alive() is anchored so pgrep no longer matches its own wrapper (every lane read ok with 0 processes).
ox_campaign gater: the ledger's wall_min counts drafting + queue wait since the ready marker's t0;
gate_min is the gate alone (the '30-67 min gates' picture was this conflation).
maintenance lane: the fleet R22 sweep skipped whenever any gate was in flight, i.e. always (last
real sweep 12:54 08-25); it now takes .run/auto/draw.lock and waits its turn, skipping only for gate_main.
2026-08-25 23:49:43 -06:00

173 lines
7.6 KiB
Python

#!/usr/bin/env python3
"""campaign_status.py — ONE status view covering EVERY lane, not just the loud one.
WHY (Drew, P31 S59): status checks kept reporting the overlay drafter and the gater — the lanes
whose logs scroll — while the main lane, the maintenance lane and the distill lane went unmentioned
for hours. A lane you do not report is a lane you do not notice failing: the main lane spent an
afternoon on an old config and bisected a whole batch to zero banks without that ever reaching a
status line. This prints all of them, each with ITS OWN metrics, from the artefacts rather than
from memory.
tools/campaign_status.py # everything
tools/campaign_status.py --brief # one block per lane, no history
"""
import argparse
import glob
import json
import os
import re
import subprocess
import sys
import time
REPO = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
sys.path.insert(0, os.path.join(REPO, "tools"))
LANES = ("drafter", "gater", "maintenance", "stallguard", "distill", "main")
def sh(cmd):
return subprocess.run(["bash", "-c", cmd], capture_output=True, text=True, cwd=REPO).stdout.strip()
def alive(lane):
"""Anchored: `pgrep -f 'bash .run/x.sh'` matched THIS tool's own `bash -c "pgrep ..."` wrapper, so
every lane read `ok` with zero processes alive (S60 shutdown; verified S61 with a clean ps)."""
return bool(sh(f"pgrep -f '^bash \\.run/{lane}\\.sh' | head -1"))
def agents_by_wave():
out = {}
for p in sh("pgrep -f '^\\.venv/bin/python -u tools/api_agent' || true").split():
try:
cl = open(f"/proc/{p}/cmdline", "rb").read().decode(errors="replace").replace("\0", " ")
except OSError:
continue
m = re.search(r"wave_([a-z0-9]+)", cl)
w = m.group(1) if m else "?"
env = ""
try:
env = open(f"/proc/{p}/environ", "rb").read().decode(errors="replace")
except OSError:
pass
mt = re.search(r"MAXTOK=(\d+)", env)
out.setdefault((w, mt.group(1) if mt else "?"), 0)
out[(w, mt.group(1) if mt else "?")] += 1
return out
def rate(window=300):
now, req, f429 = time.time(), 0, 0
try:
for line in open(os.path.join(REPO, ".run/api_rate.jsonl"), errors="replace"):
line = line.strip()
if not line:
continue
try:
r = json.loads(line)
except Exception:
continue
if now - r.get("t", 0) > window:
continue
if r.get("ev") == "POST":
req += 1
elif "429" in str(r.get("ev")):
f429 += 1
except OSError:
return None
return req, f429, req / (window / 60.0)
def wave_rows(pattern, n):
"""(tag, cards, drafts, truncated) for the most recent waves matching a card-file glob."""
rows = []
for cp in sorted(glob.glob(os.path.join(REPO, pattern)), key=os.path.getmtime)[-n:]:
tag = os.path.basename(cp).split("_")[1]
try:
c = json.load(open(cp))
c = c if isinstance(c, list) else c.get("cards", [])
except Exception:
c = []
d = len(glob.glob(os.path.join(REPO, f".run/wave_{tag}/shard*/*.c")))
tr = 0
for lp in glob.glob(os.path.join(REPO, f".run/wave_{tag}/shard*.log")):
try:
tr += sum(1 for l in open(lp, errors="replace") if "finish=length" in l)
except OSError:
pass
rows.append((tag, len(c), d, tr))
return rows
def main():
ap = argparse.ArgumentParser()
ap.add_argument("--brief", action="store_true")
a = ap.parse_args()
os.chdir(REPO)
print(f"=== CAMPAIGN STATUS {time.strftime('%H:%M:%S')} ===")
# ---- lanes ---------------------------------------------------------------------------------
print("lanes : " + " ".join(f"{l}={'ok' if alive(l) else 'DEAD'}" for l in LANES))
ab = agents_by_wave()
tot = sum(ab.values())
detail = " · ".join(f"{w}:{n}@{mt}tok" for (w, mt), n in sorted(ab.items(), key=lambda x: -x[1]))
print(f"agents: {tot} {detail}")
r = rate()
if r:
print(f"rate : {r[2]:.0f} req/min · {r[1]} 429 in 5 min")
# ---- overlay drafting lane -----------------------------------------------------------------
print("\n-- OVERLAY LANE (drafter -> gater) --")
for tag, cards, drafts, tr in wave_rows(".run/wave_??_cards.json", 4 if not a.brief else 2):
print(f" wave {tag}: {cards:4} cards · {drafts:4} drafts ({100*drafts/max(cards,1):3.0f}%) · {tr:4} truncated turns")
for line in sh("grep -E 'GATE [a-z]+: banked' .run/gater.log 2>/dev/null | tail -3").splitlines():
print(" " + line.strip()[:110])
print(f" ready queue: [{sh('ls .run/ready/ 2>/dev/null | tr \"\\n\" \" \"')}]")
# ---- main lane -------------------------------------------------------------------------------
print("\n-- MAIN LANE (its own gate cadence) --")
for tag, cards, drafts, tr in wave_rows(".run/wave_m??_cards.json", 3 if not a.brief else 1):
print(f" wave {tag}: {cards:4} cards · {drafts:4} drafts ({100*drafts/max(cards,1):3.0f}%) · {tr:4} truncated turns")
for line in sh("grep -E 'banked|nothing banked|BISECT' .run/main_lane.log 2>/dev/null | tail -4").splitlines():
print(" " + line.strip()[:110])
try:
import corpus
print(f" open main stubs: {len(corpus.stubs('main'))}")
except Exception as e:
print(f" (main stub count unavailable: {type(e).__name__})")
# ---- the free lanes ---------------------------------------------------------------------------
print("\n-- MAINTENANCE / DISTILL (zero model tokens) --")
for line in sh("tail -3 .run/maintenance.log 2>/dev/null").splitlines():
print(" maint: " + line.strip()[:104])
pend = sh("ls .run/distill_ready/ 2>/dev/null | tr '\\n' ' '")
try:
mined = len(json.load(open(".run/distill_state.json")).get("mined", {}))
except Exception:
mined = "?"
print(f" distill: {mined} wave(s) mined · ready batches: [{pend}]")
# ---- totals ----------------------------------------------------------------------------------
# BANKED TODAY IS DERIVED FROM THE INCLUDE_ASM INVARIANT, NOT FROM COMMIT SUBJECTS (S61, R33).
# The subject regex summed "— N banked" and missed every bank that rode in a chore/maint commit:
# the S60 A-prop pass put 357 banks in commit:2904 ("chore: ...") and the regex reported 2,185
# for a day whose stub count fell by 2,644 net (frontier-analysis-s60 §1). Stubs are counted at
# the last commit before midnight, at HEAD, and in the working tree (uncommitted banks — R42
# says commit them, the number says whether anyone did). Carves re-add stubs, so this is NET.
def _stubs(rev):
if rev == "tree":
return int(sh("grep -rc 'INCLUDE_ASM(' src/ --include='*.c' | awk -F: '{s+=$NF} END {print s+0}'") or 0)
return int(sh(f"git grep -c 'INCLUDE_ASM(' {rev} -- src/ | awk -F: '{{s+=$NF}} END {{print s+0}}'") or 0)
base = sh("git rev-list -1 --before='00:00' HEAD")
s_base, s_head, s_tree = _stubs(base), _stubs("HEAD"), _stubs("tree")
subj = sh("git log --since='00:00' --format='%s' | grep -oE '[—-] [0-9]{1,4} banked' "
"| awk '{s+=$2} END {print s+0}'")
print(f"\ntoday : {s_base - s_head} net stubs removed since {base[:9]} ({s_base} -> {s_head} committed"
f"{'' if s_tree == s_head else f', {s_head - s_tree:+d} more uncommitted in the tree'}) · "
f"{sh('git log --since=00:00 --oneline | wc -l')} commits · subject-regex sum {subj} (undercounts)")
return 0
if __name__ == "__main__":
sys.exit(main())