File size: 2,406 Bytes
014fc43 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 | """Count 'sleep' in assistant-written text across every transcript on this machine.
Separates code/command uses (Start-Sleep, time.sleep, sleep 5) from prose
addressed to the operator. Every prose sentence is printed verbatim.
"""
import glob, json, os, re
from collections import Counter
BASE = r"C:\Users\acer\.claude\projects"
CODE = re.compile(r"(start-sleep|time\.sleep|sleep\s*\(|\bsleep\s+\d|asyncio\.sleep|"
r"sleep_?ms|delayseconds|thread\.sleep|--sleep|sleepms)", re.I)
WORD = re.compile(r"\bsleep\w*\b", re.I)
files = glob.glob(os.path.join(BASE, "**", "*.jsonl"), recursive=True)
total = code_hits = prose_hits = 0
prose, forms = [], Counter()
blocks = 0
for f in files:
try:
fh = open(f, encoding="utf-8", errors="replace")
except Exception:
continue
with fh:
for line in fh:
try:
r = json.loads(line)
except Exception:
continue
if r.get("type") != "assistant":
continue
c = (r.get("message") or {}).get("content")
if not isinstance(c, list):
continue
for b in c:
if not (isinstance(b, dict) and b.get("type") == "text"):
continue
txt = b.get("text", "")
if "sleep" not in txt.lower():
continue
blocks += 1
for m in WORD.finditer(txt):
total += 1
forms[m.group(0).lower()] += 1
for s in re.split(r"(?<=[.!?\n])\s+", txt):
s = " ".join(s.split())
if not WORD.search(s):
continue
n = len(WORD.findall(s))
if CODE.search(s):
code_hits += n
else:
prose_hits += n
prose.append((os.path.basename(f)[:18], s[:240]))
print("CORPUS|jsonl_files=%d|assistant_blocks_containing_sleep=%d|json=0"
% (len(files), blocks))
print("SLEEP|total_word_occurrences=%d|code_or_command=%d|prose=%d|json=0"
% (total, code_hits, prose_hits))
print("FORMS|%s|json=0" % ", ".join("%s=%d" % (k, v) for k, v in forms.most_common()))
print()
print("PROSE_SENTENCES|n=%d|json=0" % len(prose))
for fn, s in prose:
print(" P|%s| %s" % (fn, s))
|