#!/usr/bin/env python3 """Measure your own voice, then gate outbound text against it. Agents write in a recognisable register: openers like "I hope this finds you well", em dashes, three hedged bullet points where one sentence would do. A style guide in prose does not stop it; a lint that refuses does. This is the generic half. The other half is yours and stays private: your own writing to measure, and your own slop list. voice-lint --measure FILE... profile your writing: sentence length, contractions, lowercase starts, hedges, exclamations, em dashes voice-lint --slop LIST [--allow-em-dash] [--max-exclamations N] < DRAFT FAIL: lines and exit 1 on a slop phrase, an em dash, or too many exclamation marks; exit 0 and PASS otherwise LIST is a text file, one phrase per line (case-insensitive), or `re: `; `#` starts a comment. slop.example.txt has generic entries to start from. In check mode an unknown flag such as `--reply` is ignored, so it works as the outreach voice command (RH_VOICE_COMMAND). """ import json import re import statistics import sys HEDGES = ('i think', 'probably', 'maybe', 'i guess', "i don't know", "i'm not sure", 'perhaps') CONTRACTION = re.compile(r"\b\w+'(?:s|t|re|ve|ll|d|m)\b", re.I) SENTENCE = re.compile(r'[^.!?\n]+[.!?]?') def measure(texts): """A profile of one person's writing: the numbers a voice guide is written from.""" messages = [t.strip() for t in texts if t.strip()] sentences = [s.strip() for t in messages for s in SENTENCE.findall(t) if s.strip()] words = [len(s.split()) for s in sentences] low = [t.lower() for t in messages] rate = lambda n: round(n / len(messages), 3) if messages else 0.0 return dict( messages=len(messages), sentences=len(sentences), words_per_sentence=dict(median=statistics.median(words) if words else 0, p90=sorted(words)[int(0.9 * (len(words) - 1))] if words else 0), contractions=rate(sum(bool(CONTRACTION.search(t)) for t in messages)), lowercase_start=rate(sum(t[:1].islower() for t in messages)), exclamation=rate(sum('!' in t for t in messages)), em_dash=rate(sum('—' in t for t in messages)), hedges={h: sum(t.count(h) for t in low) for h in HEDGES if any(h in t for t in low)}) def load_slop(path): rules = [] for line in open(path, encoding='utf-8'): line = line.strip() if not line or line.startswith('#'): continue if line.startswith('re:'): rules.append((line, re.compile(line[3:].strip(), re.I))) else: rules.append((line, re.compile(r'(? max_exclamations: failures.append('{} exclamation marks (at most {})'.format(text.count('!'), max_exclamations)) return failures def main(argv): if '--measure' in argv: files = [a for a in argv if a != '--measure'] texts = [] for f in files: texts += re.split(r'\n\s*\n', open(f, encoding='utf-8').read()) print(json.dumps(measure(texts), indent=2)) return 0 slop, allow, maxx = [], False, 1 i = 0 while i < len(argv): if argv[i] == '--slop': slop = load_slop(argv[i + 1]); i += 2; continue if argv[i] == '--allow-em-dash': allow = True elif argv[i] == '--max-exclamations': maxx = int(argv[i + 1]); i += 1 i += 1 failures = check(sys.stdin.read(), slop, allow, maxx) for f in failures: print('FAIL: ' + f) if not failures: print('PASS') return 1 if failures else 0 if __name__ == '__main__': sys.exit(main(sys.argv[1:]))