idio

Classical-language, human-readable identifiers with >32-bit entropy using Latin and Ancient Greek lexemes. Designed for enterprise-grade uniqueness while keeping outputs memorable and ASCII-only.

Why idio?

• Uses cryptographically secure randomness (secrets.SystemRandom).
• Exceeds 32 bits of entropy by composing two grammar-conformant segments.
• Deterministic grammar drawn from AGENT.md: three patterns mixing Latin adjectives/nouns and Greek nouns/modifiers.
• Pure-Python, zero dependencies.

Installation

pip install idio
# or locally
pip install .

Usage

from idio import generate_identifier, generate_segment, entropy_bits

# Default: two segments to exceed 32 bits, domain-safe with hyphens and dots
identifier = generate_identifier()
print(identifier)  # e.g., "unicus-nomen-onoma.idion-sema-signum"

# Single segment (useful for display, but ~18.9 bits with bundled lists)
segment = generate_segment()

# Entropy budget for your configuration
bits = entropy_bits(segments=2)
print(f"{bits:.2f} bits")

Subdomain-safe format

By default, words are joined with - and segments with . to produce valid DNS labels (e.g., normalis-gradus-kosmos.perfectus-signum-onoma). If you prefer the original underscore format, pass word_separator="_" (and optionally segment_separator="--").

Patterns (all supported)

• M1: <latin_adjective>-<latin_noun>-<greek_noun>
• M2: <greek_modifier>-<greek_noun>-<latin_noun>
• M3: <latin_noun>-<latin_adjective>-<greek_noun>

generate_identifier joins segments with . by default to keep each label comfortably under DNS's 63-character limit and avoid underscores. If you want the original spec's format, pass word_separator="_" and segment_separator="--".

Entropy math

Current bundled lists are roughly: ~75 Latin adjectives, ~3.5k Latin nouns, ~900 Greek nouns, and 11 Greek modifiers.

• Combinations per pattern set per segment: on the order of 5e8
• Bits per segment: approximately 29 bits
• Default identifier (2 segments): approximately 58 bits

Add more words (≤100 per list) or more segments if you need an even larger space.

API surface

• generate_identifier(segments=2, patterns=None, bank=DEFAULT_BANK, word_separator="-", segment_separator=".")
• generate_segment(pattern=None, bank=DEFAULT_BANK, word_separator="-")
• entropy_bits(segments=2, patterns=Pattern, bank=DEFAULT_BANK)
• IdioGenerator class if you want to own the RNG or word bank.

Wordlists

All lexemes are ASCII transliterations with no diacritics, chosen to stay abstract, technical, and non-mythological per the spec in AGENT.md.

Development

python -m venv .venv && source .venv/bin/activate
pip install -e .[dev]
pytest

License

MIT License © Causum