namekit

Deterministic pop-culture name generator. Map any string key to a stable, memorable name. Filter by franchise, entity type, or affiliation; choose how the name is formatted.

pip install namekit

import namekit

namekit.name("user-42")                              # 'fred_weasley'
namekit.name("user-42", franchise="lotr")            # 'gollum'
namekit.name("user-42", case="title")                # 'Fred Weasley'
namekit.name("user-42", name_part="last")            # 'weasley'
namekit.name("user-42", suffix=True)                 # 'fred_weasley_6d894'

The same key always produces the same name. The mapping is stable across processes, machines, and Python versions because it uses SHA-256 rather than Python's salted built-in hash().

Install

pip install namekit

Or, from source:

pip install -e .

Why namekit?

Common use cases:

• Naming training runs, experiments, or sweeps from a config hash, so the same config always lands on the same name.
• Generating friendly identifiers for users, sessions, or test fixtures.
• Replacing opaque hashes with human-readable labels in logs and dashboards.

Filtering

Three independent filters compose freely.

Each filter accepts a single value, a list/tuple of values, or None (no filter).

from namekit import name, list_names

# A villain from Game of Thrones (rendered Title Case)
name("user-1", franchise="got", affiliation="bad", case="title")
# 'Petyr Baelish'

# LOTR strongholds (the bad guys' places)
list_names(franchise="lotr", entity_type="place", affiliation="bad", case="title")
# ['Mordor', 'Isengard', 'Mount Doom']

# Disney villains
list_names(franchise="disney", affiliation="bad", case="title")
# ['Scar', 'Ursula', 'Gaston', 'Jafar']

If a filter combination yields no entries (e.g., bad scientists), name() raises ValueError.

Format options

Two more knobs control the shape of the returned string.

name_part — which slice of the name

list_names(franchise="scientists", name_part="last", case="title")[:8]
# ['Newton', 'Einstein', 'Curie', 'Darwin', 'Galilei', 'Tesla', 'Feynman', 'Hawking']

case — how the parts are joined

Apostrophes and hyphens are preserved in title only; other cases strip them (e.g., King's Landing → kings_landing / kingslanding / kings-landing).

Hash suffix for uniqueness

A short hex suffix makes collisions vanishingly rare while keeping the prefix readable.

name("config-v3", suffix=True)
# 'aragorn_8c1d2'

name("config-v3", suffix=True, suffix_length=8, separator="-")
# 'aragorn-8c1d24a9'

Reusable namer

When you call the namer many times with the same configuration, build a NameKit once.

from namekit import NameKit

kit = NameKit(
    franchise=["lotr", "ghibli"],
    affiliation="good",
    case="title",
    suffix=True,
)
for user_id in user_ids:
    print(kit(user_id))

Inspecting the corpus

from namekit import (
    FRANCHISES, ENTITY_TYPES, AFFILIATIONS, NAME_PARTS, CASES,
    ENTITIES, list_names, list_entities, format_name,
)

FRANCHISES                              # ('avatar', 'starwars', 'lotr', ...)
ENTITY_TYPES                            # ('character', 'place')
AFFILIATIONS                            # ('good', 'bad', 'neutral')
NAME_PARTS                              # ('first', 'last', 'full')
CASES                                   # ('snake', 'title', 'compact', 'kebab')

list_names(franchise="harrypotter", affiliation="good", case="title")
list_entities(entity_type="place")      # full Entity dataclasses

# Format a single Entity yourself
e = ENTITIES[0]
format_name(e, name_part="last", case="title")

Each Entity is a frozen dataclass:

from namekit import Entity
Entity(parts=("Frodo", "Baggins"), franchise="lotr",
       entity_type="character", affiliation="good")

The parts tuple holds Title Case word pieces. entity.first and entity.last are convenience properties (last is None when there is only one part).

Bundled franchises

avatar, starwars, lotr, pokemon, harrypotter, marvel, got, ghibli, mario, zelda, disney, greek, norse, scientists, philosophers.

The corpus has 434 entries (~430 unique formatted names) across 361 characters and 73 places. Affiliation tags are best-effort labels for casual filtering; places default to neutral and only explicit villain strongholds (Mordor, Bowser's Castle, etc.) are tagged bad.

Notes on collisions

Without a suffix, the chance of two different keys producing the same name is roughly 1 / len(corpus). For the default corpus that's about 0.23% per pair. If uniqueness matters, pass suffix=True; with the default 5-char hex suffix, collisions become astronomically unlikely.

Development

pip install -e '.[dev]'
pytest

License

MIT