Skip to main content

onlyworlds (Python)

OnlyWorlds is an open standard for worldbuilding data: 22 element types (characters, locations, events, laws and the rest) that any tool can read and write. This is a Python client for its API, and a reader and writer for OnlyWorlds world folders (a world as a folder of JSON files). Python 3.12+, no runtime dependencies (the optional OKF export needs PyYAML).

Install

pip install onlyworlds

The newest unreleased code: pip install git+https://github.com/OnlyWorlds/python-sdk.

First run

Moppetopia is a public demo world. Its read-only key, 0000000001, needs no account:

from onlyworlds import Client

client = Client("0000000001")
print(client.get_world()["name"])
for c in client.list_page("character", filter={"name__icontains": "fluffington"})["data"]:
    print(" ", c["name"])
Moppetopia
  Admiral Fluffington

Your own world

Create a world at onlyworlds.com; its keys are on the world's page. A new key is shown once, so copy it then.

  • An ow_r_ key reads, with no PIN: Client("ow_r_…").
  • An ow_w_ key reads and writes. Writes also send a PIN: Client("ow_w_…", "…").
  • For code that writes, give it its own agent seat, a key for one tool or script. The seat's key and its ow_s_ secret (passed as the PIN) work in one world, and you can remove them without touching your account PIN.

Elements are plain dicts:

writer = Client("ow_w_…", "ow_s_…")
peak = writer.create("location", {"name": "Dragon Peak"})
writer.patch("location", peak["id"], {"supertype": "Mountain"})

for character in client.iter_elements("character", expand=["species"]):
    print(character["name"])

World folders

A world folder is one JSON file per element, in the world folder format. Atlas, the OnlyWorlds editor, reads and writes the same format.

from onlyworlds import export_world, read_folder

export_world("moppetopia", api_key="0000000001")
folder = read_folder("moppetopia")
print(folder.world["name"], len(folder.elements), "elements")

An exported folder's world.json says where it came from and which part of the world it holds: source names the server, the world's id and when it was exported, and exported_as says whether the key read the whole world as its owner, as a member (with its role), or as one Character (a guest's slice). When the server can't say which (GET /me refused or missing), exported_as is left out rather than guessed. producer says which tool made the folder and from what: {"tool": "onlyworlds", "version": "<this package's version>", "input": "server"}. A folder written from another folder (a branch, a copy) keeps its maker's producer as it found it, and one with none gets none.

write_folder(root, world, elements) writes a folder from a world dict and a dict of element lists by type, keeping every key the world dict has. Reading never changes a file, and a file without an id is skipped and left alone.

A whole world also travels as one file, an .owworld (a zip by another name, the folder format's §4.7):

from onlyworlds import read_world_file, write_world_file

write_world_file("moppetopia", "moppetopia.owworld")
world = read_world_file("moppetopia.owworld")
world.to_folder("moppetopia-copy")

write_world_file carries world.json byte for byte (its producer, source and format_version included) and the folder's parts, elements/, spatial/, media/ and records/, and nothing else: never an api block, never a tool's own folder unless you hand it in (private_dir=".atlas", which makes the file a private copy rather than one to send around), and never a producer's records its records/RECORDS.json entry marks "private": true unless you ask (include_private_records=True); the manifest itself always travels as it is. The same folder gives the same bytes on one Python. read_world_file checks the whole archive before it unpacks anything and refuses a file it cannot open safely with a WorldFileError, whose reason names one of the spec's twelve causes (unsafe_path, collision, too_big, ...). It gives the same verdict as the TypeScript and Unity SDKs on every case of the format's conformance set.

To send a folder's edits back, keep a copy exported before you edited, then push the difference:

from onlyworlds import push

result = push("moppetopia-before", "moppetopia", api_key="ow_w_…", pin="ow_s_…", log_path="push.jsonl")

push sends only the fields that changed, retries what the server asks it to retry, checks that what came back is what it sent, and logs every write, so running it again skips what already landed. dry_run=True shows the plan without sending it.

The world's origin is a world's birth line, written once: a push sets it only on a first fill from a folder made elsewhere, never from a folder exported from the same world, and never over a first import. When get_world() serves origin null in both keys and the folder's source.world is not that world (or the folder has no source line, as a converter's first output), push sends the folder's source and producer from world.json in one PATCH /world and reports it in result.origin. patch_world sends the two only when you name them (source=, producer=; None clears one); a folder world.json handed to it keeps its own pair home, since that pair names the folder's maker.

A world kept in Notion becomes a folder without retyping it. Export the workspace or page from Notion as "Markdown & CSV", then:

python -m onlyworlds.notion Export.zip my-world --plan   # what would become what; writes nothing
python -m onlyworlds.notion Export.zip my-world
from onlyworlds.notion import import_notion

report = import_notion("Export.zip", "my-world")
print(report.render())

A database whose title is one of the 22 types (Characters, Locations, any case, singular or plural), or that has a Type column naming one per row, becomes elements of that type. A column becomes a field when its name matches the field's (Birth Date is birth_date), and a relation column becomes a link when it matches a link field and its pages are in the export. Every other page becomes a Narrative, and so does each row of a database whose type the door cannot tell (supertype notion, the database's name as subtype). Nothing is dropped: a column that doesn't fit is kept as x_notion_<column>, and attachments a page links to are copied into media/notion/. Ids come from Notion's page ids, so importing the same export twice gives the same folder. The run ends with a report naming every column it could not place and every link it could not resolve, so you can see what to rename in Notion before the next export. No AI, no account. It is built on Notion's documented export shape; no real export has been through it yet.

A world kept in LegendKeeper becomes a folder from LegendKeeper's own export file (Project Settings, Export; the .lk file or the JSON one), with no key and no account:

python -m onlyworlds.legendkeeper --file my-world.lk my-world --plan   # what would become what; writes nothing
python -m onlyworlds.legendkeeper --file my-world.lk --map faction=Institution my-world

or through LegendKeeper's own API, read-only, with your project's API key (Project Settings, Developers, API Keys; LegendKeeper gives the API to projects whose owner is on a paid plan) and the project's id:

export LEGENDKEEPER_API_KEY=lk_...                                          # or --key-file PATH
python -m onlyworlds.legendkeeper --project ID my-world --plan
python -m onlyworlds.legendkeeper --project ID --map faction=Institution my-world
from onlyworlds.legendkeeper import ApiSource, import_legendkeeper

report = import_legendkeeper("my-world.lk", out_dir="my-world")  # the export file
report = import_legendkeeper(ApiSource(key), "ID", "my-world", type_map={"faction": "institution"})
print(report.render())

Both roads give the same elements. The export file names no project, so its ids are keyed on the project's first top-level page; pass --project ID with --file to key them on the project's own id, the same ids the API road gives. LegendKeeper's Markdown export is not read: it loses the tags, the properties and the hidden marks.

LegendKeeper has no element types, so a page's type comes from its tags: a tag that is one of the 22 type names (character, Location) types it, --map TAG=TYPE adds your own tags, and every other page becomes a Narrative (supertype legendkeeper). The page text becomes the description as Markdown. A sidebar or infobox row becomes a field when its key matches one (Height is height), and a titled link list (Allies) becomes a link field when it matches one and its targets fit; otherwise each link becomes a Relation element named by the list's title. A timeline's events become Events with LegendKeeper's raw minutes and formatted dates kept beside them; their start_date is left for you to set against the world's own calendar. What LegendKeeper hides stays hidden: a hidden page is imported with hidden: true (and every page under it), a hidden tab on a visible page becomes its own hidden Narrative, and so does each Secret block, the page keeping only its visible text, so nothing is dropped and nothing secret turns public. Templates are imported marked x_legendkeeper_template, and the tutorial pages a new project starts with are imported like the rest, since nothing marks them. Whatever has no field is kept as x_legendkeeper_<key>, and the report names every key, link, document and hidden thing. Ids come from LegendKeeper's ids, so a second import of the same project gives the same elements. The key is sent to LegendKeeper only, never written or printed. No AI. Map and board documents are not imported yet. Read on one real export file (a tutorial project with six pages typed by hand); the API road has been walked on a test project.

A campaign kept in Kanka becomes a folder from Kanka's own JSON export (Campaign, Settings, Data, Export, JSON; free campaigns get it):

python -m onlyworlds.kanka export.zip my-world --plan   # what would become what; writes nothing
python -m onlyworlds.kanka export.zip my-world
from onlyworlds.kanka import import_kanka

report = import_kanka("export.zip", "my-world")
print(report.render())

Each Kanka module becomes its type: characters Characters, locations Locations, organisations Institutions, notes, quests and journals Narratives (the text in story), families, races, items, creatures, events, abilities and maps their types, and any other module a Narrative with supertype kanka. The entry becomes the description as Markdown, with the posts after it under their own headings; a mention becomes the entity's name. A location's parent becomes parent_location, a character's places location, its appearance and personality traits physicality and mentality, and an attribute named like a field (Architecture) fills it. A relationship becomes a Relation, and a two-way one is one Relation, not two. Exact duplicates (the same relationship or post made twice) fold into one, and the report names each fold. What Kanka hides stays hidden: a private entity is imported with hidden: true, a private post becomes its own hidden Narrative, and a visible entity's private details go into one hidden Narrative beside it, so nothing is dropped and nothing private turns public. Kanka's starter entries ([Your Character Name], ...) are imported like the rest, since nothing in the data marks them, and the report names them. Everything without a field is kept as x_kanka_<key>. Ids come from Kanka's ids, so a second import of the same campaign gives the same elements. No AI. Read on a real Kanka 3.15 export of characters, locations, organisations and notes; the other modules are built on the same shape and have not been seen in a file yet.

A world kept in World Anvil becomes a folder from World Anvil's own export (world settings, Export World, Export this World; free accounts get it). It opens a page, not a download: save that page from the browser as HTML, then:

python -m onlyworlds.worldanvil --file the-quay-world-export.html my-world --plan   # what would become what; writes nothing
python -m onlyworlds.worldanvil --file the-quay-world-export.html my-world
from onlyworlds.worldanvil import import_worldanvil

report = import_worldanvil("the-quay-world-export.html", "my-world")
print(report.render())

Each article becomes its template's type: Character a Character, Species a Species, Geographic Location a Location (and Settlement, Organization, Item, Ethnicity, Myth / Legend, Vehicle, Profession and a few more, by one table, TEMPLATE_TYPES); an article of any other template becomes a Narrative (supertype Article, the template's name as subtype). The body becomes the description as Markdown. A link to another article becomes its name in the text, with the linked element kept by id in x_worldanvil_mentions; a mention World Anvil left unresolved (@Ada) stays as you typed it, and the report names it. A template field named like a field (Height) fills it; every other (Gender, Eyes) is kept as x_worldanvil_<field>. The article's category is kept as x_worldanvil_category, and its creation, last update and publication dates as ISO dates (x_worldanvil_created, _updated, _published; a date the door cannot read exactly is kept as printed and named). The author is named in the report, never written into the world. What World Anvil hides stays hidden: an article that is not public is imported with hidden: true, and a Secret block becomes its own hidden Narrative, so nothing is dropped and nothing private turns public. The report names every field, link, mention and hidden article, and anything in the page the door did not read. Ids come from World Anvil's article ids, so a later export of the same world gives the same elements. No AI. Read on two real exports of a small world on a free account (three articles, links, a category); private articles and Secrets (a paid feature) have not been seen in a file yet.

A project kept in Campfire becomes a folder from Campfire's own export (File, Export Elements, the html format, Select all Elements; free accounts get it). It is a document, not data, and the door reads it as one:

python -m onlyworlds.campfire --file "Campfire Export.html" my-world --plan   # what would become what; writes nothing
python -m onlyworlds.campfire --file "Campfire Export.html" my-world --project "Harbour Town"
from onlyworlds.campfire import import_campfire

report = import_campfire("Campfire Export.html", "my-world", project="Harbour Town")
print(report.render())

Each element heading becomes an element, its description the description. The file names no module, so the type is read from the template's own panel words (a Basic Information panel with Place of Residence is a Character, Location Details a Location; Species, Cultures, Items, Languages, Religions, Philosophies, Systems, Magic and Maps are mapped too, by one table, MODULE_TYPES); anything else becomes a Narrative (supertype Article). The report names, for each element, the panel words its type was read from. Every filled panel value is kept as x_campfire_<panel>_<label> (x_campfire_basic_information_full_name), a text panel as x_campfire_<panel>; an em dash is empty and not written. A document carries no ids, no links and no hidden marks, and the report says so each time: ids come from each element's type and name, so a later export gives the same elements only while the names hold. The file names no project either: --project NAME names the world and keys every id on it, so two projects exported under Campfire's default title, "Campfire Export", stay apart; without it the world takes the document's title. The one link made is a Character's Place of Residence that names a Location in the same export. No AI. Read on one real export (a Character and a Location); the other modules have not been seen in a file yet.

Export to OKF

A world folder can be exported as a valid Open Knowledge Format (OKF) 0.2 bundle, as the spec stands at one pinned commit (onlyworlds.okf.OKF_COMMIT): one markdown file per element, with its fields in the frontmatter, its text under headings and its links (in-prose ow:// references included) as markdown links, plus an index per type. This needs PyYAML, which the rest of the package does not:

pip install "onlyworlds[okf]"
python -m onlyworlds.okf moppetopia moppetopia-okf
from onlyworlds.okf import export_okf

report = export_okf("moppetopia", "moppetopia-okf")
print(report.elements, "elements,", len(report.notes), "notes")

The output directory must not exist yet. Anything the export leaves out (a file that isn't an element, a key that isn't a field, a link to the wrong type) is listed in report.notes; the command prints these on stderr. The same folder always gives the same bytes. What a bundle never carries: the server's bookkeeping (change_seq, created_at, who wrote what) and the world's key, so a bundle is a copy to read elsewhere, not a folder to sync.

The Markdown pack

A world as a few merged Markdown files, sized for an AI notebook with source caps (NotebookLM: 50 sources a notebook, 500,000 words a source; Claude Projects: 30 MB a file). No extra needed:

python -m onlyworlds.pack moppetopia moppetopia-pack          # a world folder
OW_API_KEY=ow_r_… python -m onlyworlds.pack --api moppetopia-pack   # the key's world

00-world.md says which kind of key made the pack (the owner's, a member's with its role, a Character's view; never the key), then holds the world's own fields and an index of every element; then one file per element type, each element a ## Name section with its fields in schema order and every link written as the name and type of what it points to. Room elements, proposed elements and extension fields are left out unless asked (--rooms, --proposed, --extensions). A proposed element is an AI seat's new element not yet reviewed (it carries x_review_proposed, keel #565) and is not canon: 00-world.md counts the ones left out in its opening lines, and with --proposed each kept one is labelled proposed, not yet reviewed under its heading, on its index line and on every link to it; a file past a cap splits into parts, and small files merge when the parts would pass the source cap (--words, --bytes, --files). Nothing is invented and nothing is cut: an unresolved link stays as its id, an oversize element gets a file of its own. A world with its own words for the types (x_names, below) gets them on its headings and element lines, the standard's type in brackets for the AI: ### Employees (Character); file names keep the standard's slugs. The output directory must not exist yet; the same world gives the same bytes. Beside it: the OKF bundle (above) is one file per element with frontmatter and links as paths, for knowledge tools that follow links; the pack is merged prose for a notebook that reads it whole.

Proposed changes

A member can propose a change for a keeper to accept instead of writing it (keel #519), and a world's owner can make every AI seat work that way:

row = client.propose("character", cid, {"reputation": "Feared"}, evidence="Chapter 12")
for p in client.walk_world_proposals(state="open"):  # newest first; keepers see every row, others their own
    ...
client.accept_proposal("character", cid, row["id"])  # the keeper's own write; StaleProposal when a proposed field moved
client.decline_proposal("character", cid, row["id"], reason="not yet")
client.withdraw_proposal("character", cid, row["id"])  # the proposer's
client.review_new_element("character", new_id, "accept")  # a marked new element; decline deletes it

With AI seats propose, a person accepts on for the world (agent_writes_proposed on get_world(), read-only), a seat's patch, upsert or link edit raises ProposedError (.proposal) instead of returning a body that is not the element; push counts those as proposed and never resends one. A seat's new element lands marked (x_review_proposed) and the queue lists it as a kind: "new_element" row beside the proposals (kind: "proposal"); a keeper settles it with review_new_element (422 not_proposed, is_not_proposed, when there is no mark). A proposal's fields holds only what differs from the element, and a PUT's omitted names the fields it would empty; a seat's write that changes nothing makes no row and answers 200. The doctor's flag 12 reports the queue's open and lapsed rows, the proposals and the marked new elements counted apart.

The doctor

The doctor checks a world and reports what its keeper should look at, each finding with the element and the reason. It only reads: it never changes a file or sends a write.

python -m onlyworlds.doctor moppetopia            # a world folder
OW_API_KEY=ow_r_… python -m onlyworlds.doctor --api   # the key's world (OW_API_PIN when the key needs one)
from onlyworlds import read_folder
from onlyworlds.doctor import doctor_folder

report = doctor_folder(read_folder("moppetopia"))
for finding in report.findings:
    print(finding["flag"], finding["element"], finding["reason"])

It looks for links to elements that are not there, Events that end before they start, elements with no links in or out, two elements of one type whose names collide once case, character width, spacing and punctuation are folded (a name shared across types, such as a Location and its Pin, is not one), stubs never filled (a known-of element older than --days, or an element with no text in any of its type's text fields, on every type but map, marker, pin and relation), and a charter Law untouched after --changes changes in the feed. On a folder it also checks world.json: a format_version older than this package's (flag 9), and an api_key in a world.json that git tracks (flag 10). On both doors it reads the world's own house rules (flag 11, the checks profile: x_checks_rules on the Charter Law, applied by onlyworlds.apply_checks) and names every element a rule finds short, with each rule's scoped and short counts, and every rule it cannot apply, with why. Each check says how much it could compare, and one it cannot run (a folder has no change feed; nothing records whether a message was read) is listed with the reason, never counted as clean. A finding on a message or a room thing (supertype Message, Board, Work item, Spec, Heart, Vote or Sitting) is marked "kind": "message" and listed last, outside the count of things to look at. --json prints the same report for a tool, and --packet prints the findings as work packets (the packet extension profile: a Relation with supertype Work item, its elements with a reason each under x_packet_elements, a JSON array, empty when clean) for a tool that lands them in a world or a parlor. The exit code is 0 when there is nothing to look at, 1 when there is (a message finding too), 2 when nothing could be read.

The API door reads the key from OW_API_KEY and the PIN from OW_API_PIN, or from the toolkit's ONLYWORLDS_API_KEY and ONLYWORLDS_API_PIN when those are not set.

Errors

A failed call raises ApiError with the API's code, param (the field that failed) and doc_url, and flags for the cases you handle differently: is_validation_error, is_auth_error, is_not_author, is_owner_only, is_id_conflict, is_resync_required, is_busy, is_precondition_failed.

To write only if nobody changed an element since you read it, pass its change_seq: client.patch("character", el["id"], {"name": "Mine"}, if_match=el["change_seq"]) (on edit_links too). If it moved, nothing is written and PreconditionFailed is raised, whose current is the element as it is now. The server never merges: compare the copy you edited from with current, and send again with if_match=current["change_seq"] or ask which version to keep.

The change feed

walk = client.walk_changes(saved_cursor)  # None for everything
for op in walk:
    ...  # op["op"] is "upsert" or "delete"
saved_cursor = walk.cursor  # opaque: keep it, don't parse it

To poll for news, client.head() reads where the feed is now, {head, cursor, epoch, unchanged}, without a page. The client keeps the last answer's ETag and sends it back as If-None-Match, so a poll where nothing moved is a 304 that keel does not count against the request budget, and head() then answers what it last saw with unchanged=True. Walk when head or epoch differs from the ones you last applied, and sleep between polls yourself: the package never sleeps for you. The method's docstring has the loop.

Schema

The 22 element types and each field's kind are generated from the published schema distribution; onlyworlds.SCHEMA_VERSION and onlyworlds.SCHEMA_DIST_TAG say which one. onlyworlds.PUBLIC_FIELDS names, per type, the fields OnlyWorlds shows on a known-of element (one a guest sees only one step out) beside its name and description. The package's own version doesn't track the schema's.

The room

A parlor (a chat room on a world) keeps its lines, board entries, work items, specs, hearts, votes, sittings and member cards in the world as ordinary elements, told apart by type and supertype and marked with x_room_of (the id of the world or sitting the room belongs to). onlyworlds.ROOM_PAIRS is the room profile's list of those (type, supertype) pairs, and onlyworlds.is_room_element(element) is true when a body or a FolderElement matches one (trimmed, case-blind, the whole word) and carries the mark. A pair without the mark is lore: a novel's letters are Narratives with supertype Message.

from onlyworlds import is_room_element

# a lore view, unless the person asks for the room
lore = [el for el in folder.all_elements().values() if not is_room_element(el)]

A lore view (a list, a count, search, a graph, a story's context, a survey) leaves room elements out unless the person asks; the room's own views show them whole, and a writer writes them like any element. A Law with supertype Charter is lore on purpose. The profile: profiles/room.md in the schema distribution.

A world's own words

A world may name its types in its own words: in a work world a Character is an Employee. The map is x_names on the World (world.json in a folder), and only what a person reads changes; the API, the folder and every id keep the standard's names. onlyworlds.type_word(world, element_type, supertype=None, many=False) gives the word to print, or None for the standard's name:

from onlyworlds import type_word

word = type_word(folder.world, "character", many=True) or "Characters"

The profile: profiles/names.md.

Not yet: typed element models, the account routes, a snapshot writer.

Develop

uv sync
uv run pytest -q
uv run ruff check . && uv run mypy --strict src
uv run python codegen/generate_schema.py --check
uv run python codegen/generate_casefold.py --check

Three groups of tests run only when asked:

  • OW_CONFORMANCE_FIXTURE=<path>: the folder-format conformance fixture from the Atlas repo.
  • OW_OPENAPI=https://www.onlyworlds.com/api/v2/openapi.json (or a saved copy): checks the client's routes, parameters and fields against the API's own OpenAPI document.
  • OW_LIVE=1: read-only calls against the demo world.

OKF_REFERENCE_SRC=<path> (the src/ of an open-knowledge-format checkout) also runs OKF's own parser and viewer over an exported bundle.

License

MIT. See LICENSE.

Metadata

Release files for onlyworlds 0.41.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for onlyworlds 0.41.0
File Size Uploaded Provenance
onlyworlds-0.41.0.tar.gz 599.4 kB With provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for onlyworlds 0.41.0
File Interpreter ABI Platform
onlyworlds-0.41.0-py3-none-any.whl Python 3 none any Details

Total release size: 810.0 kB

Release files / onlyworlds-0.41.0.tar.gz

Download URL onlyworlds-0.41.0.tar.gz
Size 599.4 kB
Tags Source
SHA-256 checksum
How to use checksums
e8d7bf05c8ecb0e016622ab30cc36bfe5944ca81d3395c6031aff3b399363649
BLAKE2b-256 checksum
How to use checksums
95e6d22fa0426e1eacdb9b17f2e67d6b98344f5b40ff35717b0f34cb7e40e443
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 23, 2026.

Transparency log

Release files / onlyworlds-0.41.0-py3-none-any.whl

Download URL onlyworlds-0.41.0-py3-none-any.whl
Size 210.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
a832bd703e62767641e890a378816caf1ac5c090c0ff9566a28254c7ed0472af
BLAKE2b-256 checksum
How to use checksums
efb1c7c5cda32521bff3e980489a98fdcfc62600f9870cdd60c9194c27cf7b90
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 23, 2026.

Transparency log

Security for onlyworlds 0.41.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

All 2 files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for onlyworlds-0.41.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 23, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.41.0 This release

2 release files Full provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page