Skip to main content

Coworld certification, upload, and runner tooling

Project description

Coworld

Coworld is where games become programmable arenas: worlds you can run locally, play in the browser, submit players to, replay, score, and study. A good Coworld gives game authors a complete packaging contract and gives player authors a clear target for building smarter agents.

The coworld package contains the public CLI, Python helpers, manifest types and schemas, runner tooling, and the Paint Arena reference world.

Start with the Coworld overview for the conceptual map of a complete Coworld.

What Is A Coworld?

A Coworld is a game environment built around a player-improvement loop. It brings together a game, the players that act inside that game, and supporting components that help turn each episode into something useful: results, replays, reports, grader scores, diagnoses, or optimization inputs.

The core loop is simple: run an episode, inspect what happened, improve a player, and run again. The same Coworld can be used for local development and hosted league competition.

Most readers are here to build a player for an existing Coworld. If that is you, start with Developing Players.

If you are building a complete Coworld, start with Developing Coworlds.

Developing Players (Work In Progress)

Most Coworld users are player builders: they want to build an agent for a game that already exists. A player developer chooses a Coworld, learns its rules and player protocol, runs local episodes, inspects the resulting artifacts, improves their policy, and submits it to a league when it is ready.

The player development user guide is still under construction. For now, use:

Developing Coworlds (Work In Progress)

Coworld builders create the worlds that player developers target. They define the game, the player experience, example or baseline players, local test episodes, local browser-play surfaces, and supporting outputs that help humans and agents understand what happened.

The Coworld development user guide is still under construction. For now, start with the Paint Arena example and use the Documentation Map for the current reference docs.

Main Workflows

Workflow Start with
Build or improve a player Cookbook: Upload And Submit A Player and Player role
Run local episodes or browser play Cookbook: Build And Run Paint Arena Locally
Inspect league status, logs, results, and replays Cookbook: Watch Results And Find Episodes
Build, certify, and upload a Coworld Cookbook: Certify And Upload A Coworld
Improve a policy in the optimizer workbench uv run coworld optimize and Optimizer role
Understand package structure and manifest fields Manifest reference

What This Package Provides

  • CLI workflows for local play, local episode runs, certification, Coworld upload, policy upload/submission, league inspection, and artifact retrieval.
  • Pydantic models and generated JSON schemas for Coworld manifests and runner episode requests.
  • Local and Kubernetes runner code for executing Coworld episodes.
  • Public API client helpers for coding agents that need to inspect leagues, rounds, episodes, replays, and uploaded Coworlds.
  • The Paint Arena example, which is the canonical example used by this package documentation.

Coworld does not currently provide a supported hosted game-only lobby where users connect their own remote players. Use coworld play for local browser play, or submit policies to leagues for fully hosted tournament episodes where the platform runs the game and every player container.

Documentation Map

The Coworld docs are being reorganized. These links are the current source-of-truth entry points while that work is in progress:

Need Current doc
Understand what a complete Coworld is Coworld overview
Build or operate from recipes Coworld cookbook
Understand manifest fields Manifest reference
Understand roles and artifact flow Coworld overview
Implement a game runnable Game role
Implement or submit a player Player role and Coworld cookbook
Understand artifact contracts Artifact reference
Consume episode artifacts as a unit Episode bundle reference
Understand the episode lifecycle Lifecycle overview
Debug local or hosted execution Local runner and Kubernetes runner
Start from the canonical example Paint Arena
Look up exact CLI or API reference uv run coworld --help, uv run coworld <command> --help, and Observatory OpenAPI

Planned cleanup will shrink or move several of these pages into a smaller set of cookbook, lifecycle, runnable, and artifact documents. Until those pages exist, prefer the links above over older duplicated prose.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

coworld-0.1.13.tar.gz (233.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

coworld-0.1.13-py3-none-any.whl (208.2 kB view details)

Uploaded Python 3

File details

Details for the file coworld-0.1.13.tar.gz.

File metadata

  • Download URL: coworld-0.1.13.tar.gz
  • Upload date:
  • Size: 233.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for coworld-0.1.13.tar.gz
Algorithm Hash digest
SHA256 2b0e7a1405eccb6ea8b035d25d6ee94420db0fcfbab7dce9bc86a4bddef76e4d
MD5 129ce45a9450b839ebb771a0c56a4fda
BLAKE2b-256 5e6abf2187f296e23aa92b91b2aa9c772e4ebc5f2e6287473b282a626e72c773

See more details on using hashes here.

Provenance

The following attestation bundles were made for coworld-0.1.13.tar.gz:

Publisher: release-coworld.yml on Metta-AI/metta

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file coworld-0.1.13-py3-none-any.whl.

File metadata

  • Download URL: coworld-0.1.13-py3-none-any.whl
  • Upload date:
  • Size: 208.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for coworld-0.1.13-py3-none-any.whl
Algorithm Hash digest
SHA256 e1891f57d3c2b22692e958559500a865b08ba0c66442937384d2380d0eac72f1
MD5 6cb323386c655042368971cafbd9662c
BLAKE2b-256 7bc7af2a7db4f64b06d49aa6237940ee4db8c1f2c59dc012d52e9682ad156d69

See more details on using hashes here.

Provenance

The following attestation bundles were made for coworld-0.1.13-py3-none-any.whl:

Publisher: release-coworld.yml on Metta-AI/metta

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page