Skip to content

AIsketcher 0.2.1

Published on PyPI

AIsketcher 0.2.1 is available from the PyPI project page. The commands on this page pin the public version for a reproducible install.

Install profiles

Profile Command Includes
Core python -m pip install "aisketcher==0.2.1" preparation, studies, export, replay
Studio python -m pip install "aisketcher[demo]==0.2.1" core plus local Gradio interface
Local python -m pip install "aisketcher[local]==0.2.1" core plus Diffusers runtime
Full local python -m pip install "aisketcher[local,demo]==0.2.1" local generation and Studio

The base package remains lightweight: it does not install Torch, Diffusers, model weights, or Gradio. Model files are never bundled in the distribution and are downloaded only after a user reviews and confirms a pinned preset plan.

Current package presentation

The package is a model-agnostic SDK for the traceable design-study workflow:

prepare → explore → pick → vary → export → replay

README.md is the package metadata description, so the 0.2.1 publication also updates the PyPI page from the historical AWS-era description to the current SDK, Studio, replay, and privacy model.

The English Studio guide uses an actual local capture rather than a synthetic UI composite. Its documentation-only heritage fixture records the HPO-selected direction at seed 6764547109648557242 with pinned preset sdxl-canny-lite@1. The fixture is separate from the packaged Pocket Kingdom Guided Sample and does not imply that a seed is universally best.

Release-published Trusted Publishing

Publishing a GitHub Release for the matching v0.2.1 tag starts the protected PyPI workflow automatically. Before upload, the workflow:

  1. checks that the release tag and built package metadata both name 0.2.1;
  2. runs the release test suite;
  3. builds and checks the wheel and source archive;
  4. scans distributable contents and smoke-tests the exact wheel;
  5. publishes with PyPI Trusted Publishing through a short-lived GitHub OIDC identity, without storing a PyPI API token.

The protected pypi environment and PyPI Trusted Publisher must still be configured for this repository and workflow. A GitHub Release updates the PyPI description only by publishing a new immutable package version; editing an existing Release or README does not rewrite an already-published PyPI artifact.

Compatibility and provenance

The SDK and Studio boundary remains the v0.2 contract. The deprecated AIsketcher.img2img facade is still available for caller-owned compatible pipelines and remains scheduled for removal in 0.3.0. Historical 0.2.0 release notes and sample provenance remain unchanged.

Model weights, private inputs, and artwork are not part of the MIT-licensed package. Review the artwork notice before reusing repository images.