Auto1111SDK¶

Development line: project:auto1111sdk · thread auto1111sdk-development
Last event: 2024-03-05 · 2 dated since 2024-02-06 · Researched: 2026-09-04 · confidence: medium

What it is¶

Auto1111SDK is a Python library for developers driving local Stable Diffusion workflows from code instead of the Automatic1111 WebUI.

• Text-to-image, image-to-image, inpainting, outpainting, and Stable Diffusion upscale.
• ESRGAN/Real-ESRGAN upscaling, CivitAI downloads, custom VAE/SDXL support, and a ControlNet interface.

Development line¶

• 2024-02-06 — Auto1111SDK project repository and Colab notebook appeared. The project used the personal saketh12/Auto1111SDK repository path and a Colab demo.
• 2024-03-05 — Auto1111SDK organization repository appeared with the Colab notebook. On 2024-03-05, the repository moved under the Auto1111SDK GitHub organization alongside the linked Colab notebook. The change marks a public entry point, but the links do not show whether it was a repository migration, a release, or a repost.

What changed¶

Auto1111SDK stopped development in early-2024 with no later PyPI release. On 2024-02-06, the project started under the personal saketh12/Auto1111SDK repository path with a Colab demo. On 2024-02-16, PyPI published 0.0.95, which remains the latest release in its history. On 2024-03-05, the repository path changed to Auto1111SDK/Auto1111SDK while retaining the same Colab file; this shows a publishing path change rather than new code. On 2026-09-04, the GitHub README still documents virtualenv-only installation, the old personal Git install path, and FP32-only ControlNet. The package is now a legacy dependency with unverified current-runtime compatibility.

How to use this¶

From 2024-03-05, treat the Auto1111SDK organization repository as the historical project entry point and verify current setup separately.

1. Create a clean Python virtual environment; the project explicitly recommends a virtualenv and says Conda is unsupported. — https://github.com/Auto1111SDK/Auto1111SDK
2. Install the reproducible last package release with pip install auto1111sdk==0.0.95, then verify the version and files against PyPI. — https://pypi.org/project/auto1111sdk/
3. Place a compatible local checkpoint or .safetensors file on disk, construct StableDiffusionPipeline(path), and call generate_txt2img with a prompt, dimensions, and steps. — https://github.com/Auto1111SDK/Auto1111SDK
4. Save the first generated image with output[0].save(...); use this minimal flow as the environment smoke test before adding a larger workflow. — https://github.com/Auto1111SDK/Auto1111SDK
5. For ControlNet, create ControlNetModel with the model name without its extension and an input image, pass it to the pipeline, and provision FP32 VRAM. — https://github.com/Auto1111SDK/Auto1111SDK

Best practices¶

• Use an isolated virtual environment rather than Conda; this is the only environment support boundary stated upstream. — https://github.com/Auto1111SDK/Auto1111SDK
• Pin Auto1111SDK, Python, PyTorch, CUDA, and model-weight versions in a lockfile or image: the latest published package is pre-alpha 0.0.95 from 2024-02-16. — https://pypi.org/project/auto1111sdk/
• Do not use the README’s unpinned pip install git+https://github.com/saketh12/Auto1111SDK.git for a reproducible workflow; use a reviewed immutable source revision if source installation is necessary. — https://github.com/Auto1111SDK/Auto1111SDK
• Treat ControlNet as FP32-only and measure GPU memory before committing to it; do not assume the README’s planned FP16 support shipped. — https://github.com/Auto1111SDK/Auto1111SDK

Superseded by this¶

• 2024-02-06 — the personal saketh12/Auto1111SDK link is superseded as the canonical reference by the organization path recorded on 2024-03-05; both resolve to the same named project today.
• 2024-02-16 — “install the latest” must not be read as current-generation diffusion support: it resolves to pre-alpha 0.0.95, released on 2024-02-16.
• 2024-03-05 — treating the Git-source install as evidence of newer ControlNet support is obsolete; the current README still documents ControlNet as FP32-only.

Still unknown¶

• The Colab notebook at https://colab.research.google.com/drive/1SekiJ-mdB2V8ogWbyRyF_yDnoMuDGWTl could not be read without a valid Google session, so its current runtime, package pins, hardware, and execution state are unverified.
• The dated links establish a personal-to-organization GitHub path change on 2024-03-05, but do not establish whether it was a repository transfer, a mirror, or a code release; no release note or commit diff for 2024-03-05 was available.
• Current compatibility with 2026 Python, PyTorch, CUDA, and model formats is not established by the sources read.

Sources¶

Agent brief¶

• Subject: project:auto1111sdk, thread auto1111sdk-development, 2 dated events 2024-02-06 → 2024-03-05.
• Practical note: From 2024-03-05, treat the Auto1111SDK organization repository as the recorded historical project entry point and validate current setup instructions separately.
• Confidence: medium. Dated supersedes above are the authority for what is obsolete.