Skip to main content

Python Client and Model Management for CivitAI.

Project description

pycivitai

PyPI PyPI - Python Version Loc Comments

Code Test Package Release codecov

GitHub Org's stars GitHub stars GitHub forks GitHub commit activity GitHub issues GitHub pulls Contributors GitHub license

Python Client and Model Management for civitai,

The design was inspired by huggingface/huggingface_hub, implementing automatic management of downloaded models locally, and automatically detecting file changes and the latest versions.

Installation

You can install pycivitai with pip

pip install pycivitai

If you need TUI for manually deleting downloaded models, just install like this

pip install pycivitai[cli]

Quick Start

Use Model From CivitAI

pycivitai follows a similar usage pattern as the huggingface_hub, as shown below:

from pycivitai import civitai_download

if __name__ == '__main__':
    # get the latest version of DEN_barbucci_artstyle (either model title or id is okay)
    # Homepage of this model: https://civitai.com/models/85716
    # the return value of civitai_download is the path of downloaded file you need
    print(civitai_download('DEN_barbucci_artstyle'))

    # get the same model (DEN_barbucci_artstyle) with model id
    print(civitai_download(85716))

    # get the specific version (either version name of id is okay)
    print(civitai_download('DEN_barbucci_artstyle', version='v1.0'))
    print(civitai_download('DEN_barbucci_artstyle', 91158))

    # This is a model with multiple files inside.
    # It contains a safetensors and a vae, the safetensors is primary.
    # Homepage of this model: https://civitai.com/models/6755/cetus-mix
    print(civitai_download('Cetus-Mix'))  # the safetensors file (when file not specified, primary file will be chosen)
    print(civitai_download('Cetus-Mix', file='*.vae.pt'))  # get the vae file

If you are using command line, you can get the models with the following commands

pycivitai get -m 'DEN_barbucci_artstyle'           # get model, use primary file of the latest version
pycivitai get -m 'DEN_barbucci_artstyle' -v 'v1.0' # get model, use primary file the v1.0 version
pycivitai get -m 'Cetus-Mix'                       # get the safetensors file of Cetus-Mix
pycivitai get -m 'Cetus-Mix' -f '*.vae.pt'         # get the vae file of Cetus-Mix

Get Information of Model Resource

If you only need to obtain information about the model's resource files, for example, if you need to download the files yourself, you can use the civitai_find_online function, which will return a Resource object.

from pycivitai import civitai_find_online, Resource

if __name__ == '__main__':
    # get resource of this model file, arguments are the same as `civitai_download`
    resource: Resource = civitai_find_online('DEN_barbucci_artstyle')

    # information of this resource
    print(resource.model_name)
    print(resource.model_id)
    print(resource.version_name)
    print(resource.version_id)
    print(resource.filename)
    print(resource.url)
    print(resource.size)
    print(resource.sha256)
    print(resource.is_primary)

Clear the Downloaded Models

If you need to delete all the local models, just

pycivitai delete-cache -A  # download all models

or use the TUI to choose which one to delete

pip install pycivitai[cli]  # this step is necessary
pycivitai delete-cache

TUI is like this, build with InquirerPy:

? Choose model versions to delete:
  Model cetus_mix(ID: 6755, 2 files, size: 3.899 GiB):
❯ ○ cetusmix_whalefall2(ID: 105924, 2 files, size: 3.899 GiB)

  Model den_barbucci_artstyle(ID: 85716, 2 files, size: 17.181 KiB):
  ○ v1_0(ID: 91158, 1 file, size: 7.160 KiB)
  ○ v2_0(ID: 113049, 1 file, size: 10.021 KiB)

F.A.Q.

Where will the downloaded model be saved?

The downloaded model will be saved by default in the ~/.cache/civitai directory. If you need to change the save path, you can set the value of the CIVITAI_HOME environment variable to your desired save path.

How are downloaded models managed?

After downloading, the models are managed using the file system, and file locks are set to ensure thread safety during reading and writing. The structure is similar to the following:

~/.cache/civitai
├── cetus_mix__6755
│   └── cetusmix_whalefall2__105924
│       ├── files
│       │   ├── cetusMix_Whalefall2.safetensors
│       │   └── vae-ft-mse-840000-ema-pruned.vae.pt
│       ├── hashes
│       │   ├── cetusMix_Whalefall2.safetensors.hash
│       │   └── vae-ft-mse-840000-ema-pruned.vae.pt.hash
│       └── primary
└── den_barbucci_artstyle__85716
    ├── v1_0__91158
    │   ├── files
    │   │   └── DEN_barbucci_artstyle.pt
    │   ├── hashes
    │   │   └── DEN_barbucci_artstyle.pt.hash
    │   └── primary
    └── v2_0__113049
        ├── files
        │   └── DEN_barbucci_styleMK2.pt
        ├── hashes
        │   └── DEN_barbucci_styleMK2.pt.hash
        └── primary

This structure does not have any additional dependencies. Therefore, when it is necessary to migrate the storage path, you can simply move it and modify the environment variable CIVITAI_HOME.

What happens when in an offline environment?

When a network anomaly or an offline environment is detected (such as connection errors or timeouts), the library will attempt to find a suitable model from the locally downloaded models to provide to the user. If no locally available model is found, an exception will be raised.

Additionally, if you are in a clearly offline environment and using pre-downloaded models, you can set the value of the CIVITAI_OFFLINE environment variable to avoid attempts for network access.

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

pycivitai-0.2.0.tar.gz (27.1 kB view details)

Uploaded Source

Built Distribution

pycivitai-0.2.0-py3-none-any.whl (27.1 kB view details)

Uploaded Python 3

File details

Details for the file pycivitai-0.2.0.tar.gz.

File metadata

  • Download URL: pycivitai-0.2.0.tar.gz
  • Upload date:
  • Size: 27.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.11.8

File hashes

Hashes for pycivitai-0.2.0.tar.gz
Algorithm Hash digest
SHA256 985fad1b6c693dad65e7d66d7b21932d153fbd605ebee041451b0765bdaa59cd
MD5 fe77a286d58ea8beed73737c4dfddad5
BLAKE2b-256 8f82a48e166bd026c1781e78a4cc6c193980ba6e163448624356006830aad54b

See more details on using hashes here.

File details

Details for the file pycivitai-0.2.0-py3-none-any.whl.

File metadata

  • Download URL: pycivitai-0.2.0-py3-none-any.whl
  • Upload date:
  • Size: 27.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.11.8

File hashes

Hashes for pycivitai-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 be8489077db0cde49e63ec771a57604ac4fe305be6f86499133af0ba55ffd390
MD5 150b50de97e7ee4ae7c0c13e77409065
BLAKE2b-256 6e3fefbfb36d4332064a883499ee68eac220e9127c20425f92a3e3d87d43f996

See more details on using hashes here.

Supported by

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