app.py

import requests
import os
import gradio as gr
from huggingface_hub import update_repo_visibility, whoami, upload_folder, create_repo, upload_file, update_repo_visibility
from slugify import slugify
import gradio as gr
import re
import uuid
from typing import Optional
import json
from bs4 import BeautifulSoup

TRUSTED_UPLOADERS = ["KappaNeuro", "CiroN2022", "multimodalart", "Norod78", "joachimsallstrom", "blink7630", "e-n-v-y", "DoctorDiffusion", "RalFinger", "artificialguybr"]

def get_json_data(url):
    url_split = url.split('/')
    print("url_split: ", url_split)
    api_url = f"https://civitai.com/api/v1/models/{url_split[4]}"
    try:
        response = requests.get(api_url)
        response.raise_for_status()
        return response.json()
    except requests.exceptions.RequestException as e:
        print(f"Error fetching JSON data: {e}")
        return None

def check_nsfw(json_data, profile):
    if json_data["nsfw"]:
        return False
    print(profile)
    if(profile.username in TRUSTED_UPLOADERS):
        return True
    for model_version in json_data["modelVersions"]:
        for image in model_version["images"]:
            if image["nsfw"] > 5:
                return False
    return True

def get_prompts_from_image(image_id):
    print("image_id: ", image_id)
    url = f'https://civitai.com/images/{image_id}'
    response = requests.get(url)
    soup = BeautifulSoup(response.text, 'html.parser')
    content = soup.find_all(class_='mantine-Code-root mantine-Code-block mantine-2v44jn')
    print(content)
    if(content):
        if(len(content) > 1):
            return content[0].text, content[1].text
        else:
            return content[0].text, ""
    else:
        return "", ""

def extract_info(json_data):
    if json_data["type"] == "LORA":
        for model_version in json_data["modelVersions"]:
            if model_version["baseModel"] in ["SDXL 1.0", "SDXL 0.9", "SD 1.5", "SD 1.4", "SD 2.1", "SD 2.0", "SD 2.0 768", "SD 2.1 768"]:
                for file in model_version["files"]:
                    print(file)
                    if "primary" in file:
                        # Start by adding the primary file to the list
                        urls_to_download = [{"url": file["downloadUrl"], "filename": file["name"], "type": "weightName"}]
                        
                        # Then append all image URLs to the list
                        for image in model_version["images"]:
                            image_id = image["url"].split("/")[-1].split(".")[0]
                            prompt, negative_prompt = get_prompts_from_image(image_id)
                            if image["nsfwLevel"] > 5:
                                pass #ugly before checking the actual logic
                            else:
                                urls_to_download.append({
                                    "url": image["url"],
                                    "filename": os.path.basename(image["url"]),
                                    "type": "imageName",
                                    "prompt": prompt, #if "meta" in image and "prompt" in image["meta"] else ""
                                    "negative_prompt": negative_prompt
                                })
                        model_mapping = {
                            "SDXL 1.0": "stabilityai/stable-diffusion-xl-base-1.0",
                            "SDXL 0.9": "stabilityai/stable-diffusion-xl-base-1.0",
                            "SD 1.5": "runwayml/stable-diffusion-v1-5",
                            "SD 1.4": "CompVis/stable-diffusion-v1-4",
                            "SD 2.1": "stabilityai/stable-diffusion-2-1-base",
                            "SD 2.0": "stabilityai/stable-diffusion-2-base",
                            "SD 2.1 768": "stabilityai/stable-diffusion-2-1",
                            "SD 2.0 768": "stabilityai/stable-diffusion-2"
                        }
                        base_model = model_mapping[model_version["baseModel"]]
                        print(json_data)
                        info = {
                            "urls_to_download": urls_to_download,
                            "id": model_version["id"],
                            "baseModel": base_model,
                            "modelId": model_version.get("modelId", ""),
                            "name": json_data["name"],
                            "description": json_data["description"],
                            "trainedWords": model_version["trainedWords"] if "trainedWords" in model_version else [],
                            "creator": json_data["creator"]["username"],
                            "tags": json_data["tags"],
                            "allowNoCredit": json_data["allowNoCredit"],
                            "allowCommercialUse": json_data["allowCommercialUse"],
                            "allowDerivatives": json_data["allowDerivatives"],
                            "allowDifferentLicense": json_data["allowDifferentLicense"]
                        }
                        return info
    return None

def download_files(info, folder="."):
    downloaded_files = {
        "imageName": [],
        "imagePrompt": [],
        "imageNegativePrompt": [],
        "weightName": []
    }
    for item in info["urls_to_download"]:
        download_file(item["url"], item["filename"], folder)
        downloaded_files[item["type"]].append(item["filename"])
        if(item["type"] == "imageName"):
            prompt_clean = re.sub(r'<.*?>', '', item["prompt"])
            negative_prompt_clean = re.sub(r'<.*?>', '', item["negative_prompt"])
            downloaded_files["imagePrompt"].append(prompt_clean)
            downloaded_files["imageNegativePrompt"].append(negative_prompt_clean)
    return downloaded_files

def download_file(url, filename, folder="."):
    headers = {}
    try:
        response = requests.get(url, headers=headers)
        response.raise_for_status()
    except requests.exceptions.HTTPError as e:
        print(e)
        if response.status_code == 401: 
            headers['Authorization'] = f'Bearer {os.environ["CIVITAI_API"]}'
            try:
                response = requests.get(url, headers=headers)
                response.raise_for_status()
            except requests.exceptions.RequestException as e:
                raise gr.Error(f"Error downloading file: {e}")
        else:
            raise gr.Error(f"Error downloading file: {e}")
    except requests.exceptions.RequestException as e:
        raise gr.Error(f"Error downloading file: {e}")

    with open(f"{folder}/{filename}", 'wb') as f:
        f.write(response.content)

def process_url(url, profile, do_download=True, folder="."):
    json_data = get_json_data(url)
    if json_data:
        if check_nsfw(json_data, profile):
            info = extract_info(json_data)
            if info:
                if(do_download):
                    downloaded_files = download_files(info, folder)
                else:
                    downloaded_files = []
                return info, downloaded_files
            else:
                raise gr.Error("Only SDXL LoRAs are supported for now")
        else:
            raise gr.Error("This model has content tagged as unsafe by CivitAI")
    else:
        raise gr.Error("Something went wrong in fetching CivitAI API")

def create_readme(info, downloaded_files, user_repo_id, link_civit=False, is_author=True, folder="."):
    readme_content = ""
    original_url = f"https://civitai.com/models/{info['modelId']}"
    link_civit_disclaimer = f'([CivitAI]({original_url}))'
    non_author_disclaimer = f'This model was originally uploaded on [CivitAI]({original_url}), by [{info["creator"]}](https://civitai.com/user/{info["creator"]}/models). The information below was provided by the author on CivitAI:'
    default_tags = ["text-to-image", "stable-diffusion", "lora", "diffusers", "template:sd-lora", "migrated"]
    civit_tags = [t.replace(":", "") for t in info["tags"] if t not in default_tags]
    tags = default_tags + civit_tags
    unpacked_tags = "\n- ".join(tags)
    
    trained_words = info['trainedWords'] if 'trainedWords' in info and info['trainedWords'] else []
    formatted_words = ', '.join(f'`{word}`' for word in trained_words)
    if formatted_words:
        trigger_words_section = f"""## Trigger words
You should use {formatted_words} to trigger the image generation.
    """
    else:
        trigger_words_section = ""
    
    widget_content = ""
    for index, (prompt, negative_prompt, image) in enumerate(zip(downloaded_files["imagePrompt"], downloaded_files["imageNegativePrompt"], downloaded_files["imageName"])):
        escaped_prompt = prompt.replace("'", "''")
        negative_prompt_content = f"""parameters:
    negative_prompt: {negative_prompt}
            """ if negative_prompt else ""
        widget_content += f"""- text: '{escaped_prompt if escaped_prompt else ' ' }'
  {negative_prompt_content}
  output:
    url: >-
      {image}
"""

    content = f"""---
license: other
license_name: bespoke-lora-trained-license
license_link: https://multimodal.art/civitai-licenses?allowNoCredit={info["allowNoCredit"]}&allowCommercialUse={info["allowCommercialUse"][0]}&allowDerivatives={info["allowDerivatives"]}&allowDifferentLicense={info["allowDifferentLicense"]}
tags:
- {unpacked_tags}

base_model: {info["baseModel"]}
instance_prompt: {info['trainedWords'][0] if 'trainedWords' in info and len(info['trainedWords']) > 0 else ''}
widget:
{widget_content}
---

# {info["name"]} 

<Gallery />

{non_author_disclaimer if not is_author else ''}

{link_civit_disclaimer if link_civit else ''}

## Model description

{info["description"]}

{trigger_words_section}

## Download model

Weights for this model are available in Safetensors format.

[Download](/{user_repo_id}/tree/main) them in the Files & versions tab.

## Use it with the [🧨 diffusers library](https://github.com/huggingface/diffusers)

```py
from diffusers import AutoPipelineForText2Image
import torch

pipeline = AutoPipelineForText2Image.from_pretrained('{info["baseModel"]}', torch_dtype=torch.float16).to('cuda')
pipeline.load_lora_weights('{user_repo_id}', weight_name='{downloaded_files["weightName"][0]}')
image = pipeline('{prompt if prompt else (formatted_words if formatted_words else 'Your custom prompt')}').images[0]
```

For more details, including weighting, merging and fusing LoRAs, check the [documentation on loading LoRAs in diffusers](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading_adapters)
"""
    #for index, (image, prompt) in enumerate(zip(downloaded_files["imageName"], downloaded_files["imagePrompt"])):
    #    if index == 1:
    #        content += f"## Image examples for the model:\n![Image {index}]({image})\n> {prompt}\n"
    #    elif index > 1:
    #        content += f"\n![Image {index}]({image})\n> {prompt}\n"
    readme_content += content + "\n"
    print(readme_content)
    with open(f"{folder}/README.md", "w") as file:
        file.write(readme_content)

def get_creator(username):
    url = f"https://civitai.com/api/trpc/user.getCreator?input=%7B%22json%22%3A%7B%22username%22%3A%22{username}%22%2C%22authed%22%3Atrue%7D%7D"
    headers = {
        "authority": "civitai.com",
        "accept": "*/*",
        "accept-language": "en-BR,en;q=0.9,pt-BR;q=0.8,pt;q=0.7,es-ES;q=0.6,es;q=0.5,de-LI;q=0.4,de;q=0.3,en-GB;q=0.2,en-US;q=0.1,sk;q=0.1",
        "content-type": "application/json",
        "cookie": f'{os.environ["COOKIE_INFO"]}',
        "if-modified-since": "Tue, 22 Aug 2023 07:18:52 GMT",
        "referer": f"https://civitai.com/user/{username}/models",
        "sec-ch-ua": "\"Not.A/Brand\";v=\"8\", \"Chromium\";v=\"114\", \"Google Chrome\";v=\"114\"",
        "sec-ch-ua-mobile": "?0",
        "sec-ch-ua-platform": "macOS",
        "sec-fetch-dest": "empty",
        "sec-fetch-mode": "cors",
        "sec-fetch-site": "same-origin",
        "user-agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36"
    }
    response = requests.get(url, headers=headers)

    return response.json()

def extract_huggingface_username(username):
    data = get_creator(username)
    links = data.get('result', {}).get('data', {}).get('json', {}).get('links', [])
    for link in links:
        url = link.get('url', '')
        if url.startswith('https://huggingface.co/') or url.startswith('https://www.huggingface.co/'):
            username = url.split('/')[-1]
            return username

    return None


def check_civit_link(profile: Optional[gr.OAuthProfile], url):
    info, _ = process_url(url, profile, do_download=False)
    hf_username = extract_huggingface_username(info['creator'])
    attributes_methods = dir(profile)
    
    if(profile.username == "multimodalart"):
        return '', gr.update(interactive=True), gr.update(visible=False), gr.update(visible=True)
        
    if(not hf_username):
        no_username_text = f'If you are {info["creator"]} on CivitAI, hi! Your CivitAI profile seems to not have information about your Hugging Face account. Please visit <a href="https://civitai.com/user/account" target="_blank">https://civitai.com/user/account</a> and include your 🤗 username there, here\'s mine:<br><img width="60%" src="https://i.imgur.com/hCbo9uL.png" /><br>(if you are not {info["creator"]}, you cannot submit their model at this time)'
        return no_username_text, gr.update(interactive=False), gr.update(visible=True), gr.update(visible=False)
    if(profile.username != hf_username):
        unmatched_username_text = '<h4>Oops, the Hugging Face account in your CivitAI profile seems to be different than the one your are using here. Please visit <a href="https://civitai.com/user/account">https://civitai.com/user/account</a> and update it there to match your Hugging Face account<br><img src="https://i.imgur.com/hCbo9uL.png" /></h4>'
        return unmatched_username_text, gr.update(interactive=False), gr.update(visible=True), gr.update(visible=False)
    else:
        return '', gr.update(interactive=True), gr.update(visible=False), gr.update(visible=True)
        
def swap_fill(profile: Optional[gr.OAuthProfile]):
    if profile is None:
        return gr.update(visible=True), gr.update(visible=False)
    else:
        return gr.update(visible=False), gr.update(visible=True)

def show_output():
    return gr.update(visible=True)

def list_civit_models(username):
    url = f"https://civitai.com/api/v1/models?username={username}&limit=100"
    json_models_list = []

    while url:
        response = requests.get(url)
        data = response.json()

        # Add current page items to the list
        json_models_list.extend(data.get('items', []))

        # Check if there is a nextPage URL in the metadata
        metadata = data.get('metadata', {})
        url = metadata.get('nextPage', None)
    urls = ""
    for model in json_models_list:
        urls += f'https://civitai.com/models/{model["id"]}/{slugify(model["name"])}\n'
    
    return urls

def upload_civit_to_hf(profile: Optional[gr.OAuthProfile], oauth_token: gr.OAuthToken, url, link_civit=False):
    if not profile.name:
        return gr.Error("Are you sure you are logged in?")
    
    folder = str(uuid.uuid4())
    os.makedirs(folder, exist_ok=False)
    info, downloaded_files = process_url(url, profile, folder=folder)
    print("Civit URL: ", url)
    print("Downloaded files: ", downloaded_files)
    username = {profile.username}
    slug_name = slugify(info["name"])
    user_repo_id = f"{profile.username}/{slug_name}"
    create_readme(info, downloaded_files, user_repo_id, link_civit, folder=folder)
    try:
        print("Repo ID: ", user_repo_id)
        print("Current folder: ", folder)
        print(oauth_token.token)
        print("Creating repo...")
        print(user_repo_id)
        create_repo(repo_id=user_repo_id, private=True, exist_ok=True, token=oauth_token.token)
        print("Uploading repo...")
        upload_folder(
            folder_path=folder,
            repo_id=user_repo_id,
            repo_type="model",
            token=oauth_token.token
        )
        update_repo_visibility(repo_id=user_repo_id, private=False, token=oauth_token.token)
    except Exception as e:
        print(e)
        raise gr.Error("Your Hugging Face Token expired. Log out and in again to upload your models.")
        
    #transfer_repos = gr.load("multimodalart/transfer_repos", hf_token=os.environ["HUGGING_FACE_HUB_TOKEN"], src="spaces")
    
    #response_code = transfer_repos(repo_id, user_repo_id)
    #i = 0
    #while response_code != "200":
    #    message = None
    #    if response_code == "409":
    #        if i < 3:
    #            user_repo_id = f"{profile.preferred_username}/{slug_name}-{i}"
    #            response_code = transfer_repos(repo_id, user_repo_id)
    #            i += 1
    #        else:
    #            message = "It seems this model has been uploaded already in your account."
    #    elif response_code == "404":
    #        message = "Something went wrong with the model upload. Try again."
    #    else:
    #        message = f"Unexpected response code: {response_code}."
    #
    #    if message:
    #        api.delete_repo(repo_id=repo_id, repo_type="model")
    #        raise gr.Error(message)

    return f'''# Model uploaded to 🤗!
    ## Access it here [{user_repo_id}](https://huggingface.co/{user_repo_id}) '''

def bulk_upload(profile: Optional[gr.OAuthProfile], oauth_token: gr.OAuthToken, urls, link_civit=False, progress=gr.Progress(track_tqdm=True)):
    urls = urls.split("\n")
    print(urls)
    for url in urls:
        print(url)
        if(url):
            try:
                upload_result = upload_civit_to_hf(profile, oauth_token, url, link_civit)
                gr.Warning(f"Model {link_civit} uploaded")
            except Exception as e:
                print(e)
                gr.Warning(f"Error uploading the model {link_civit}")
css = '''
#login {
    width: 100% !important;
    margin: 0 auto;
}
#disabled_upload{
    opacity: 0.5;
    pointer-events:none;
}
'''

with gr.Blocks(css=css) as demo:
    gr.Markdown('''# Upload your CivitAI LoRA to Hugging Face 🤗
By uploading your LoRAs to Hugging Face you get diffusers compatibility, a free GPU-based Inference Widget, and possibility to submit your model to the [LoRA the Explorer](https://huggingface.co/spaces/multimodalart/LoraTheExplorer) ✨
    ''')
    gr.LoginButton(elem_id="login")
    with gr.Column(elem_id="disabled_upload") as disabled_area:
        with gr.Row():
            submit_source_civit = gr.Textbox(
                placeholder="https://civitai.com/models/144684/pixelartredmond-pixel-art-loras-for-sd-xl",
                label="CivitAI model URL",
                info="URL of the CivitAI LoRA",
            )
        submit_button_civit = gr.Button("Upload model to Hugging Face and submit", interactive=False)
    with gr.Column(visible=False) as enabled_area:
        with gr.Column():
            submit_source_civit = gr.Textbox(
                placeholder="https://civitai.com/models/144684/pixelartredmond-pixel-art-loras-for-sd-xl",
                label="CivitAI model URL",
                info="URL of the CivitAI LoRA",
                
            )
            with gr.Accordion("Bulk upload (bring in multiple LoRAs)", open=False):
                civit_username_to_bulk = gr.Textbox(label="CivitAI username (optional)", info="Type your CivitAI username here to automagically fill the bulk models URLs list below (optional, you can paste links down here directly)")
                submit_bulk_civit = gr.Textbox(
                    label="CivitAI bulk models URLs",
                    info="Add one URL per line",
                    lines=6,
                )
                link_civit = gr.Checkbox(label="Link back to CivitAI?", value=False)
                bulk_button = gr.Button("Bulk upload")
                
        instructions = gr.HTML("")
        try_again_button = gr.Button("I have added my HF profile to my account (it may take 1 minute to refresh)", visible=False)
        submit_button_civit = gr.Button("Upload model to Hugging Face", interactive=False)
        output = gr.Markdown(label="Output progress", visible=False)

    demo.load(fn=swap_fill, outputs=[disabled_area, enabled_area], queue=False)
    submit_source_civit.change(fn=check_civit_link, inputs=[submit_source_civit], outputs=[instructions, submit_button_civit, try_again_button, submit_button_civit], queue=False)
    civit_username_to_bulk.change(fn=list_civit_models, inputs=[civit_username_to_bulk], outputs=[submit_bulk_civit], queue=False)
    try_again_button.click(fn=check_civit_link, inputs=[submit_source_civit], outputs=[instructions, submit_button_civit, try_again_button, submit_button_civit], queue=False)
    submit_button_civit.click(fn=show_output, inputs=[], outputs=[output]).then(fn=upload_civit_to_hf, inputs=[submit_source_civit, link_civit], outputs=[output], queue=False)
    bulk_button.click(fn=show_output, inputs=[], outputs=[output]).then(fn=bulk_upload, inputs=[submit_bulk_civit, link_civit], outputs=[output], queue=False)
    #gr.LogoutButton(elem_id="logout")
    
demo.queue()
demo.launch()