stable-diffusion-telegram-bot/main.py

import os
import re
import io
import uuid
import base64
import requests
from datetime import datetime
from PIL import Image, PngImagePlugin
from pyrogram import Client, filters
from pyrogram.types import InlineKeyboardButton, InlineKeyboardMarkup
from dotenv import load_dotenv

# Load environment variables
load_dotenv()
API_ID = os.environ.get("API_ID")
API_HASH = os.environ.get("API_HASH")
TOKEN = os.environ.get("TOKEN_givemtxt2img")
SD_URL = os.environ.get("SD_URL")

app = Client("stable", api_id=API_ID, api_hash=API_HASH, bot_token=TOKEN)
IMAGE_PATH = 'images'

# Ensure IMAGE_PATH directory exists
os.makedirs(IMAGE_PATH, exist_ok=True)

# Model-specific emmbedings for negative prompts
# see civit.ai model page for specific emmbedings recommnded for each model
model_negative_prompts = {
    "Anything-Diffusion": "",
    "Deliberate": "",
    "Dreamshaper": "",
    "DreamShaperXL_Lightning": "",
    "icbinp": "",
    "realisticVisionV60B1_v51VAE": "realisticvision-negative-embedding",
    "v1-5-pruned-emaonly": ""
}


def encode_file_to_base64(path):
    with open(path, 'rb') as file:
        return base64.b64encode(file.read()).decode('utf-8')


def decode_and_save_base64(base64_str, save_path):
    with open(save_path, "wb") as file:
        file.write(base64.b64decode(base64_str))



    # Set default payload values
default_payload = {
        "prompt": "",
        "seed": -1, # Random seed
        "negative_prompt":  "extra fingers, mutated hands, poorly drawn hands, poorly drawn face, deformed, ugly, blurry, bad anatomy, bad proportions, extra limbs, cloned face, skinny, glitchy, double torso, extra arms, extra hands, mangled fingers, missing lips, ugly face, distorted face, extra legs",
        "enable_hr": False,
        "Sampler": "DPM++ SDE Karras",
        "denoising_strength": 0.35,
        "batch_size": 1,
        "n_iter": 1,
        "steps": 35,
        "cfg_scale": 7,
        "width": 512,
        "height": 512,
        "restore_faces": False,
        "override_settings": {},
        "override_settings_restore_afterwards": True,
    }

def update_negative_prompt(model_name):
    if model_name in model_negative_prompts:
        suffix = model_negative_prompts[model_name]
        default_payload["negative_prompt"] += f", {suffix}"

def parse_input(input_string):
    payload = default_payload.copy()
    prompt = []

    matches = re.finditer(r"(\w+):", input_string)
    last_index = 0

    script_args = [0, "", [], 0, "", [], 0, "", [], True, False, False, False, False, False, False, 0, False]
    script_name = None


    script_args = [0, "", [], 0, "", [], 0, "", [], True, False, False, False, False, False, False, 0, False]
    script_name = None

    slot_mapping = {0: (0, 1), 1: (3, 4), 2: (6, 7)}
    slot_index = 0

    for match in matches:
        key = match.group(1).lower()
        value_start_index = match.end()
        if last_index != match.start():
            prompt.append(input_string[last_index: match.start()].strip())
        last_index = value_start_index
        value_end_match = re.search(r"(?=\s+\w+:|$)", input_string[value_start_index:])
        if value_end_match:
            value_end_index = value_end_match.start() + value_start_index
        else:
            value_end_index = len(input_string)
        value = input_string[value_start_index: value_end_index].strip()
        if key == "ds":
            key = "denoising_strength"
        if key == "ng":
            key = "negative_prompt"
        if key == "cfg":
            key = "cfg_scale"

        if key in default_payload:
            payload[key] = value
        elif key in ["xsr", "xsteps", "xds", "xcfg", "nl", "ks", "rs"]:
            script_name = "x/y/z plot"
            if slot_index < 3:
                script_slot = slot_mapping[slot_index]
                if key == "xsr":
                    script_args[script_slot[0]] = 7  # Enum value for xsr
                    script_args[script_slot[1]] = value
                elif key == "xsteps":
                    script_args[script_slot[0]] = 4  # Enum value for xsteps
                    script_args[script_slot[1]] = value
                elif key == "xds":
                    script_args[script_slot[0]] = 22  # Enum value for xds
                    script_args[script_slot[1]] = value
                elif key == "xcfg":
                    script_args[script_slot[0]] = 6  # Enum value for CFG Scale
                    script_args[script_slot[1]] = value
                slot_index += 1
            elif key == "nl":
                script_args[9] = False  # Draw legend
            elif key == "ks":
                script_args[10] = True  # Keep sub images
            elif key == "rs":
                script_args[11] = True  # Set random seed to sub images
        else:
            prompt.append(f"{key}:{value}")

        last_index = value_end_index

    payload["prompt"] = " ".join(prompt).strip()
    if not payload["prompt"]:
        payload["prompt"] = input_string.strip()

    if script_name:
        payload["script_name"] = script_name
        payload["script_args"] = script_args

    return payload


def create_caption(payload, user_name, user_id, info):
    caption = f"**[{user_name}](tg://user?id={user_id})**\n\n"
    prompt = payload["prompt"]
    print(payload["prompt"])
    print(info)
    # Steps: 3, Sampler: Euler, CFG scale: 7.0, Seed: 4094161400, Size: 512x512, Model hash: 15012c538f, Model: realisticVisionV60B1_v51VAE, Denoising strength: 0.35, Version: v1.8.0-1-g20cdc7c

    # Define a regular expression pattern to match the seed value
    seed_pattern = r"Seed: (\d+)"

    # Search for the pattern in the info string
    match = re.search(seed_pattern, info)

    # Check if a match was found and extract the seed value
    if match:
        seed_value = match.group(1)
        print(f"Seed value: {seed_value}")
        caption += f"**{seed_value}**\n"
    else:
        print("Seed value not found in the info string.")

    caption += f"**{prompt}**\n"

    if len(caption) > 1024:
        caption = caption[:1021] + "..."

    return caption


def call_api(api_endpoint, payload):
    try:
        response = requests.post(f'{SD_URL}/{api_endpoint}', json=payload)
        response.raise_for_status()
        return response.json()
    except requests.RequestException as e:
        print(f"API call failed: {e}")
        return None


def process_images(images, user_id, user_name):
    def generate_unique_name():
        unique_id = str(uuid.uuid4())[:7]
        return f"{user_name}-{unique_id}"

    word = generate_unique_name()

    for i in images:
        image = Image.open(io.BytesIO(base64.b64decode(i.split(",", 1)[0])))
        png_payload = {"image": "data:image/png;base64," + i}
        response2 = requests.post(f"{SD_URL}/sdapi/v1/png-info", json=png_payload)
        response2.raise_for_status()

        pnginfo = PngImagePlugin.PngInfo()
        pnginfo.add_text("parameters", response2.json().get("info"))
        image.save(f"{IMAGE_PATH}/{word}.png", pnginfo=pnginfo)

        return word, response2.json().get("info")

@app.on_message(filters.command(["draw"]))
def draw(client, message):
    msgs = message.text.split(" ", 1)
    if len(msgs) == 1:
        message.reply_text("Format :\n/draw < text to image >\nng: < negative (optional) >\nsteps: < steps value (1-70, optional) >")
        return

    payload = parse_input(msgs[1])
    print(payload)

    # Check if xds is used in the payload
    if "xds" in msgs[1].lower():
        message.reply_text("`xds` key cannot be used in the `/draw` command. Use `/img` instead.")
        return

    K = message.reply_text("Please Wait 10-15 Seconds")
    r = call_api('sdapi/v1/txt2img', payload)

    if r:
        for i in r["images"]:
            word, info = process_images([i], message.from_user.id, message.from_user.first_name)
            caption = create_caption(payload, message.from_user.first_name, message.from_user.id, info)
            message.reply_photo(photo=f"{IMAGE_PATH}/{word}.png", caption=caption)
        K.delete()
    else:
        message.reply_text("Failed to generate image. Please try again later.")
        K.delete()


@app.on_message(filters.command(["img"]))
def img2img(client, message):
    if not message.reply_to_message or not message.reply_to_message.photo:
        message.reply_text("Reply to an image with\n`/img < prompt > ds:0-1.0`\n\nds stands for `Denoising_strength` parameter. Set that low (like 0.2) if you just want to slightly change things. defaults to 0.35\n\nExample: `/img murder on the dance floor ds:0.2`")
        return

    msgs = message.text.split(" ", 1)
    if len(msgs) == 1:
        message.reply_text("dont FAIL in life")
        return

    payload = parse_input(msgs[1])
    print(f"input:\n{payload}")
    photo = message.reply_to_message.photo
    # prompt_from_reply = message.reply_to_message.
    # orginal_prompt = app.reply_to_message.message
    # print(orginal_prompt)
    photo_file = app.download_media(photo)
    init_image = encode_file_to_base64(photo_file)
    os.remove(photo_file)  # Clean up downloaded image file

    payload["init_images"] = [init_image]

    K = message.reply_text("Please Wait 10-15 Seconds")
    r = call_api('sdapi/v1/img2img', payload)

    if r:
        for i in r["images"]:
            word, info = process_images([i], message.from_user.id, message.from_user.first_name)
            caption = create_caption(payload, message.from_user.first_name, message.from_user.id, info)
            message.reply_photo(photo=f"{IMAGE_PATH}/{word}.png", caption=caption)
        K.delete()
    else:
        message.reply_text("Failed to process image. Please try again later.")
        K.delete()


@app.on_message(filters.command(["getmodels"]))
async def get_models(client, message):
    try:
        response = requests.get(f"{SD_URL}/sdapi/v1/sd-models")
        response.raise_for_status()
        models_json = response.json()
        print(models_json)
        buttons = [
            [InlineKeyboardButton(model["title"], callback_data=model["model_name"])]
            for model in models_json
        ]
        await message.reply_text("Select a model [checkpoint] to use", reply_markup=InlineKeyboardMarkup(buttons))
    except requests.RequestException as e:
        await message.reply_text(f"Failed to get models: {e}")


@app.on_callback_query()
async def process_callback(client, callback_query):
    sd_model_checkpoint = callback_query.data
    options = {"sd_model_checkpoint": sd_model_checkpoint}

    try:
        response = requests.post(f"{SD_URL}/sdapi/v1/options", json=options)
        response.raise_for_status()

        # Update the negative prompt based on the selected model
        update_negative_prompt(sd_model_checkpoint)
        
        await callback_query.message.reply_text(f"Checkpoint set to {sd_model_checkpoint}")
    except requests.RequestException as e:
        await callback_query.message.reply_text(f"Failed to set checkpoint: {e}")
        print(f"Error setting checkpoint: {e}")


@app.on_message(filters.command(["info_sd_bot"]))
async def info(client, message):
    await message.reply_text("""
now support for xyz scripts, see [sd wiki](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#xyz-plot) !
currently supported 
`xsr` - search replace text/emoji in the prompt, more info [here](https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#prompt-sr) 
`xds` - denoise strength, only valid for img2img
`xsteps` - steps
**note** limit the overall `steps:` to lower value (10-20) for big xyz plots

aside from that you can use the usual `ng`, `ds`, `cfg`, `steps` for single image generation.
""", disable_web_page_preview=True)


app.run()