app.py

import mimetypes
import os
from typing import Any
import uuid

import gradio as gr
from dotenv import load_dotenv
from google import genai
from google.genai import types
from gradio.components.multimodal_textbox import MultimodalValue

load_dotenv()

generate_content_config = types.GenerateContentConfig(
    temperature=1,
    top_p=0.95,
    top_k=40,
    max_output_tokens=8192,
    response_modalities=[
        "image",
        "text",
    ],
    response_mime_type="text/plain",
)
model = "gemini-2.0-flash-exp-image-generation"


def save_binary_file(file_name, data):
    f = open(file_name, "wb")
    f.write(data)
    f.close()


def response(
    message: MultimodalValue,
    history: list[dict[str, Any]],
    chat_history_gemini,
    client: genai.Client | None,
):
    if client is None:
        raise gr.Error("Client not initialized. Please enter your API key.")

    files = [client.files.upload(file=f) for f in message.get("files", "")]
    file_parts = [
        types.Part.from_uri(file_uri=f.uri, mime_type=f.mime_type) for f in files
    ]
    user_msg = types.Content(
        role="user",
        parts=file_parts + [types.Part.from_text(text=message.get("text", ""))],
    )
    chat_history_gemini.append(user_msg)
    msg = {"files": [], "text": ""}
    for chunk in client.models.generate_content_stream(
        model=model,
        contents=chat_history_gemini,
        config=generate_content_config,
    ):
        if (
            not chunk.candidates
            or not chunk.candidates[0].content
            or not chunk.candidates[0].content.parts
        ):
            continue
        if chunk.candidates[0].content.parts[0].inline_data:
            file_name = str(uuid.uuid4())
            inline_data = chunk.candidates[0].content.parts[0].inline_data
            file_extension = mimetypes.guess_extension(inline_data.mime_type)
            save_binary_file(f"{file_name}{file_extension}", inline_data.data)
            msg["files"] = [f"{file_name}{file_extension}"]
        else:
            msg["text"] += chunk.text
        yield msg


api_key = gr.Textbox(
    label="Enter your API key", type="password", value=os.getenv("GOOGLE_API_KEY")
)
deep_link = gr.DeepLinkButton()

with gr.Blocks() as demo:
    chat_history_gemini = gr.State([])
    client = gr.State(value=None)
    gr.ChatInterface(
        response,
        title="Gemini Image Generation",
        description="Edit images and share chat with deep links",
        additional_inputs=[chat_history_gemini, client],
        type="messages",
        multimodal=True,
    )
    gr.DeepLinkButton()
    api_key.render()
    api_key.submit(
        lambda api_key: genai.Client(api_key=api_key),
        inputs=api_key,
        outputs=client,
    )
    demo.load(
        lambda: genai.Client(api_key=os.getenv("GOOGLE_API_KEY"))
        if os.getenv("GOOGLE_API_KEY")
        else None,
        inputs=[],
        outputs=[client],
    )


if __name__ == "__main__":
    demo.launch()