Introduction

Gemini Nano Banana project demonstrate how to quickly build and deploy a chatbot powered with Gemini to generate your own personalized content using the latest model from Gemini.

This chatbot allows you to load an image and instruct Gemini 2.5 Flash Image (Gemini Nano Banana) model (gemini-2.5-flash-image-preview) to modify the image.

The following code excerpt shows how we generate the multimodal output (Text, Image) from the multimodal input (same modalities):

    contents = ([system_prompt,input_text], st.session_state.image)
    
    response = client.models.generate_content(
        model="gemini-2.5-flash-image-preview",
        contents=contents,
        config=types.GenerateContentConfig(
        response_modalities=['Text', 'Image']
        )
    )

The user interface is developed using Streamlit.

The application is deployed with App Engine.

Deployment using Google Cloud SDK

To init the Google Cloud SDK, run the command:

~/google-cloud-sdk/bin/gcloud init

See the list of projects with:

~/google-cloud-sdk/bin/gcloud projects list

Check what project is set:

~/google-cloud-sdk/bin/gcloud config get-value project

To set the current project, run:

~/google-cloud-sdk/bin/gcloud config set project <PROJECT_NUMBER>

To deploy the application, run the command:

~/google-cloud-sdk/bin/gcloud app deploy app.yaml --project <PROJECT_ID>

Start app

To start the application, run from the root:

streamlit run --server.port 8080 --server.enableCORS false --server.enableXsrfProtection=false --server.enableWebsocketCompression=false app.py

On local, if you want to test the app, just run:

streamlit run app.py