Streamlit Chat Prompt

A Streamlit component that provides a modern chat-style prompt with image attachment and paste support. This component was built to mimic the style of streamlit.chat_input while expanding functionality with images. Future work may include addition of speech-to-text input.

Author: Tyler House (@tahouse)

Features

• 📝 Chat-style text input with multi-line support
• 📎 Image attachment support via button or drag-and-drop
• 📋 Paste image support (paste images directly from clipboard)
• 🖼️ Image preview with ability to remove attached images
• ⌨️ Submit with Enter key (Shift+Enter for new line)
• 🎨 Automatic theme integration with Streamlit
• 📱 Responsive design that works well on mobile and desktop
• 🗜️ Automatic image compression/scaling to stay under size limits (customizable, default 5MB)
• 📌 Optional pinned-to-bottom placement for main chat interface (one per app)
• 🔄 Flexible positioning for use in dialogs, sidebars, or anywhere in the app flow
• ✏️ Support for default/editable content - perfect for message editing workflows
• 🔤 Smart focus management - automatically returns to text input after interactions

Installation

pip install streamlit-chat-prompt

Usage

import streamlit as st
from streamlit_chat_prompt import prompt

# Create a chat prompt
response = prompt(
    name="chat",  # Unique name for the prompt
    key="chat",   # Unique key for the component instance
    placeholder="Hi there! What should we talk about?",  # Optional placeholder text
    main_bottom=True,  # Pin prompt to bottom of main area
    max_image_size=5 * 1024 * 1024,  # Maximum image size (5MB default)
    disabled=False,  # Optionally disable the prompt
)

# Handle the response
if response:
    if response.text:
        st.write(f"Message: {response.text}")
    
    if response.images:
        for i, img in enumerate(response.images):
            st.write(f"Image {i+1}: {img.type} ({img.format})")

Examples

Here are some usage patterns, or check out rocktalk for a full working example.

1. Main Chat Interface

   import base64
   from io import BytesIO
   import streamlit as st
   from streamlit_chat_prompt import PromptReturn, prompt, ImageData
   from PIL import Image
   
   
   st.chat_message("assistant").write("Hi there! What should we chat about?")
   
   prompt_return: PromptReturn | None = prompt(
       name="foo",
       key="chat_prompt",
       placeholder="Hi there! What should we chat about?",
       main_bottom=True,
   )
   
   if prompt_return:
       with st.chat_message("user"):
           st.write(prompt_return.text)
           if prompt_return.images:
               for image in prompt_return.images:
                   st.divider()
                   image_data: bytes = base64.b64decode(image.data)
                   st.markdown("Using `st.image`")
                   st.image(Image.open(BytesIO(image_data)))
   
                   # or use markdown
                   st.divider()
                   st.markdown("Using `st.markdown`")
                   st.markdown(f"![Image example](data:image/png;base64,{image.data})")

2. Dialog Usage and Starting From Existing Message

   if st.button(
       "Dialog Prompt with Default Value", key=f"dialog_prompt_with_default_button"
   ):
       with open("example_images/vangogh.png", "rb") as f:
           image_data = f.read()
           image = Image.open(BytesIO(image_data))
           base64_image = base64.b64encode(image_data).decode("utf-8")
           test_dg(
               default_input=PromptReturn(
                   text="This is a test message with an image",
                   images=[
                       ImageData(data=base64_image, type="image/png", format="base64")
                   ],
               ),
               key="dialog_with_default",
           )

Component API

prompt()

Main function to create a chat prompt.

Parameters:

• name (str): Unique name for this prompt instance
• key (str): Unique key for the component instance
• placeholder (str, optional): Placeholder text shown in input field
• default (Union[str, PromptReturn], optional): Default value for the prompt. Can include text and images using the PromptReturn object type.
• main_bottom (bool, optional): Pin prompt to bottom of main area (default: True)
• max_image_size (int, optional): Maximum image size in bytes (default: 5MB)
• disabled (bool, optional): Disable the prompt (default: False)

Returns:

Optional[PromptReturn]: Object containing message and images if submitted, None otherwise

PromptReturn

Object returned when user submits the prompt.

Properties:

• text (Optional[str]): Text message entered by user
• images (Optional[List[ImageData]]): List of attached images

ImageData

Object representing an attached image.

Properties:

• type (str): Image MIME type (e.g. "image/jpeg")
• format (str): Image format (e.g. "base64")
• data (str): Image data as base64 string

Development

This repository is based on the Streamlit Component template system. If you want to modify or develop the component:

1. Clone the repository

2. Install development dependencies:

   pip install -e ".[devel]"

3. Start the frontend development server:

   cd streamlit_chat_prompt/frontend
   npm install
   npm run start

4. In a separate terminal, run your Streamlit app:

   streamlit run your_app.py

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.