dasha2ui 0.1.0

A2UI Protocol Implementation for Dash - Declarative UI generation for LLM agents

DashA2UI

A2UI Protocol Implementation for Dash - Declarative UI generation for LLM agents.

Overview

DashA2UI implements the A2UI (Agent-to-User Interface) Protocol v0.8 for generating declarative, streaming UI payloads that can be rendered by A2UI-compatible clients.

A2UI is designed to be:

• LLM-friendly: Flat adjacency list model, easy for LLMs to generate incrementally
• Secure: Declarative data format, not executable code
• Framework-agnostic: Same payload renders on web, Flutter, etc.

Installation

From GitHub (recommended until published to PyPI)

# Core package (no framework dependencies)
pip install "dasha2ui @ git+https://github.com/Cemberk/dasha2ui.git"

# With Dash renderer support
pip install "dasha2ui[dash] @ git+https://github.com/Cemberk/dasha2ui.git"

# Everything
pip install "dasha2ui[all] @ git+https://github.com/Cemberk/dasha2ui.git"

From Local Clone

git clone https://github.com/Cemberk/dasha2ui.git
cd dasha2ui
pip install -e .[dash]  # Editable install with Dash support

From PyPI (when published)

pip install dasha2ui[dash]
pip install dasha2ui[all]

Quick Start

Building A2UI Surfaces

from dasha2ui import (
    A2UISurface, text, row, column, card,
    TextUsageHint, Distribution
)

# Create a surface
surface = A2UISurface("my-dashboard")

# Add components
title = surface.add(text("Dashboard", usage_hint=TextUsageHint.H1))
subtitle = surface.add(text("Welcome!", usage_hint=TextUsageHint.BODY))

# Create layout
content = surface.add(column([title.id, subtitle.id]))

# Build A2UI messages
messages = surface.build(root_id=content.id)

# Output as JSON
print(surface.to_json_array(root_id=content.id))

Using Pre-built Templates

from dasha2ui import build_dashboard_ui

messages = build_dashboard_ui(
    title="Sales Dashboard",
    metrics=[
        {"title": "Total Sales", "value_path": "/sales/total", "icon": "shoppingCart"},
        {"title": "Orders", "value_path": "/sales/orders", "icon": "payment"},
    ],
    data={"sales": {"total": "$12,345", "orders": "156"}}
)

Rendering in Dash

from dasha2ui.renderers.dash_renderer import A2UIRenderer

# Process A2UI messages
renderer = A2UIRenderer()
renderer.process_messages(messages)

# Get Dash component
dash_component = renderer.render()

# Use in your Dash app
app.layout = html.Div([dash_component])

Components

DashA2UI supports all A2UI v0.8 standard components:

Layout

• row() - Horizontal flex layout
• column() - Vertical flex layout
• list_component() - Scrollable list
• card() - Card container

Content

• text() - Text with typography hints (h1-h5, body, caption)
• image() - Image with sizing hints
• icon() - Material Design icons
• divider() - Horizontal/vertical divider

Interactive

• button() - Button with action
• text_field() - Text input
• checkbox() - Checkbox
• slider() - Slider
• multiple_choice() - Selection
• tabs() - Tab navigation
• modal() - Modal dialog

Media

• video() - Video player
• audio_player() - Audio player

Data Binding

Components can bind to a data model using paths:

from dasha2ui import text, BoundString, A2UISurface

surface = A2UISurface("example")

# Bind text to data model path
value = surface.add(text(BoundString.bound("/metrics/total")))

# Set data
surface.set_data({"metrics": {"total": "1,234"}})

Streaming

For progressive UI generation:

from dasha2ui import A2UIStream, text, column

stream = A2UIStream("dashboard")

# Send initial structure
yield stream.begin_rendering("root")

# Stream components
title = text("Loading...", id="title")
yield stream.surface_update([title])

# Update data progressively
yield stream.data_update({"status": "Ready"})

LLM Integration

A2UI is designed for LLM output. Example prompt:

Generate A2UI JSON to display a dashboard with:
- Title: "Performance Metrics"
- 3 metric cards showing CPU, Memory, Disk usage

Output format: JSON array of A2UI messages

License

MIT License