Galvanized Pukeko

AI-powered User Interface (UI) framework

Who loves GPukeko?

• Software developers
• DIY enthusiasts

When use GPukeko?

• Building Data-Heavy Dashboards Fast When you need to display complex datasets—analytics, reports, or metrics — without hand-building a visualization for every payload.
• Creating Dynamic User Workflows When your application needs to adapt its interface based on user actions, serving different forms, tables, or controls as users navigate through multi-step processes.
• Rapid Prototyping When you're validating product ideas or building MVPs and want functional, on-brand interfaces without hand-building a component for every interaction.
• Reducing UI Maintenance Overhead When your team is tired of maintaining dozens of similar components (forms, tables, cards) and wants the agent to render the common ones on demand.
• Scaling Products Across Multiple Contexts When you're building white-label solutions or multi-tenant applications that need consistent functionality but different visual treatments for each client or brand.

What Galvanized Pukeko Does

Core Functions

• 1. Agent-Rendered UI
  • Your agent calls a tool to render a component; the client draws it and streams it in live
  • Collect input with forms, present data with charts and tables, or build richer interactive surfaces
  • User interactions are sent back to the agent so the conversation can continue
• 2. The Agent Picks the Component
  • The LLM decides whether a task needs a form, chart, table or an interactive surface, and calls the matching tool
  • You expose the tools; the model chooses how and when to use them
• 3. Design System Integration
  • Built-in components follow the Pukeko design tokens
  • Maintains visual consistency across your application
  • Restyle with your own tokens, or supply your own components
• 4. Open Protocol
  • Talks to any AG-UI server over HTTP/SSE
  • Use the Gaunt Sloth CLI, the ADK (Java) agent, or your own AG-UI backend

Key Features

• Built on Vue 3
  • Drop in the all-in-one app component, or compose the chat panel and widgets yourself
  • Use the built-in components, or supply your own Vue components
• Streaming Responses
  • Text, reasoning and tool calls stream in as the agent works
  • Input from rendered components flows back to the agent
• Developer-Friendly
  • Minimal setup — install, point at an AG-UI server, mount
  • Spend your time on the agent and its tools, not on UI scaffolding
• The Components
  • Forms, tables, bar and pie charts, and interactive A2UI surfaces
  • Plus a webcam panel for vision and robotics clients

Why GPukeko?

Chatbots that only return text are limiting.

When your AI assistant needs the user to fill in a form, compare numbers in a chart, or pick a row from a table, plain text falls short — and hand-building a bespoke UI for every interaction is slow.

Galvanized Pukeko gives the agent real UI.

Expose a set of UI tools to your agent and it can render a form, a chart, a table or an interactive surface right in the conversation, then read back what the user does with it. The components are Vue 3 and follow the Pukeko design tokens out of the box — restyle them or bring your own. It works with the Gaunt Sloth CLI, the ADK (Java) agent, or any AG-UI server.

To make GPukeko work,

you need an API key from some AI provider, such as:

• OpenRouter
• Groq
• DeepSeek
• Google AI Studio and Google Vertex AI
• Anthropic
• OpenAI (and other providers using OpenAI format, such as Inception)
• Ollama with JS config (some of the models, see #107 )
• xAI

* Any other provider supported by LangChain.JS should also work with JS config .

The promise of Galvanized Pukeko:

• Less Repetitive UI Code Instead of hand-building a form, table or chart for every interaction, give the agent the tools and let it render them.
• Your Design Built-in components follow the Pukeko design tokens. Restyle them, or bring your own Vue components.
• Interactive, Not Just Text The agent can ask for input, show data, and react to what the user does — inside the conversation.
• Works With Your Stack Pair it with the Gaunt Sloth CLI, the ADK (Java) agent, or any AG-UI server.
• Open Source Built on the open AG-UI and A2UI protocols — no lock-in.