Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.livepeer.org/llms.txt

Use this file to discover all available pages before exploring further.

Storyboard is a web application at storyboard-rust.vercel.app that lets users generate, edit, animate, and live-stream media through a conversational interface on an infinite canvas. It is built on two packages: @livepeer/agent (provider-agnostic agent runtime) and @livepeer/creative-kit (reusable framework for creative AI apps). Both packages are available for developers building their own creative applications on the same infrastructure.

​
Capabilities

Storyboard exposes 40+ AI capabilities through a single conversation interface. Type a prompt and the agent plans, selects models, runs inference in parallel, and places results as draggable cards on the canvas: 
"Create a 6-scene story about a brave knight, animate the best scene,
 and start a live stream with cyberpunk style"
Generation: All models have automatic fallback chains: if a model rejects a request (content policy, size limit), the next sibling model tries automatically. Slash commands (25+): /story (6-scene illustrated story), /film (4-shot mini-film with camera directions), /stream (real-time LV2V with Scope pipeline), /talk (talking video with voice clone), /analyze (Gemini Vision style/character extraction), /project (manage and replay sessions), and more. Type /help in the app for the full list.

​
Three-Layer Architecture

Your Application
  @livepeer/creative-kit        ArtifactStore · ChatPanel · InfiniteBoard
  @livepeer/agent               AgentRunner · ToolRegistry · WorkingMemory

Livepeer Infrastructure
  SDK Service (sdk.daydream.monster)
  BYOC Orchestrator (go-livepeer, 40 capabilities)
  Scope Orchestrators (LV2V, real-time video)
@livepeer/agent (Layer 0): Provider-agnostic agent runtime. AgentRunner manages the LLM-to-tool loop. Swap LLM providers (Gemini, Claude, OpenAI, Livepeer) without changing application code. Includes working memory (800-token budget) and queryable session memory. @livepeer/creative-kit (Layer 1): Reusable framework for creative AI apps. ArtifactStore manages canvas state, ProjectPipeline handles batch generation, CommandRouter handles slash commands, CapabilityResolver manages model selection and fallback chains. Includes InfiniteBoard, ChatPanel, and ArtifactCard UI components. Applications (Layer 2): Storyboard itself, plus Creative Lab (an educational app for children aged 8-16 in apps/creative-lab/).

​
LLM Provider Configuration

@livepeer/agent supports four LLM backends: The Livepeer provider routes all LLM calls through Livepeer infrastructure with a single Daydream API key. Get one at daydream.live.

​
Building with the SDK

The @livepeer/agent CLI is the fastest way to use Storyboard capabilities programmatically: 
curl -fsSL https://storyboard.daydream.monster/install.sh | bash
livepeer setup --api-key sk_live_xxx
livepeer run flux-schnell --prompt "a mountain at dawn"
For library usage and MCP server mode, see .

Agent SDK

@livepeer/agent library API: AgentRunner, providers, ToolRegistry, and MCP server mode.

Agents Overview

Eliza integration, BYOC agent patterns, and Agent SPE context.

AI Pipelines

The underlying inference endpoints Storyboard routes through.

Real-Time AI

The Cascade/LV2V pipeline powering Storyboard’s /stream command.
Last modified on May 19, 2026