Skip to content

vishwas-vs/slideforge

Repository files navigation

slideforge

Agent-authored content, compiled to animated slides and rendered video — from one shared content model.

slideforge lets an LLM agent (with a human steering conversationally, not approving fixed checkpoints) turn a brief, a document, or a topic into a Deck — a small structured content model — and then compile that same Deck two different ways:

  • slideforge slides → an interactive, animated Slidev deck (click-through reveals, live in the browser)
  • slideforge video → a deterministic, rendered HyperFrames video (MP4/WebM, via headless Chrome + FFmpeg)

Slides and video are genuinely different outputs with different constraints (click-driven vs. hard-timed), so slideforge treats them as two explicit modes sharing one content model rather than pretending one derives from the other.

Status

Early-stage. The content model (packages/schema) and its MCP tool surface (packages/mcp-server) are implemented and tested against a real MCP client — an agent can already create a Deck and build it up beat by beat. The compilers (compiler-slidev, compiler-hyperframes) and the slideforge CLI itself are still open issues (see Milestones): nothing here renders a real Slidev deck or HyperFrames video yet. See docs/architecture.md for the intended design and docs/content-model.md for the content model reference.

Why slideforge, not a wrapper around either tool alone

  • Slidev has no video export — it's a live, click-driven presentation tool. Great for interactive decks, not for a rendered video artifact.
  • HyperFrames has no concept of a "deck" or click-through reveals — it's a deterministic, hard-timed HTML→video renderer. Great for video, wrong shape for a live presentation.
  • Both already do the hard part well (markdown parsing + MCP server for Slidev; HTML/GSAP validation + rendering + a 150+ block catalog for HyperFrames). slideforge's job is the shared content model and the two thin adapters — not reimplementing either renderer.

Agent integration

slideforge does not run its own LLM loop or manage model API keys. It exposes an MCP server (content-model CRUD, compile-to-slides, compile-to-video) plus a plain CLI, so any host agent you already use (Claude Code, Cursor, etc.) supplies the "brain" and drives the tools — the same approach Slidev itself takes with its own slidev mcp server. See AGENTS.md for the contributor-facing side of this.

Packages

Package Purpose
packages/schema Shared Deck/Beat/Block content model, validation, patch-based versioning
packages/mcp-server MCP tool surface for host agents (content CRUD, compile, render)
packages/compiler-slidev Deck → Slidev project
packages/compiler-hyperframes Deck → HyperFrames composition
packages/cli The slideforge command-line entry point

Development

Requires Node.js >=22 and pnpm.

pnpm install
pnpm build
pnpm lint
pnpm typecheck
pnpm test

See CONTRIBUTING.md for the full contributor workflow, and docs/ for design docs as they land.

License

MIT

About

Agent-authored content, compiled to animated Slidev decks and rendered HyperFrames video, from one shared content model.

Resources

License

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors