OpenRAL — website

The public landing page for OpenRAL, the open-source harness for physical AI: a typed orchestration layer over ROS 2 that turns vision-language-action models, perception and reasoning into safe, runnable robot behavior.

→ Project repo: github.com/OpenRAL/openral

Stack

• Vite + React + Framer Motion — component-based single page under src/, built to static assets. Refined-dark theme with a single muted-clay accent. Fonts: Clash Display + Satoshi (Fontshare) and JetBrains Mono (Google Fonts).
• One serverless function — api/contact.js (Vercel Node runtime) delivers the contact form to the right inbox via Resend.

Sections

Hero (floating L0–L7 layer stack) → terminal install (curl installer) → what it solves → architecture diagram (Helix-style dual-system S1 ⇄ S2 flow) → capabilities → rSkills → showcase (autoscrolling benchmark / simulation / deployment clips) → contact → footer (GitHub · Hugging Face · Discord).

Showcase videos

The "See it run" section is an autoscrolling strip of benchmark, simulation and deployment clips. Videos are not committed to this repo — they're hosted on the public OpenRAL/website-media Hugging Face dataset (CDN-backed) and listed in src/videos/manifest.json.

To make a video show up on the site:

# 1. Drop the raw clip (any size/format) into the gitignored media/ folder.
#    The sub-folder picks the tab; the filename fills the on-clip labels:
#    <benchmark>_<rskill>_<success|fail>.<ext>
media/benchmarks/libero-spatial_pi05_success.mp4
media/simulation/warehouse-pick_rtdetr-v2_fail.mp4
media/deployment/franka-stack_qwen35_success.mp4

# 2. Encode (poster + square preview + full clip), upload to HF, refresh manifest:
npm run media

# 3. Commit the updated src/videos/manifest.json — that's the only tracked change.
git add src/videos/manifest.json && git commit -m "Add showcase clip"

Use hyphens inside a name (libero-spatial); underscores only separate the three parts, and the last must be success or fail. Adding clips needs ffmpeg and the hf CLI signed in with write access to the OpenRAL org (hf auth whoami); npm run media -- --no-upload encodes + updates the manifest locally without publishing. Full reference: src/videos/README.md.

Local preview

npm install
npm run dev      # Vite dev server on http://localhost:5173
npm run build    # production build → dist/
npm run preview  # serve the production build

To exercise the /api/contact function locally, use the Vercel CLI (npm i -g vercel then vercel dev) — it serves the Vite build and the serverless function together.

Deploy (Vercel)

1. Import this repo in Vercel (Framework preset: Vite — it builds dist/ and serves the api/ function alongside).
2. Add two environment variables (Project → Settings → Environment Variables):

   Variable	Value
   RESEND_API_KEY	An API key from your Resend account
   CONTACT_FROM	A verified sender on the domain, e.g. OpenRAL <noreply@openral.com>

3. In Resend, verify the openral.com domain so the function can send from it.
4. Map the custom domain (openral.com) in Vercel → Settings → Domains.

Contact form behavior

The form posts { audience, name, email, message } to /api/contact. audience routes the message to hello@openral.com (both the general and partnership topics currently land there; the subject line distinguishes them); the visitor's address is set as reply_to.

Until RESEND_API_KEY and CONTACT_FROM are set, the endpoint returns 503 { code: "not_configured" } and the form tells the visitor to email us directly — it never silently drops a message.

License

Apache-2.0.