Kebivo demo source

Motion Canvas project behind ../demo.gif, the website's demo.mp4 and demo-poster.png, and ../social-preview.png (rendered from ../social-preview.html). Every product frame is a screenshot of a real ghcr.io/ocnai/kebivo:0.1.1 container filled with synthetic data (example.org, no real names). Only the chat in scene 1 and the end card are drawn.

Scene	What it shows
1	A generic chat: the assistant can't find the answer
2	The agent's report_knowledge_gap call over MCP and the real status: accepted response
3	The routing rule and the delivery row marked delivered
4	Repeats join the same inbox entry; the occurrence count goes 1 to 4
5	The owner's note and the resolved status, then the end card

Re-render

cd docs/assets/demo-src
npm install
npm run serve          # editor at http://localhost:9000

In the editor, open Video Settings and press Render. It writes PNG frames (1280x720, 30 fps) to output/project/ (ignored by git). Then encode the shipped files:

bash scripts/encode.sh   # docs/assets/demo.gif, plus output/demo.mp4 and output/demo-poster.png for the website

Re-capture the screenshots

docker run -d --name kdemo -p 127.0.0.1:18321:8000 -v kdemo-data:/data \
  -e SECRET_KEY=$(openssl rand -hex 32) -e ENCRYPTION_KEY=$(openssl rand -hex 32) \
  -e ALLOWED_ORIGINS=http://127.0.0.1:18321 ghcr.io/ocnai/kebivo:0.1.1
printf 'DemoPassw0rd-2026\nDemoPassw0rd-2026\n' | docker exec -i kdemo python -m kebivo.cli bootstrap --email admin@example.org --password-stdin
node capture/capture.mjs   # needs Playwright (frontend/node_modules) and a Chromium; set CHROMIUM=/path if needed

capture.mjs seeds one project, a webhook destination and a routing rule, sends the MCP call, and writes shots/*.png. It starts the receiver in capture/receiver.py inside the container, so use a fresh container each time.