11-novnc_lan_flow - noVNC computers controlled by URI flow

Docker demo where URI commands control real Chromium browsers running inside virtual noVNC desktops. The browser opens inside the container, not on the host.

What runs

• pc1-browser..pc4-browser - Selenium Chromium desktops with noVNC.
• pc1-api..pc4-api - small URI node APIs exposing browser://, app:// and log://.
• dashboard - browser view with noVNC iframes, logs and last flow result.
• orchestrator/run_flow.py - sends a multi-step URI flow to the computers.

Optional pc3 and pc4 services exist behind the Compose profile full.

Run

make up

Open:

Dashboard: http://127.0.0.1:8192/?pc1NovncPort=7901&pc2NovncPort=7902&pc1ApiPort=9001&pc2ApiPort=9002
pc1 noVNC: http://127.0.0.1:7901/?autoconnect=1&resize=scale
pc2 noVNC: http://127.0.0.1:7902/?autoconnect=1&resize=scale

If these ports are busy, run the same demo on another set of ports. The dashboard URL accepts port overrides in the query string:

DASHBOARD_PORT=18192 \
PC1_NOVNC_PORT=17901 PC2_NOVNC_PORT=17902 \
PC1_API_PORT=19001 PC2_API_PORT=19002 \
make up

Run the URI flow:

make flow

The basic flow opens pages and captures screenshots through these URI commands:

browser://pc1/page/command/open
browser://pc1/page/command/screenshot
browser://pc2/page/command/open
browser://pc2/page/command/screenshot
log://pc1/session/command/write
log://pc2/session/query/recent

Four-computer flow

The full scenario starts four noVNC computers and gives every computer a small local service:

make up-full
make flow-full

The dashboard URL printed by make up-full includes the query parameters needed to show all four iframes:

?pcs=pc1,pc2,pc3,pc4&pc1NovncPort=7901&pc2NovncPort=7902&pc3NovncPort=7903&pc4NovncPort=7904

The full flow uses these extra application routes:

app://pc1/notes/command/add
app://pc1/notes/query/list
app://pc2/orders/command/create
app://pc2/orders/query/list
app://pc3/reports/command/render
app://pc3/reports/query/latest
app://pc4/monitor/command/check
app://pc4/monitor/query/status

It also drives browser screenshots on all four machines:

browser://pc1/page/command/screenshot
browser://pc2/page/command/screenshot
browser://pc3/page/command/screenshot
browser://pc4/page/command/screenshot

Artifacts are written to:

generated/flow-result.json
generated/registry.json
generated/routes.txt
generated/screenshots/*.png

Manual URI call

curl -fsS http://127.0.0.1:9001/run \
  -H 'content-type: application/json' \
  -d '{"uri":"browser://pc1/page/command/open","payload":{"url":"https://example.com/"}}' \
  | python3 -m json.tool

Screenshot:

curl -fsS http://127.0.0.1:9001/run \
  -H 'content-type: application/json' \
  -d '{"uri":"browser://pc1/page/command/screenshot","payload":{"url":"https://example.com/","output":"manual.png"}}' \
  | python3 -m json.tool

Application command:

curl -fsS http://127.0.0.1:9001/run \
  -H 'content-type: application/json' \
  -d '{"uri":"app://pc1/notes/command/add","payload":{"text":"Manual note from shell"}}' \
  | python3 -m json.tool

Test

make test
make test-full

The smoke tests start Docker, run the URI flow, verify generated routes and check that screenshots were produced. make test-full also checks the app:// routes and the four-computer screenshot set.

> Needs Docker and pulls the public selenium/standalone-chromium image.

Files

View on GitHub →