Arcade-style agent activity display for the Divoom Pixoo-64 LED panel.
Shows real-time AI agent status: what it's thinking, which tools it's using, context usage, token counts, and a scrolling activity pulse.
- Activity Icons — Thinking, Tool, Web, Cron, Dev — light up in neon colors when active
- Stats Bars — Context usage (CTX), total tokens (TOK), model name (LLM), version (VER)
- Scrolling Pulse — Activity history bar chart that grows with agent work and decays when idle
- Live Monitoring — Watches OpenClaw session logs and updates the display in real-time
- Pixel Art Renderer — Custom 3x5 and 4x5 bitmap fonts, 8x8 icons, all rendered in code
- Python 3.8+
- Divoom Pixoo-64 on the same network
curl(for HTTP communication with the Pixoo)nmap(optional, for device discovery)- No pip dependencies — uses only Python stdlib
# Auto-detect Pixoo on your network
./scripts/find-pixoo.sh
# Or specify a subnet
./scripts/find-pixoo.sh 192.168.1.0/24The script scans your local network for Divoom devices and reports their IP, current channel, and brightness.
python3 pixoo_display.py 192.168.178.190 RUNE "#FF3030"python3 activity_monitor.py --name RUNE --color "#FF3030" --ip 192.168.178.190python3 activity_monitor.py --demoThe display shows five activity indicators. Each lights up in a distinct neon color when the corresponding activity is detected in OpenClaw's session logs:
| Icon | Label | Color | Lights up when... |
|---|---|---|---|
| 🧠 | THK | Cyan | The LLM is generating a response (thinking/reasoning) |
| 🔧 | TL | Yellow | A tool is being executed (exec, Read, Write, Edit, process) |
| 🌐 | WEB | Green | Web tools are active (web_search, web_fetch, browser) |
| ⏰ | CRN | Magenta | A heartbeat/cron job is running (scheduled tasks, reminders) |
| 💻 | DEV | Orange | Code is being written or edited (file writes to .py, .ts, .js, .yml, etc.) |
Icons dim to ~15% brightness when inactive and glow with a border effect when active. Multiple icons can be lit simultaneously (e.g., thinking + tool + web during a complex operation).
The monitor watches OpenClaw session log files (.jsonl) for patterns:
- Thinking: Model API calls, assistant responses
- Tool:
exec,Read,Write,Edit,processtool invocations - Web:
web_search,web_fetch,browsertool calls - Cron: Heartbeat messages, scheduled reminders, cron-triggered sessions
- Dev: File writes to code files,
git commit,git push,npm run, build commands
Each activity has a 5-second decay — the icon stays lit for 5 seconds after the last matching log entry, then fades out.
| Row | Label | Shows |
|---|---|---|
| 1 | CTX | Context window usage as percentage (green → yellow → orange → red as it fills) |
| 2 | TOK | Total tokens used this session in thousands (e.g., "130K") |
| 3 | LLM | Current model name (e.g., "opus-4-6") |
| 4 | VER | Software version |
Stats are refreshed from openclaw status --json every 30 seconds (configurable with --stats-interval).
The bottom row shows a scrolling bar chart of recent activity intensity:
- Bar height = number of simultaneously active indicators (more activity = taller bars)
- Bars decay over time with randomized fade-out
- Scrolls left-to-right as new activity is recorded
- Shows a dim baseline when idle
--ip IP Pixoo IP address (default: 192.168.178.190)
--name NAME Agent name displayed in title bar
--color HEX Agent color as hex (e.g. "#FF3030" for red)
--interval SECS Update interval in seconds (default: 1.0)
--brightness 0-100 Display brightness (default: 80)
--stats-interval S How often to refresh stats from OpenClaw (default: 30)
--log-dir PATH OpenClaw session log directory
--demo Run in demo mode (cycles through activity states)
Run one monitor per agent host, each with a different name and color:
# On Rune (red)
python3 activity_monitor.py --name RUNE --color "#FF3030"
# On Colossus (green)
python3 activity_monitor.py --name COLOSSUS --color "#30FF30"Install as a user service so it starts automatically on login:
# Copy the service file
cp openclaw-pixoo.service ~/.config/systemd/user/
# Edit it to match your setup (agent name, color, Pixoo IP)
nano ~/.config/systemd/user/openclaw-pixoo.service
# Enable and start
systemctl --user daemon-reload
systemctl --user enable openclaw-pixoo.service
systemctl --user start openclaw-pixoo.service
# Check status
systemctl --user status openclaw-pixoo.service
# View logs
journalctl --user -u openclaw-pixoo.service -fTo keep the service running even when you're not logged in:
sudo loginctl enable-linger $USEREdit ~/.config/systemd/user/openclaw-pixoo.service and adjust the ExecStart line:
ExecStart=/usr/bin/python3 -u %h/path/to/activity_monitor.py \
--name YOUR_AGENT \
--color "#FF3030" \
--ip YOUR_PIXOO_IP \
--brightness 80┌────────────────────────────────────┐
│ RUNE (title) │ y=0-8 Title bar (agent color)
├╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┤ y=9 Separator
│ 🧠 🔧 🌐 ⏰ 💻 (icons) │ y=12-19 Activity icons (8×8 px)
│ THK TL WEB CRN DEV (labels) │ y=21-25 Labels (3×5 font)
├╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┤ y=27 Separator
│ CTX [████████░░░░] 72% │ y=29 Context usage
│ TOK [██████░░░░░░] 130K │ y=34 Token count
│ LLM opus-4-6 │ y=39 Model name
│ VER 2026.3.2 │ y=45 Version
│ │
│ ▁▃▅▇▅▃▁ ▁▃▅▃▁ (pulse) │ y=56 Activity pulse
└────────────────────────────────────┘
If you're an AI agent setting this up, here's how to find the Pixoo on the network:
# Option 1: Use the discovery script
./scripts/find-pixoo.sh
# Option 2: Manual probe
# The Pixoo runs an HTTP API on port 80. Scan and probe:
for ip in $(nmap -sn 192.168.1.0/24 -T4 | grep -oP '\d+\.\d+\.\d+\.\d+'); do
response=$(curl -s -m 1 -X POST "http://$ip:80/post" \
-H 'Content-Type: application/json' \
-d '{"Command":"Channel/GetIndex"}' 2>/dev/null)
if echo "$response" | grep -q '"error_code"'; then
echo "Pixoo found at: $ip"
fi
done
# Option 3: Check the Divoom app on a phone connected to the same networkVerification: Once you have the IP, confirm it's a Pixoo:
# Should return: Hello World divoom!
curl -s http://PIXOO_IP:80/get
# Should return JSON with error_code: 0
curl -s -X POST http://PIXOO_IP:80/post \
-H 'Content-Type: application/json' \
-d '{"Command":"Channel/GetAllConf"}'Critical: Sequential PicIDs required. The Pixoo silently ignores frames with non-sequential PicIDs. Always query Draw/GetHttpGifId first, then send with current_id + 1.
# ✅ Correct
current = query("Draw/GetHttpGifId")["PicId"]
send(PicID=current + 1)
# ❌ Wrong — frame accepted but silently not displayed
send(PicID=1)
send(PicID=random())Other gotchas:
Draw/ResetHttpGifIdcrashes the Pixoo firmware — never use it- Channel 3 is the custom/API draw channel (
Channel/SetIndexwithSelectIndex: 3) - Use
curlsubprocess for HTTP — Python'surllibsilently fails on large payloads - The Pixoo persists the last frame to flash (survives reboots)
- Re-assert
Channel/SetIndexperiodically (~30s) if the Pixoo auto-reverts to its clock face - All API responses return
{"error_code": 0}even when the frame is silently dropped
MIT