ClawWatch 2.0 is here

[ThinkOffApp]

ClawWatch 2.0 is out. Here is what changed since 1.0:

Vector avatars. Six hand-drawn characters (ant, lobster, orange lobster, robot, boy, girl) replace the old static screens. Swipe to switch. The avatar system is built to react to real context, not just loop animations.

Live sensor access. The agent reads Galaxy Watch hardware directly: heart rate, SpO2, barometric pressure, ambient light, step count, motion, and altitude. Ask "what is my pulse" and it reads the sensor. Ask "check my vitals" for a full snapshot. Real readings, not hallucinated numbers.

Agent Kit integration. ClawWatch connects to Ant Farm rooms and other agents in real time through the Agent Kit. Ask "how is the family" and it summarizes what your people have been talking about.

Local command routing. Timers, pulse checks, vitals snapshots, and room summaries are handled on-device without hitting the LLM. Faster responses and works with spotty connectivity.

Admin panel. Configure everything from your browser at localhost:4747. Push API keys, switch models, change avatars, edit the system prompt, capture logs, rebuild and reinstall.

Release: https://github.com/ThinkOffApp/ClawWatch/releases/tag/v2.0.0

Video demo: https://x.com/petruspennanen/status/2030743559742083549

[DonPrus]

I will work with nullclaw binary size & performance next wek, but you already could build smaller binnary with
zig build -Doptimize=ReleaseSmall -Dchannels=telegram,cli -Dengines=markdown or other Dchannels=... & Dengines=... :)

[ThinkOffApp]

Hello Igor, thank you very much for your work! NullClaw seems to work really well, it was great immediately on the first install I did with my Samsung watch.

The binary size is already very small and works well with the Galaxy smartwatches I've tried, so we don't have any pressing need to make it even smaller. But good to hear about the ReleaseSmall option with selected channels. Right now ClawWatch bundles the full binary, and dropping unused engines should reduce the size quite a lot.

If there is anything specific about how ClawWatch uses NullClaw that would help you optimize, happy to share our build config and usage patterns. We're also making an Apple Watch version now by compiling Nullclaw into a static library which is then wrapped into a Swift package - but I don't have an Apple Watch yet to test it :D