Docker¶

Everything lives under .docker/. The image is a two-stage build on python:3.12-slim — no compiler at runtime, no ROS.

docker compose -f .docker/docker-compose.yml build

Try it with no robot¶

docker compose -f .docker/docker-compose.yml run --rm btview demo

Watch Nav2 running on the host¶

The btview service uses network_mode: host, so 127.0.0.1:1667 inside the container is the same socket the robot's publisher bound on the host:

docker compose -f .docker/docker-compose.yml run --rm btview
docker compose -f .docker/docker-compose.yml run --rm btview snapshot --active

Host networking is a Linux feature. On Docker Desktop (macOS, Windows) use the published address instead:

docker run --rm -it btview:latest tui --host host.docker.internal --port 1667

Watch a robot elsewhere on the network¶

docker run --rm -it btview:latest tui --host 192.168.1.42 --port 1667

Two containers, private network¶

The split profile runs a synthetic publisher and a monitor on a compose bridge — useful for exercising the network path without a robot:

docker compose -f .docker/docker-compose.yml --profile split up

mock binds 0.0.0.0:1667 inside the network; monitor connects to tcp://mock:1667.

Getting a good terminal¶

The image sets TERM=xterm-256color, and compose forwards your own $TERM. Two things matter for the full-screen UI:

-it — without a TTY, Textual has nothing to draw on. docker compose run includes it; a bare docker run needs it spelled out.
Font — if glyphs come out as boxes, add --ascii.

Non-interactive commands need neither:

docker run --rm btview:latest snapshot --host 192.168.1.42 --ascii --no-color

Getting files out¶

export, record and the TUI's s key write to the working directory. Mount one:

docker run --rm -v "$PWD:/out" -w /out btview:latest \
    export --host 192.168.1.42 --format mermaid -o tree.mmd

docker run --rm -v "$PWD:/out" -w /out --network host btview:latest \
    record -o /out/run.btv.jsonl --duration 60

Building for a robot¶

The image is architecture-agnostic; for an ARM64 robot from an x86 laptop:

docker buildx build --platform linux/arm64 -f .docker/Dockerfile -t btview:arm64 --load .

Honestly though: on the robot itself, pipx install btview is smaller, starts faster, and sidesteps the networking entirely. Containers are the better answer on your workstation.