feat(demo): close the OTA loop for the rep demo

Makefile

@@ -262,6 +262,13 @@ sim-drive-stop: ## Stop the rover
 	@$(CONTAINER_ENGINE) exec $(SIM_CONTAINER) bash -c \
 	  'ign topic -t $(GZ_DRIVE_TOPIC) -m ignition.msgs.Twist -p "linear: {x: 0}, angular: {z: 0}"'
 
+# Demo helpers
+.PHONY: demo-swap
+demo-swap: ## Fire an OTA rollout: circle → figure-eight (robot-app on sim-robot-01)
+	curl -sX POST http://localhost:8081/v1/ota/rollouts \
+	  -H "content-type: application/json" \
+	  -d '{"image_ref":"localhost:14050/robot-app:figure-eight-v1","smoke_command":"true","cohort_selector":{"robot_ids":["sim-robot-01"]}}'
+
 # =============================================================================
 # CI cluster (smoke / pre-push parity) — alternate ports so it can run
 # alongside `make lab-up` on the same host. Used by .git-hooks/installer-smoke.sh

installer/docker-compose/docker-compose.sim.yml

@@ -53,6 +53,11 @@ services:
       # this, OTA runs the container on the default bridge network and
       # the new behaviour never reaches the simulated robot.
       OTA_RUN_ARGS: "--network=temporal-hack-lab_default,-e,ROS_DOMAIN_ID=42,-e,RMW_IMPLEMENTATION=rmw_cyclonedds_cpp"
+      # Controller images are built --platform=linux/amd64 (they layer on
+      # ROS Humble's amd64-only base). On Apple Silicon, docker pull defaults
+      # to arm64 and refuses the amd64-only manifest. Force amd64 for both
+      # the pull and the run so OTA can swap controllers cleanly.
+      DOCKER_DEFAULT_PLATFORM: linux/amd64
       # The repo is bind-mounted read-only. Go's workspace mode tries
       # to update go.work.sum on `go run` and fails; turn the workspace
       # off so the agent module's own go.mod/go.sum are authoritative.
@@ -71,13 +76,50 @@ services:
       # perform OTA. CONTAINER_SOCK is supplied by the Makefile at
       # compose time (resolves to docker.sock or podman.sock).
       - ${CONTAINER_SOCK:-/var/run/docker.sock}:/var/run/docker.sock
-    # apt is needed to install the C toolchain for go-sqlite3 (cgo).
-    command: >
-      bash -c "
-        apt-get update -qq && apt-get install -y -qq build-essential >/dev/null &&
-        cd /src/agent &&
-        CGO_ENABLED=1 go run ./cmd/agent
-      "
+    # apt installs: build-essential for go-sqlite3 (cgo), docker-ce-cli for
+    # OTA's `docker run`. The Bookworm-default `docker.io` package's CLI is
+    # API 1.41 — too old for Docker Desktop's daemon (needs 1.44+).
+    # Block scalar style is `|` (literal, line-preserving) instead of `>`
+    # (folded). Folded scalar plus yamlfmt will helpfully wrap the long
+    # `echo "..." > docker.list` line and break the redirect — bash then
+    # sees the redirect as a standalone command and truncates docker.list
+    # to empty. Literal scalar keeps each `&&` clause on its own line and
+    # yamlfmt leaves it alone.
+    command: |
+      bash -c '
+        set -e
+        apt-get update -qq
+        apt-get install -y -qq build-essential ca-certificates curl gnupg >/dev/null
+        install -m 0755 -d /etc/apt/keyrings
+        curl -fsSL https://download.docker.com/linux/debian/gpg | gpg --dearmor -o /etc/apt/keyrings/docker.gpg
+        chmod a+r /etc/apt/keyrings/docker.gpg
+        ARCH=$$(dpkg --print-architecture)
+        echo "deb [arch=$$ARCH signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/debian bookworm stable" > /etc/apt/sources.list.d/docker.list
+        apt-get update -qq
+        apt-get install -y -qq docker-ce-cli >/dev/null
+        cd /src/agent
+        exec env CGO_ENABLED=1 go run ./cmd/agent
+      '
+
+  # Initial robot-app behaviour. Auto-starts on `make sim-up` so the rover
+  # starts driving in a circle with zero clicks. OTA replaces this container
+  # by name (see agent/internal/ota/docker.go: docker rm -f robot-app +
+  # docker rename robot-app-new robot-app) — `container_name: robot-app` is
+  # required for that swap to find the right container.
+  robot-app:
+    image: localhost:14050/robot-app:circle-v1
+    container_name: robot-app
+    depends_on:
+      sim:
+        condition: service_started
+      mqtt:
+        condition: service_healthy
+    environment:
+      ROS_DOMAIN_ID: 42
+      RMW_IMPLEMENTATION: rmw_cyclonedds_cpp
+    networks:
+      - default
+    restart: unless-stopped
 
 volumes:
   bridge-sock: