Bug fixes
Streaming indicator stuck after agent error
The orchestrator ... waiting indicator in the chat UI could get stuck visible
after an agent finished — visible until a page refresh. The root cause was that
neither SessionLive nor ChatComponent cleared their streaming state on an
:error event. The agent would error out, the error entry would appear in chat,
but the indicator never cleared.
Both components now reset streaming, waiting, and streaming_agent_id when
an error event arrives for the active orchestrator.
Worker turn_end incorrectly cleared the orchestrator indicator
In multi-agent sessions, a worker finishing its turn would fire turn_end and
clear the streaming flag — even while the orchestrator was still active. This
could cause the send button to re-enable mid-turn, and queued prompts to be sent
prematurely.
turn_start, turn_end, and :error events in SessionLive are now gated on
orchestrator_event?/2: only events from the orchestrator update the top-level
streaming state and prompt queue. Worker lifecycle events are still forwarded to
the chat component for display, but do not affect session-level state.
Upgrading
You can run the following for upgrading planck:
Linux/MacOS
curl -fsSL https://thebroken.link/planck/install_docker.sh | shOr manually run the following:
cd ~/planck
if docker compose version >/dev/null 2>&1; then
alias dc="docker compose"
else
alias dc="docker-compose"
fi
dc -f compose.yml --env-file .env pull
dc -f compose.yml --env-file .env run --rm setup
dc -f compose.yml --env-file .env up -dWindows
irm https://thebroken.link/planck/install_docker.ps1 | iexOr manually run the following:
cd $HOME\planck
$Compose = if (docker compose version 2>$null) { @("docker","compose") } else { @("docker-compose") }
& $Compose -f compose.yml --env-file .env pull
& $Compose -f compose.yml --env-file .env run --rm setup
& $Compose -f compose.yml --env-file .env up -d