v0.31.1

• Give the terminal a consistent OpenRUA palette, with lime brand accents,
  quiet local surfaces and distinct status colors. Adapt to light and dark
  terminals while preserving the terminal's overall background and supporting
  limited-color and no-color displays.
• Introduce the project workflow film in the README's inline player. Keep the
  original robot task video directly playable in the detailed demo guide.

v0.31.0

• Remove the duplicate plain text chat interface, including --cli and the
  chat command. Use openrua for terminal chat, openrua --gui for the
  browser, or openrua --name NAME to reconnect. Scripts can submit messages
  and inspect results with session send, session status and session events.
  Keep the shared session service, history, native agent and exec paths.

• Restore the README's product positioning, inline demo and visible agent,
  robot, simulator and benchmark coverage. Keep quick start focused on a first
  task, illustrate the native ROS 2 workflow and link paper results from the
  benchmark section. Select major news while keeping the full history linked.

• Align setup labels, project settings and exit instructions in the user guides
  with the current terminal, including No as the default resource choice.

v0.30.5

• Restore information omitted during the wording update: conversation recovery
  and viewing limits, waiting-message behavior, robot request results and IDs,
  connection errors, environment update reasons and skill validation limits.
• Keep the clearer English while preserving supported input routes, permission
  boundaries, file-preview limitations and the scope of recurring reviews.

v0.30.4

• Use plain English across setup, history, chat, robot controls, files, skills,
  scheduled reviews, updates and the browser. Explain what happened and what
  to do next. Keep each agent's permission names and behavior.
• Simplify the resume page and explain saved conversations when they open.
• Record separate shutdown times for pending robot work, the agent and resource
  cleanup in the service log, so a slow stop can be traced to the right step.

v0.30.3

• Cancel robot readiness and scene initialization when ending a session during
  startup, then join the existing owner's cleanup. Stop no longer waits for
  the robot to finish starting before releasing it.
• Cancel only the owned readiness probes or simulation resources. Existing
  robot services and configured real-driver shutdown grace remain unchanged.

v0.30.2

• Count physical presses for the double exit shortcut. Key releases and held
  Ctrl+C/Ctrl+D repeats no longer trigger exit shortcuts with Kitty input.
• Keep the resource choice open after further Ctrl+C or Ctrl+D input. Use Back
  or Escape to return, or Enter to confirm the selected option.

v0.30.1

• Keep short service liveness checks separate from normal command timeouts,
  so resource shutdown can finish before leaving or switching conversations.
• Confirm a lost shutdown reply from explicitly closed retained records,
  without repeating the shutdown request or returning to a dead connection.
  Unconfirmed shutdown failures still block departure.

v0.30.0

• Give Ctrl+C Claude Code's interrupt priority: stop a running response while
  keeping the input draft; when idle, clear input or open the leave choice.
  Keep Claude Code's double press before the No/Yes leave choice, with a visible
  hint and a Back option to cancel leaving.
• Explain that No stops the agent, frees memory and stops any simulator or
  drivers OpenRUA started. Retain files and chat history.
• Remove dash punctuation from the maintained workspace guide.

v0.29.1

• Use plain language when asking whether to keep a session running in the
  background. State that files and chat history are saved, and describe the
  choices as stopping the session or keeping it running.

v0.29.0

• Ask whether to keep execution resources when leaving a terminal conversation
  or switching through history or setup. Default to No: stop owned execution
  resources while retaining project files and history. Yes keeps execution in
  the background. Cancelling selection or setup leaves the current session
  unchanged; an unconfirmed shutdown prevents the switch.

• Keep one stable model field in the conversation status. Show a response's
  different model only when the native agent confirms that new settings apply
  after it finishes; retain the start snapshot in history without displaying it
  on every active response.