AutoCua v0.1.3: iPhone, browser ,coding and computer use at once

[FunctionFreak]

What's new

New chat UI

• The main area works like a chat. Each turn has its own recording, its own bot and a reply box.
• The reply glides to its last line, and the agent strip shows a thinking orb.

Computer use

• Browser, iOS and coder sub-agents each get their own box.
• Screens keep one shape, and browser screenshots sit in the centre.

Shell use

• Shell use is now a chat with its bots.
• The terminal box is the run's tool box.

Intro and modes

• New splash with all seven bots.
• A picked mode's bot comes in big, then becomes the run's bot.

Left bar

• Flat and collapsible, with a pure white app background.
• The chat list runs down to the footer and reads clearer.

Recording

• Every run is recorded and compressed.
• On Linux, recording picks a GPU encoder that works, or falls back to CPU.

Web scrape

• more continues from exactly where the last read stopped.

Security

• The app's API needs a per-run secret token.
• The simulator's WebDriverAgent listens only on this Mac.
• Upgraded pyo3 to 0.29.
• Upgraded Pillow, requests, python-dotenv, Flask and Werkzeug.

Docs

• Updated agent_capabilities.md.

Merged PRs: #8 to #16

---

This discussion was created from the release AutoCua v0.1.3: iPhone, browser ,coding and computer use at once.