Context
Currently the user message and model response flow together visually.
Claude Code and similar tools visually separate the user input area with
horizontal rule dividers, making turns easier to scan.
Proposed behaviour
After the user submits a message, echo it inside a visual block before 'thinking':
────────────────────────────────────────────────────────
listame los archivos de mi carpeta src
────────────────────────────────────────────────────────
⠋ thinking… 42 tok
Or just a top rule + indented message with no bottom rule (cleaner).
Implementation sketch
In Agent.run(), after receiving user_input and before the Live spinner:
width = console.width
console.print(f"[{TEXT_MUTED}] {'─' * (width - 2)}[/]")
console.print(f" {user_input}")
console.print(f"[{TEXT_MUTED}] {'─' * (width - 2)}[/]\n")
Context
Currently the user message and model response flow together visually.
Claude Code and similar tools visually separate the user input area with
horizontal rule dividers, making turns easier to scan.
Proposed behaviour
After the user submits a message, echo it inside a visual block before 'thinking':
Or just a top rule + indented message with no bottom rule (cleaner).
Implementation sketch
In
Agent.run(), after receivinguser_inputand before the Live spinner: