Allow redirecting ChatInterface mid-task with adaptive

[ahuang11]

demo.mp4

Inspired by https://x.com/_catwu/status/1922352915076849743

Finally: Real-time steering. Send feedback to Claude Code while it's working, without waiting for completion. Claude incorporates your input immediately, adjusting its approach based on new requirements or clarifications.

To start building out that capability to use in Lumen, I added an adaptive param in ChatInterface which allows users to continue sending messages while the first callback is still generating.

import param
import panel as pn
import asyncio
pn.extension()

async def adaptive_callback(message, user, instance):
    # Simulate a long-running response
    for i in range(5):
        await asyncio.sleep(0.5)
        instance.send(f"Step {i+1}: Processing '{message}'...", respond=False)
    return f"Completed processing: {message}"

# Enable adaptive mode for real-time interruption
ci = pn.chat.ChatInterface(
    callback=adaptive_callback,
    adaptive=True,  # Allow interrupting responses
    placeholder_text="Send follow-ups anytime to interrupt and redirect!",
    callback_exception="verbose",
    sizing_mode="stretch_both"
)
ci.show()