Problem

When a session approaches the model's context limit, the user gets a hard failure:

⚠️ Context overflow — prompt too large for this model. Try a shorter message or a larger-context model.

This is abrupt and loses the user's message. There's no way to know you're approaching the limit until you hit it.

Proposed Solution

Add a warning threshold (e.g., 80-90% of context) that:

1. Warns the user before overflow:

   ⚠️ Context at 85% capacity. Consider using /forget or /new soon.
   

2. Optionally auto-compact when threshold is reached (configurable)

3. Show context usage in /status output (tokens used / max)

Benefits

• Users can proactively manage context before losing messages
• Better UX than hard failures
• Helps users understand when to start fresh sessions
• Could enable smarter auto-compaction strategies

Configuration (suggested)

contextWarning:
  enabled: true
  threshold: 0.85  # warn at 85%
  autoCompact: false  # or true to auto-compact at threshold