Current Issue: We completely ignore thinking text from thinking models.

Expected Behavior: Display thinking text/chain-of-thought from models that support it (like o1, Claude 3.5 Sonnet, etc.) in the chat interface.

Implementation Notes:

• Parse and display intermediate thinking steps
• Show a "thinking..." indicator during processing
• Display the thinking process in a visually distinct way
• Ensure compatibility with different model providers

Priority: High - This is a key UX improvement for transparency in AI responses