Feature Request: Display model reasoning chain (e.g., <think> tags) in the User Interface

[RangerVzla]

Description

When integrating local models (such as the Qwen or DeepSeek families) via llama-server or similar backends, these models often generate a chain of thought or internal reasoning process prior to delivering the final answer. This reasoning is typically enclosed within specific tokens or tags, most commonly <think> and </think>.

Currently, Row-Bot does not expose this reasoning process within the primary chat interface. The raw text appears to be parsed out, hidden by default, or only accessible by inspecting system logs.

Proposed Solution

I request the implementation of a UI feature to make the reasoning chain visible to the user. Possible approaches include:

• An expandable and collapsible panel (e.g., an accordion component) located above the final response.
• A toggle option within the application settings to globally enable or disable the "Show reasoning process" view.
• Rendering the content within <think> blocks as a distinct visual element (such as a specialized code block or quote container) instead of omitting it from the rendered output.

Additional Context

Exposing the chain of thought is highly valuable for users operating local inference servers. It is essential for debugging model behavior, understanding the logic behind the agent's tool execution, and verifying the deductive steps taken during complex queries. Adding this visibility would significantly improve the troubleshooting experience for advanced users.