Model Selection override issue when switching between regular chat and AI Character

Steps to Reproduce:

1. Open an AI character chat (e.g., "Batman") and verify the "Selected Model" in its configuration (e.g., "Venice Large").

2. Switch to a regular chat session (non-character mode).

3. In the regular chat, change the active model to a different one (e.g., "Llama 3.1").

4. Return to the AI character chat ("Batman").

5. Observe that the character now uses "Llama 3.1" instead of its configured "Venice Large".

Expected Behavior:
AI characters should consistently use the model explicitly defined in their configuration, regardless of model changes in regular chat sessions.

Actual Behavior:
The AI character adopts the last model used in a regular chat session, overriding its predefined "Selected Model" setting.

Impact:

• Breaks model consistency for AI characters.

• Forces users to reconfigure characters after switching between chat modes.

• Risk of unintended behavior in AI character responses due to incorrect model usage.

Environment:

• Platform: Venice.ai web interface (Pro version).

• Browser:

  Brave 1.79.119 (Official Build) (64-bit), Chromium: 137.0.7151.68

• Operating System: Manjaro Linux 6.14.6-2-MANJARO

• Device: Desktop

• Venice Session State: Bug occurs only after extended use

Additional Notes:

1. Workaround: Logging in as a fresh user resolves the issue temporarily.

2. Scope: The bug appears to affect all AI characters with predefined models.

3. Possible Root Cause: Improper isolation of model state between regular chat and character-specific configurations.

Request:

• Fix the model persistence logic to ensure AI characters retain their configured models across session interactions.

• Add a toggle to display the currently active model in the UI for troubleshooting.