Feature Request: Persistent Model-Specific System Prompt Mapping

[arleas]

Problem:
Currently, switching models requires the user to manually update the system prompt to match the specific requirements, formatting, or "persona" optimized for that model. This is tedious when rotating through multiple models and prone to errors (e.g., using a Gemma prompt for a Qwen model), which degrades output quality.

Proposed Solution:
Implement a configuration system that allows users to associate specific system prompts with model identifiers.

Mapping System: Introduce a prompts.json or a UI-based mapping tool where users can define a ""Prompt ID"" and associate it with one or more models.
Regex Support: Instead of just exact filename matching, allow for Regular Expression (Regex) patterns. For example, any model matching .*gemma.* would automatically load the ""Gemma_Core"" prompt, while .*qwen.* would load the ""Qwen_Optimized"" prompt.
Automatic Trigger: Trigger the prompt load sequence during the model_load event, overriding the default global system prompt.

Expected Behavior:

User selects "gemma-2b-it" →→ UI detects the 'gemma' regex →→ System Prompt is automatically populated with the saved Gemma prompt.
User selects "qwen-7b-chat" →→ UI detects the 'qwen' regex →→ System Prompt updates instantly.