Purpose

This PR improves the eSim Chatbot by fixing stability issues, improving model selection and session handling, and adding testing utilities.

Approach

Fixes

• Fixed STT backend detection priority.
• Fixed Ollama stream KeyError crashes.
• Fixed debug_error() crash when no project is open.
• Fixed Windows CMD popup during Ollama startup.
• Improved model fallback and text model selection.
• Fixed duplicate chat sessions after restart.

Improvements

• Added adaptive token budgeting for faster responses.
• Added user-friendly "No models installed" guidance.
• Centralized vision model keywords and model-fetching logic.

Testing

• Added run_chatbot.py for standalone testing.
• Added verify_chatbot.py for environment validation.

Compatibility

• No breaking changes.
• Backward compatible with existing chat history.