Chat UI and Markdown/image rendering improvements:

• Unified and enhanced Markdown rendering for assistant responses in both gl-google.js and lm-studio.js, including support for up to three [Image of ...] placeholders per response. These placeholders are replaced with actual images fetched via Google Image Search, and the rendering process safely tokenizes and restores <img> tags to avoid interfering with Markdown parsing. [1] [2]
• User messages in lm-studio.js are now sanitized and rendered as styled chat bubbles, matching the assistant's UI and preventing HTML injection.

Configuration and API key management:

• Improved logic for loading the Google Gemini API key, preferring inline config or global variables before falling back to config.json. This makes configuration more robust and flexible.
• Ensured that GOOGLE_GL_KEY is set when applying configuration in options.js.

Theme and UI modularity:

• Modularized theme stylesheet loading and unloading, enabling dynamic theme switching and ensuring that only the relevant stylesheets are active. The LCARS theme now loads its CSS dynamically, and the monitors dock is shown only when this theme is active. [1] [2]
• Added logic for monitor tab switching in the LCARS theme, improving the UI experience for theme-specific panels. [1] [2]

Context window/token usage monitoring:

• Introduced a lightweight token/context window monitor that estimates token usage for current messages and responses, displays usage as a progress bar, and updates periodically or on model change. This helps users stay within model limits.

Error handling and accessibility:

• Added global error and unhandled promise rejection handlers to display status messages in the footer, improving user feedback and debugging.
• Minor accessibility and robustness improvements, such as checking for element existence before manipulating DOM nodes.