LLM Enhance

[andyjjrt]

This pull request introduces multimodal support for the LLM chat command, adds an image generation command for bot owners, and improves the Discord UI for both chat and image generation. It also adds a new command to reload music nodes and updates configuration to support separate image API credentials. The most important changes are grouped below.

Multimodal LLM chat and UI improvements:

• Added support for image attachments in the chat command, allowing users to provide an image alongside a text prompt for vision-capable models. Only bot owners can select models; others are restricted to the default model. [1] [2]
• Enhanced the LLMView UI to display attached images, show truncated prompts, and provide a "Cancel" button during streaming. Added blockquote formatting for "thinking" content and improved streaming chunk handling to display reasoning and allow cancellation. [1] [2] [3] [4] [5]

Image generation command and UI:

• Introduced a new /image command for bot owners to generate images from prompts, with options for size, inference steps, CFG scale, and seed. The command uses a separate OpenAI client and displays results using the new ImageView UI. [1] [2] [3]
• Added the ImageView class to provide a rich Discord UI for image generation, including loading, result, error states, and timing information.

Configuration updates:

• Updated LLMConfig to support separate API keys and model names for image generation, enabling secure and flexible multimodal support.

Music bot improvements:

• Added a new owner-only reload_nodes command to reload Lavalink nodes, improving music bot reliability and maintainability.

---

References:
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]