Class AI_LLAMA_CHAT

This functionality turns a set of HTML elements into a chat interface for the LLAMA Model served by the CodBi. It enables interactive, multi-turn conversations about uploaded images and PDF documents, provides internet query access to the model via the Brave Search API and the client's location via the Geolocation API. Voice input is supported via the Media.Input.Speech.Whisper-Functionality.

If the model is not specified QWEN3-VL 2B is downloaded and utilized.

Required Elements (found by CSS class within the nearest common ancestor):

Generated CSS Classes (injected at runtime):

Behavior:

• The display textarea is made read-only and shows the full conversation history.
• When files are selected via the upload input, they are attached for subsequent messages.
• When the user clicks Send (or presses CTRL+Enter in the input), the message and any attached files are sent to the standard backend for processing by the AI model. The response is displayed in the chat.
• PDF files are automatically detected and processed (rendered to images or extracted).
• Multiple files can be attached; each is processed independently by the model.
• The send button and input are disabled during inference to prevent duplicate requests.

Config Parameters:

• MaxPages: Maximum PDF pages to process (default: 5).
• Rotation: Image rotation in degrees (90, 180, or 270). If it is known that the image to process is rotated, this can be set to avoid Tesseract OSD (if available) or the AI having to deal with it, speeding up the inference. Not setting or setting to 0 means that rotation is unknown.
• MaxPixelSize: Maximum total pixel budget (width×height). Images exceeding this are downscaled client-side while preserving the aspect ratio. Default: 3211264 (≈ 1792×1792). Set to 0 to disable client-side downscaling.
• LLAMABubble: Background color for Llama (AI) bubbles (default: #e5e5ea).
• UserBubble: Background color for user bubbles (default: #0b93f6).
• WelcomeText: Text shown after the model name(s) in the ready message (default: "Chat ready. Attach file(s) and type your question.").
• VoiceHotkey: Keyboard shortcut to toggle voice input, e.g. "Alt+A" (default: "Alt+A"). Format: modifier(s) + key separated by +. Recognized modifiers: Alt, Ctrl, Shift, Meta. The key part is case-insensitive.
• VoicePlaceholder: Placeholder text shown in the chat input when voice input is available. Default: "Alt+A = 🎙 on/off | Alt+Q = 🎙 off + send" (reflects the configured hotkeys).
• VoiceSendHotkey: Keyboard shortcut to stop recording and send, e.g. "Alt+Q" (default: "Alt+Q"). Same modifier format as VoiceHotkey.
• Language: Language code for Whisper speech-to-text (e.g. "de", "en"). Empty or unset means auto-detect.
• WaitingText: Text shown while waiting for the AI server to become available (default: "Waiting for AI server\u2026").
• LowConfidenceText: Warning text shown when the AI response has low confidence (default: "Low Confidence").
• RethinkButtonText: Button label offering to re-answer with the thinking model (default: "Rethink").
• UncertainText: Tooltip text shown when hovering over uncertain (low-confidence) tokens (default: "Low confidence").
• ShowUncertainTokens: Whether to visually highlight uncertain tokens in AI responses. Set to "false" to disable highlighting (default: "true").
• ResponseLanguage: Two-letter ISO 639-1 code (e.g. "de", "fr"). When set, the AI is forced to respond in this language — no auto-detection is performed. Overrides the AI_LLAMA_STD_Language plugin property for this instance. The chat interface reflects this language for labels where available.
• Specialist: Name of a specialist model registered via AI_LLAMA_STD_SPECIALIST_XXX plugin property. When set, requests are routed to that specialist's dedicated server instance (case-insensitive match).
• QueueBadge: If set to "true", shows a badge with the current queue position while waiting for inference. Overrides the AI_QueueBadge plugin property for this instance. Default: determined by plugin property.
• QueueText: Text appended after the queue position number in the badge (e.g. "in queue" → badge shows "3 in queue"). Default: empty.
• FilterResults: If set to "true", enables PII filtering on Brave Search queries for this instance, overriding the global AI_BraveSearch_FilterResults plugin property. Default: determined by plugin property.