CometAI Settings Features

Table of Contents

This article provides a comprehensive overview of the advanced configuration features within CometAI. Understanding and utilizing these settings can significantly enhance the model's performance, efficiency, and relevance for academic research, coursework, and development projects.

• Accessing Settings Features
• Feature Details and Technical Specifications
  • ​​​​​​Artifacts
  • Smart Focused Messages
  • Plugin Selector
  • Prompt Highlighter

Accessing Settings Features

To customize your CometAI experience, you must first access the settings panel where these features can be enabled or disabled.

1. Log in to the UTD CometAI portal.
2. Click the Settings icon (gear symbol) located in the top-left bar.
3. From the navigation menu, select Settings.

4. In the Configurations tab, you will find the features listed below. Check the box to enable a feature and click Save to apply your changes.

Feature Details and Technical Specifications

1. Artifacts

The Artifacts feature allows CometAI to generate and display rich, viewable content beyond simple text. This is ideal for creating comprehensive documents, reports, papers, or code projects. It supports various formats, including SVG graphics for charts, HTML for web content, and interactive React components.

• Example: When you ask CometAI to "generate an essay on the history of UTD," with Artifacts Settings enabled, it will produce a fully formatted document in the dedicated Artifacts view, separating the final output from the prompt conversation. More information on Artifact can be found here.

Technical Specifications

For developers, the Artifacts feature provides a powerful way to render complex outputs.

• Sandbox Environment: All generated content, particularly HTML and React components, is rendered in a sandboxed frame. This isolates the artifact's execution environment, preventing potentially malicious code from accessing your session data or the application's DOM.
• Component Rendering: The support for React components means you can prompt the AI to generate and display interactive data visualizations, custom user interfaces, or small, self-contained applications directly as an output.
• The icon to Allow Artifact Creation in the user interface’s plugin selector toggles the system's permission to generate these artifacts for a given prompt.

2. Smart Focused Messages

This feature makes your conversations with CometAI more efficient and contextually aware. Instead of sending the entire conversation history with every new prompt, it intelligently selects and sends only the most relevant messages, leading to faster and more accurate responses.

• Context Pruning: Instead of sending all the information to the AI at once, we only keep the most relevant pieces of information and ignore the rest. This way, the AI focuses on what really matters, while also saving space and cost.
• Benefits: This reduces API costs, lowers response latency, and helps prevent exceeding the model's maximum context window, which is crucial for long, complex conversations.

3. Plugin Selector

The Plugin Selector is a toolbar on the main chat screen that allows you to enable or disable specific tools for your current conversation. This gives you immediate control over the AI's capabilities without having to return to the main settings menu.

This feature acts as a front-end control panel for toggling various backend functionalities including RAG on Documents, Artifacts, Smart focused messages.      

Retrieval-Augmented Generation (RAG) on Documents

• Advantage: This is vastly more effective than sending the entire document. It allows for precise question-answering on documents that are too large to fit in the LLM's context window and yields more accurate, source-based answers. Disabling it would cause the model to receive the entire document content (if it fits).

Allow Artifacts creation

When enabled, this plugin provides the LLM with access to a sandbox code execution environment. This allows the model to write and run code to perform tasks like creating Web HTML Pages, data analysis, solving complex mathematical equations or generating plots.

Smart Focused Messages

This feature, indicated by the icon in the image, intelligently manages the conversation's memory. When active, it filters and sends only the most relevant messages from your chat history based on your current prompt. This improves response accuracy and speed by providing the AI with focused context, and it can be toggled off if you need the model to consider the entire, unfiltered conversation history.

Clear All Enabled Features

The Plugin Selector includes a convenient Clear All Enabled Features button, represented by the grid icon shown in the image. Clicking this icon instantly disables all currently active plugins in the toolbar. This serves as a quick reset, allowing you to start a new task with a clean slate without having to manually toggle off each feature one by one.

Click and Drag

You can customize the position of the vertical toolbar to fit your workflow. To move it, simply click and hold the "Click and Drag" icon. While holding the mouse button, drag the toolbar to your desired location on the screen and then release.

4. Prompt Highlighter

The Prompt Highlighter streamlines the editing and iteration process. It allows you to highlight any text within the AI's response or an artifact and use that selection as the direct context for a follow-up prompt, such as asking for a quick inline edit or an expansion of the selected topic.

There are two options to use this feature after enabling this either through Settings options or through Plugin Selector

1. Prompt
2. Fast Edit

Through Prompt

Think of it as quoting a specific message to reply to it. In CometAI, you can highlight any text in a response to "quote" it in your next prompt. This tells the assistant to focus exclusively on that selected information, allowing you to seamlessly ask for more details, get clarifications, or request edits on a specific point.

This action isolates the selection and provides it as context for a follow-up query. As shown in the image, you can then instruct the assistant to perform an action, such as "Write conclusion more detailed," ensuring the new response is precisely focused on the highlighted passage.

Through Fast Edit

For making immediate changes without generating a new message, use the "Fast Edit" feature. First, highlight the portion of the text you wish to modify. Next, click the "Fast Edit" button and type your instructions. The system will then process your request and substitute the original highlighted text with the revised content directly within the same response window, streamlining your editing workflow.

• Example:
  • Original Response:

• Suppose you want to edit inline:
  • You would highlight the text you want to edit
  • Select Fast Edit
  • Give the query/prompt you want to make changes for

• Response after inline edit changes:

After editing the response, if you want to revert back to your previous response, you can click on Revert Changes button that will bring back the previous response.