Omni Agent

The Omni Agent provides a standalone, conversational interface for exploring your data. Ask questions, generate queries, and visualize results without needing to work directly in a workbook.

​

Requirements

To use the Omni Agent, you’ll need:

• Restricted Querier, Querier, Modeler, or Connection Admin permissions
• To verify that the following settings in AI Hub > General > Features are enabled. These settings are enabled by default, but Organization Admins can modify them as needed.
  • Omni Agent - Required to use the Omni Agent
  • Omni Agent > Chat - Required to access the chat interface
  • Omni Agent > Chat > Chat in navigation - Required to show the standalone chat in the main sidebar

​

Accessing the Omni Agent

To access the Omni Agent, click Omni Agent in Omni’s main sidebar. The chat interface opens in a dedicated page where you can begin asking questions about your data.

​

Capabilities

The Omni Agent can:

• Answer natural-language questions about your data
• Search existing dashboards for tiles and visualizations that already answer your question before generating a new query
• Generate queries based on your prompts, routed through topics or any view in your model
• Create visualizations, including KPI tiles, directly from your prompts
• Create dashboards based on the queries and visualizations in the session
• Create routines that run a saved prompt on a schedule and deliver the result by email or Slack, and find and open your existing routines by describing them in plain language
• Accept files and images as context:
  • Upload files including text and Markdown files to provide additional context alongside your queries
  • Attach images by pasting from your clipboard to provide visual context for your questions
• Save information about you to memory to provide more personalized responses
• Search the Omni docs to answer questions about how to use Omni

​

Getting help

If you’re stuck and can’t remember how to do something in Omni, ask the Omni Agent. Questions like “How do I do [thing]?” will prompt the agent to search the official Omni docs and provide you with an answer, all without leaving your Omni workflow. You can also directly tell the agent to search the docs when researching the answer to your question.

​

Tuning responses

You can shape how the Omni Agent responds by scoping which data it draws from, controlling the context it can access, and choosing which underlying AI model handles your prompts.

​

Scoping responses

The Omni Agent uses pickers to scope responses to specific connections, models, and topics. Click the button at the bottom of the chat box to open the pickers: Up to three pickers can display in the chat interface:

• Connections - Lists the connections you have access to. Only displays if more than one connection is available. This setting can’t be changed if you’re using the chat in a workbook or dashboard.
  Organization Admins can configure a default connection in AI Hub > Setup > Features to set which connection the Omni Agent opens with by default.
• Models - Lists the available models in the selected connection. Only displays if:
  • The connection has more than one model, and
  • At least one topic in the model is accessible. This could mean that the model’s ai_chat_topics is unset, making all topics accessible, or that it specifies at least one topic:
    ai_chat_topics: [ Products ]
    
• Topics - Lists the available topics in the model. Only displays if at least two topics in the model are accessible. This could mean that the model’s ai_chat_topics is unset, or that it specifies at least two topics:
  ai_chat_topics: [ Products, Orders ]
  

​

Scoping topics

If a prompt is entered and a topic isn’t selected - meaning that Auto-select a topic is selected in the topic picker - the AI will attempt to select the most relevant topic. To scope the AI’s response to a specific dataset, use the pickers to select a specific topic or composite topic. The AI will remain “locked” to the selected topic until the selection is changed. Additionally, the picker menus won’t display when a model only has one AI-accessible topic. If a connection has one model with a single AI-accessible topic - determined by the value of the model’s ai_chat_topics parameter - the AI will be scoped only to that topic and the pickers will not display.

​

Managing AI integrations

The Omni Agent can use AI integrations to access context from tools like Google Drive, Google Calendar, Slack, GitHub, and Notion. To manage which integrations are available during your chat session, click the button at the bottom of the chat box. AI integrations that an Organization Admin has enabled for your organization appear at the top of the menu. Each integration shows one of two states:

• Connect - If you haven’t connected your account yet, click Connect to launch the OAuth flow. This is the same authentication flow available from your user account settings.
• On/Off toggle - If you’ve already connected your account, the integration displays an on/off toggle. Turning the toggle off temporarily withholds that integration’s tools from the Omni Agent for your session without disconnecting your account. You can turn the integration back on at any time.

The on/off toggle is per-user and doesn’t affect other users in your organization. It only controls whether the integration’s tools are available to the agent in your current and future chat sessions.

​

Saving information to memory

The Omni Agent can remember information about you between chat sessions. Memory holds personal context — such as your team, the region you cover, how you like results presented — and the agent draws on it when it’s relevant to your prompt. With memory, you add personal context once and it automatically applies in every session when relevant. For example, a prompt like “how did we do last quarter?” can be automatically filtered to your region and presented in your preferred format based on memory, instead of needing to enter the prompt in every new session. You can ask the Omni Agent to save information to memory for you, or you can add it manually on your account settings page. Memory is personal and scoped to a single organization. It isn’t shared with your team, and the agent never uses it in anyone else’s conversations. If you belong to more than one Omni organization, each keeps its own memory. Use memory for facts about you and how you work. Definitions that apply to everyone — how a metric is calculated, what a business term means — belong in your data model, where the whole organization benefits from them.

​

Selecting an AI model tier

The Omni Agent provides a model tier selector that allows you to choose between different AI models for your chat session. This lets you optimize for speed or intelligence based on your needs:

• Smarter - Uses the most capable model (Opus) for complex analysis and nuanced queries
• Standard - Uses a balanced model (Sonnet) that provides good performance for most queries
• Faster - Uses a fast model (Haiku) for quick responses to simpler questions

To select a model tier, click the model selector next to the (microphone) icon and choose your preferred tier from the dropdown menu. The selector displays the currently active tier, which defaults to your organization or model-level AI settings.

​

Selection reset

Your selected model tier will persist throughout the session, applying to prompts sent after the tier is selected. However, the selection automatically resets to the default tier when you:

• Switch to a different data model using the model picker
• Start a new chat session

The default tier for each model is determined by the model’s ai_settings configuration, falling back to organization-level settings when not specified. If neither is configured, chat starts on the Standard tier.

​

Searching existing dashboards

Before generating new queries, the Omni Agent can search for existing dashboards that might already answer your question. When you ask a question, the AI analyzes your prompt and searches through:

• Dashboard names - Searches the content index for dashboards with relevant titles
• Tile and chart names - Searches for specific visualizations that match your question

The AI uses multiple search variations of your keywords to improve the chance of finding relevant content. For example, if you ask about “inventory items count”, it might search for “inventory items”, “items count”, and “inventory” separately. You can also provide the AI with a dashboard’s ID or custom identifier. When the AI finds potentially relevant dashboards, it: This helps you discover existing content before creating new queries, promoting reuse of trusted dashboards and reducing duplicate analysis work.

​

Creating queries and dashboards

When the Omni Agent answers a question, it generates a live query you can inspect, refresh, or roll up into a dashboard alongside other charts from the session.

​

Viewing query field details

When the Omni Agent generates a query, you can view details about the fields used in the results. This helps you understand what data the AI selected and how each field is defined. To view field details, click the icon near the top-right corner of the query results, next to the icon. A panel displays information for each field in the query, including:

• Field name - The field’s label and identifier in the model
• Definition - The SQL or calculation that defines the field
• Description - Additional context about the field, if defined in the model

This feature provides transparency into AI-generated queries, helping you verify the AI selected appropriate fields and understand the underlying data structure. To improve the quality of field definitions displayed here, refer to Optimizing your models for AI.

​

Refreshing query results

When you revisit a previous chat session, query results may be stale. To update the results with the latest data, click the icon floating above the chat box: Clicking this icon re-runs all queries in the conversation, updating visualizations and results with current data. To optimize performance, only the most recent queries are re-run.

​

Creating dashboards

The Omni Agent can build dashboards. There are a few ways to get started:

• Ask the agent directly to build a dashboard — describe the dashboard you want in the chat and the agent plans the layout, generates queries, and publishes a complete dashboard to your My documents folder:

  “Create a dashboard showing sales performance by region with monthly trends and a top customers table” After the dashboard is created, you can ask the agent to iterate on it in the same session — rearranging tiles, editing charts, or adding new ones.

• From an existing chat — you’ve been exploring data and created a few visualizations in a chat session, ask the agent to pull them into a dashboard
• Upload an image — drop in an image of a dashboard you’d like to replicate. The agent parses what it sees and uses it as a starting point

The model and thinking level used for AI dashboard creation follow the model’s build_configuration setting. To disable dashboard creation from chat, users with the required permissions can add the create_dashboard_from_chat tool to a model’s ai_settings.disabled_tools. This setting is per-model, not organization-wide - you will need to set it on every model to disable dashboard creation from chat entirely.

​

Creating routines

The Omni Agent can create routines, find and open your existing routines from chat, and delete a routine when it’s open in the split view. See AI Routines for more information.

​

Adding files as context

You can upload text and Markdown files or paste images from your clipboard to give the Omni Agent additional context for your questions. Files are processed in your browser and are not stored on Omni’s servers.

​

Uploading files

You can upload text (.txt), Markdown (.md), LookML (.lkml), XML (.xml), HTML (.html, .htm), and images to the Omni Agent to provide additional context for your questions. This is useful when you need to reference additional information alongside your data queries. To upload a file:

1. Click the icon at the bottom of the chat box
2. Select Attach file
3. Select a file from your computer
4. The file appears as a file chip in the chat box, indicating it’s ready to be sent with your message

The Omni Agent can then read and reference the contents of the uploaded files when generating responses, allowing you to ask questions that combine your data with external context or documentation.

​

Attaching images

You can paste images directly from your clipboard into the chat to provide visual context alongside your questions. This works with screenshots, images copied from web pages, or any image data on your clipboard. To attach an image:

1. Copy an image to your clipboard (for example, take a screenshot or copy an image from a web page)
2. Click into the chat box
3. Paste the image using your system’s paste command (Cmd+V on Mac, Ctrl+V on Windows/Linux)

The image appears as a file chip in the chat box, indicating it’s ready to be sent with your message. You can attach multiple images to a single message by pasting them one after another.

​

Managing and sharing chat sessions

Each conversation with the Omni Agent is saved as a chat session you can revisit and share with other users in your organization.

​

Navigating prompt history

You can quickly recall previous prompts using keyboard shortcuts, similar to a command-line interface:

• Press the Up arrow key to cycle backward through your previous prompts
• Press the Down arrow key to cycle forward through your prompt history

When you reach your oldest prompt and press Up arrow again, the navigation wraps to your most recent prompt. Pressing Down arrow after your newest prompt restores any in-progress text you were typing before navigating the history. The prompt history displays only your own prompts from the current chat session.

​

Managing long conversations

As your conversation with the Omni Agent grows, it consumes more of the available context window. When a chat session reaches 75% of the context window, a visual indicator appears on the New chat button to help you recognize when starting a fresh conversation would improve performance. Context window size is determined by the conversation_prune_length parameter. This feature is automatic and requires no configuration. Starting a new chat when prompted helps maintain optimal AI performance, especially for complex queries and analysis.

​

Sharing chat sessions

You can share a chat session by copying the URL from your browser’s address bar and sending it to another user. Query results stored in the chat history are cached from when the creator ran them. The viewer of the shared chat will see those same cached results. Therefore, Omni requires that the viewer of the shared chat have Querier, Modeler, or Connection Admin permissions, as well as the same connection environment, to view the session. Users with Restricted Querier permissions or lower will not be able to access shared chat sessions. The following rules determine who can see a chat:

1. The owner of the chat can always view the chat.
2. Any user with the Querier role or above on all relevant models can view the chat. Because Queriers can already write SQL queries directly in Omni, this doesn’t expose any data they couldn’t otherwise access.
3. Viewers and Restricted Queriers are blocked from viewing, preventing access to data they are not authorized to see.

​

Customizing and embedding the Omni Agent

Organization Admins can tailor the agent’s appearance to match their brand and surface it inside external applications.

​

Applying branding

Organization Admins can customize the appearance of the AI chat interface, including the agent’s name, icon, and greeting messages. Refer to the AI branding settings for more information.

​

Embedding the agent

The Omni Agent can also be embedded into external applications. Refer to the embedding the Omni Agent guide for setup instructions.

​

Next steps

• Optimize your models for AI to improve the quality of AI-generated responses
• Configure topics to organize and scope the data the Omni Agent can access
• Customize the Omni Agent appearance with your own branding