AI Assistant Overview

The AI Assistant lets you generate visualizations by describing what you want in plain language. Instead of manually configuring charts, specify the question, date range, grouping, aggregation, and filters in a natural conversation — the assistant builds the view for you.

To start a conversation with the AI Assistant:

1. Click the Assistant icon on the left side of the Trendz workspace, then press New Chat.
2. In the popup, select the business entities the assistant should consider for this conversation. You can modify the selection at any time. Click Save to apply.
3. Type your question and send it.
4. Review the generated visualization.
5. If needed, send follow-up messages to refine the result.

The assistant works with both telemetry and attributes from your business entities.

The assistant can generate the following visualization types:

Visualization	Notes
Table	Standard table without dynamic columns
Dynamic Table	Table with dynamic columns
Line Chart
Pie Chart
Bar Chart	Vertical, Horizontal, and Stacked variants
Scatter Plot
Heatmap
Heatmap Calendar
Card	Without compared value
Card with Line

The assistant applies a date picker to generated visualizations (this week by default).

The assistant can use date fields as part of the query.

The assistant assigns names to columns and the visualization based on your request.

The assistant adds filters to the generated view when specified in your question.

The assistant responds in the language you use.

The assistant remembers the full conversation context and generates answers based on all previous messages, specifications, and responses. You can refine a result over multiple turns without restating the original request.

After sending a message or selecting a previously saved chat, you are redirected to the Chat Page. The chat history appears on the left side of the screen, and the visualization for the selected message appears on the right.

Type your message and click the Send icon on the right side of the input field.

Click the pencil icon next to the last message to modify it, then confirm or cancel the change. Use this when the assistant returned incorrect results and you want to provide more detail in the same message.

Click any message in the chat history to display its previously generated answer on the right side of the screen.

Click the Refresh button at the top right of a message block to regenerate the assistant’s answer. Use this when a response was incorrect.

Click the Assistant’s Interpretation label at the top left of a message to see how the assistant understood your request. This helps verify whether the AI correctly interpreted your question.

Click the Save config button to save the generated view to a view collection.

Click See Details at the top right of the screen to open the generated view configuration. From there, click Settings to access additional settings, or Back to exit. This helps verify whether the assistant correctly selected business entities, fields, and aggregations.

Click Get Chat Summary at the top right of the screen to generate a summary of the conversation. The summary appears at the top of the screen and helps identify anomalies or correlations across the generated views.

The assistant builds visualizations by analyzing your topology — the business entities and their fields defined in Trendz. When a topology has hundreds of entities or fields, the accuracy of the assistant may decrease.

Use the Topology Entities window to explicitly tell the assistant which entities and fields are relevant to the current conversation. You can access it by starting a new chat or by clicking the entity selection next to the Topology Entities label in the upper left corner of the screen.

• Narrow the topology: Use the Topology Entities tab to include only the entities and fields relevant to your question — especially important for large topologies.
• Add field descriptions: When field names are abbreviated or unclear, add descriptions so the assistant can interpret them correctly.
• Refine with follow-ups: If a result is close but not quite right, send a follow-up message with additional specifications instead of starting over.
• Use your own model: For self-hosted installations or to avoid cloud token limits, configure a custom large language model. See AI Settings.
• Choose the right model: When self-hosting, balance model capability against cost — more capable models produce better results but cost more per query.

The number of messages you can send per month depends on your subscription type. When the monthly limit is reached, AI Assistance is unavailable for the rest of the month. Configuring your own model removes this restriction. See AI Settings.

The number of messages per minute is capped at:

Scope	Limit
Per user	6 messages/min
Per customer	20 messages/min
Per tenant	50 messages/min

Self-hosted installations do not include a default large language model. You must configure a custom model before using AI Assistance. See AI Settings.

The following capabilities are not yet supported and will be added in future releases:

• Creating anomaly or prediction models
• Creating calculation fields
• Creating cards with a compared value
• Specifying view or view field settings