Skip to main content

8.6 Natural Language Queries

The AI Chat feature lets you query your operational data using plain language. Instead of writing TDengine SQL or navigating the asset hierarchy to find a value, you can ask questions like "what was the peak voltage for em-1 yesterday?" and receive an answer grounded in your actual data.

Opening AI Chat​

Click AI Chat in the top navigation bar. This opens the AI Chat page at /ai.

The AI assistant is named Teddy.

Interface Layout​

The AI Chat page has two areas:

Conversation history (left panel). A list of previous chat sessions. Click any session to revisit its conversation. New sessions start with an empty history.

Chat area (center/right). The main interaction area containing:

  • The greeting message: "Hi, I'm Teddy. How can I help you?"
  • The text input field with the placeholder: "Ask me anything about your data, press Enter to send, press Shift+Enter to wrap"
  • A Deep Thinking toggle button at the bottom left of the input field
  • A microphone button at the right of the input field for voice input
  • A Suggested Questions section with a refresh icon, below the input field

Asking a Question​

Type your question in the input field. Press Enter to send, or Shift+Enter to add a line break without sending.

You can also press the microphone button to use voice input. Speak your question and the system transcribes it into the input field.

Deep Thinking Mode​

The Deep Thinking toggle switches the query to the Deep Thinking Model (configured in Connection Management). Use this mode for complex multi-step questions that require extended reasoning — for example, correlating multiple attributes across a time window, or asking for a root cause hypothesis. Standard questions about data values do not require Deep Thinking mode.

Suggested Questions​

Below the input field, the Suggested Questions section displays pre-generated questions relevant to your data context. These suggestions are generated by the system based on your asset hierarchy and collected data.

Click the refresh icon next to "Suggested Questions" to load a new set of suggestions.

Click any suggestion to use it directly — the question is placed in the input field and sent automatically.

How Teddy Answers​

When you submit a question, Teddy queries your TDengine data using the appropriate SQL, retrieves the results, and returns a natural language answer. For data questions, the response includes the computed values. For analytical questions, it may include charts, tables, or a narrative explanation.

All answers are grounded in your actual data — Teddy does not fabricate values.