Section 05

Chat & querying

The chat panel is on the right side of the Dashboard. It connects to an AI agent that understands both natural language and your spatial data.

Prerequisite. The GeoJSON file for the active project must have status ready before chat is available. If it isn't, the Dashboard shows a prompt to go process the file.

Asking questions

Type any question in the input box and press Enter or click the send button. The agent:

Detects the intent of your question
Searches your spatial data using hybrid vector + spatial retrieval
Returns a grounded answer with source objects

Example questions that work well:

How many defects are detected?
Where are we losing the most power?
Show me all features in the north zone
What is the total capacity in kilowatts?
Which areas have the most anomalies?
Show NDVI on the map

Source cards

The TerraQuery Agent panel showing an NDVI analysis with percentiles and a zone breakdown, plus suggested follow-up questions — A grounded agent response — statistics, a zone breakdown, and suggested follow-up questions, alongside the map it describes.

Each assistant response includes collapsible source cards — the specific spatial features the answer is grounded in. Each card shows:

Object name and type
Status (operational / defective / etc.)
Area in m²
Primary and secondary metrics
Exact latitude / longitude

Requesting map overlays via chat

You can ask the agent to change the map view directly:

Show NDVI overlay
Calculate SAVI using band 6 and band 8
Use RedEdge and NIR for NDRE
Show ExG index

The agent will trigger a map update and confirm which index and bands are being used.

Specifying custom bands for index calculations

For multispectral imagery where you know the exact band layout, you can tell the agent which band numbers to use:

Calculate NDVI using band 8 and band 4
Use band 6 for NIR and band 2 for Red

The agent passes the band assignments to the tile renderer, producing a visualization based on exactly the bands you specified. The map reloads automatically.

Conversation context

The agent remembers the current conversation session. You can ask follow-up questions naturally:

User:  How many defects are in the south zone?
Agent: There are 14 defects in the south zone...

User:  Which one has the highest power loss?
Agent: The defect with the highest loss is POLYGON_88...