Getting Started—Using "Query Builder"

Use the dashboard's Query Builder to configure Query Metadata with checkboxes, covering contact-level and per-message settings for internal Q&A

The Query Builder is a graphical way to configure Query Metadata: use checkboxes instead of JSON, ideal for manual management in the dashboard. Two entry points correspond to the two levels on the overview page:

• Contact Editing → Set the knowledge scope for all conversations of a given user (contact level)

• Internal Q&A → Select the knowledge scope before sending each message (message level)

For nested label conditions, or when configuring via API/embed code, use the JSON format instead. Both methods produce equivalent configurations.

Contact-Level Settings

Step 1: Go to the Contacts Section in the Dashboard

Step 2: Edit a Contact and Open the Query Builder

Find the contact you want to configure and click Edit:

On the edit screen, click Query Builder to open the graphical interface:

Step 3: Select the Scope to Grant Access

Check the knowledge bases to grant access to; expand a knowledge base to select only specific documents within it:

To filter by labels, click Show Label Filter to expand the label condition settings:

Label conditions combine with OR (match any label) or AND (must match all labels), filtering documents within the selected knowledge base scope — semantically identical to label_relations in the JSON format.

Internal Q&A: Per-Message Settings

When members use Internal Q&A, they can select the knowledge scope for each message from the right-side panel (i.e., message-level Query Metadata):

You can also use Show Label Filter to set label conditions.

You can adjust the settings before each message is sent. This only affects the content that the current response can reference — message-level settings take priority over contact-level settings (see Resolution Order).

Demo

When CV, SOP, and Questionnaire Survey are checked, the AI can reference all three:

Uncheck CV before sending the next message:

The AI can no longer use the CV file, and CV no longer appears in the reference list: