Working with AI Space

This section details how to use the AI Space in Kyvos to interact with data using natural-language queries and automatically generate insights, visualizations, and analyses.

AI Space in Kyvos enables conversational analytics across multiple data sources within a unified business context. It acts as a container that consolidates relevant entities for semantic models.

By organizing these entities within a single AI Space, Kyvos supports seamless conversational Q&A across multiple semantic models, enriched with external data sources. AI Spaces provide centralized AI settings, including shared business context, tool definitions, and business rules, ensuring accurate query interpretation, consistent SQL generation, and insights beyond simple “what” questions.

To view AI Space, perform the following steps.

From the Toolbox, select AI Space.

The AI Space panel is shown, displaying a list of semantic models.
Drag and drop the semantic model to the Space Objects pane in the right.
In the Properties pane, define the General and Context for the AI Space.
- General: Specify the LLM connection and Embedding connection for metadata, query, and summary generation.
- Context: Specify the business context, querying instructions, summary instructions, verified queries for the AI Space. Once you specify the context details, the query response in Kyvos Dialogs will be generated as per the instructions you set here. Following images displays some sample images for reference.
  - Business Context: Business Context offers insight into the AI Space, helping the AI identify relevant semantic models and produce more accurate responses.
  - Querying Instructions: Provide context to be included with user input for generating queries. This helps guide query generation, refine results, and improve response accuracy.
  - Summary Instructions: Provide context to be included with user input for generating natural language summaries. This helps tailor the summaries to better align with business needs and improve relevance.
    - To generate relevant follow-up questions, specify the instructions that guide Kyvos Dialogs in producing such questions. Add these instructions to the summary instructions box. Write your summary instructions between, as described in sample below:
      <FOLLOW_UP_QUES_RULES>
      Summary instructions
      </FOLLOW_UP_QUES_RULES>
      Sample instructions to create follow up questions
      <FOLLOW_UP_QUES_RULES> Below are the columns and their dependent columns for generating follow-up questions: For Sales column, dependent columns are: Profit, Discount, Product Category For Product Category column, dependent columns are: Product Sub-Category, Product Names For Product Sub-Category column, dependent columns are: Country, State, Shipping Modes, Discount For Shipping Modes column, dependent columns are: Discount, Segment Follow-up Question Generation Rules: Generate only 3 follow-up questions. Each follow-up question must include only one dependent column. Follow-up questions should be context-aware and based on the previous response. Questions should help users drill down into patterns and move toward root cause analysis. Keep questions simple, clear, and easy to understand for end users. Do not generate generic or unrelated questions. Follow-up questions must align with the visualization and data shown. </FOLLOW_UP_QUES_RULES> between <follow up questions> to </follow up questions>. <image>
  - Verified Questions: Verified answers improve future responses and help us better understand your business queries. You can add queries, export and import queries as needed.
Click the Actions menu (…) and do the following:
- Validate the space
- Share the space via link or grant access.
- Refresh the space to get the updates.
- Add Note or Tag for the space.
Click Save. The Space will be added and displayed to the AI Space panel.