Deep Dive Mode
This documentation is for Kyvos 2026.7.x. To view documentation for other Kyvos versions, click here. |
|---|
Important
Deep Dive mode is not supported in Teams, Slack, or the Dialogs API.
Overview
Deep Dive mode enhances Kyvos Dialogs by enabling AI to solve complex analytical questions through a guided, multi-step execution process. Instead of generating a response in a single step, Thinking Mode breaks a query down into logical tasks, executes them in sequence, and displays the execution progress as ordered activities before delivering the final response.
Deep Dive Analysis also supports Thinking Mode, which allows you to control the amount of reasoning the AI performs during query execution. You can configure the Thinking Effort (for example, Low, Medium, or High, depending on your configuration) to balance response quality, reasoning depth, and execution time.
This mode offers greater transparency into how responses are generated, while also improving the AI's ability to answer questions that require information from multiple data sources.
Note
Thinking Mode is available only when Deep Dive Analysis is selected. It is not supported in Quick Analysis mode.
To get started with Deep Dive analysis mode, you must complete the prerequisites, as mentioned below in this section.
To view an architecture of the Deep Dive Analysis mode, click here.
Benefits of Deep Dive mode
This mode provides the following benefits:
Breaks down complex questions into multiple logical steps.
Retrieves information from multiple structured and contextual data sources.
Adapts execution based on intermediate results.
Displays the execution flow as ordered activities for greater transparency.
Provides an auditable execution process for enterprise environments.
When to use it
This mode is useful when queries:
Require multiple reasoning steps.
Need information from multiple semantic models.
Combine structured and contextual knowledge.
Require intermediate analysis before producing the final answer.
Benefit from visibility into the AI's execution process.
How this mode Works
When this mode is enabled, Kyvos Dialogs uses a planning-based execution model to process your query.
The execution process includes the following steps:
Analyzes the user query.
Creates an execution plan consisting of multiple steps.
Invokes the required tools to retrieve or process information.
Displays each execution step as an activity while the query is being processed.
Combines the intermediate results and generates the final response.
This execution model enables the AI to answer complex analytical questions that require reasoning across multiple semantic models and other supported knowledge sources.
Activities
During query execution, this mode displays the actions performed by the AI as Activities.
Activities provide visibility into the execution process by displaying the sequence of operations performed to answer a query. This helps users understand how the final response was generated and improves traceability for complex analytical requests.
Prerequisites for Deep Dive Analysis in Kyvos Dialogs
Complete the following tasks to enable Deep Dive Analysis.
Step | Description |
|---|---|
Step 1 | Install Python 3.12 on your machine. |
Step 2 | Post deployment configurations for Deep Dive Analysis. |
Step 3 | Configure the Kyvos Dialogs Agent Service in Kyvos Manager. |
Step 4 | Enable Deep Dive for the required AI Space and tool configurations and instructions. |
Step 5 | Start using Deep Dive Analysis in Kyvos Dialogs. |
Step 1: Python 3.12
For Cloud deployments: Python 3.12 installation is automated through the Kyvos-provided template, eliminating the need for manual installation steps. For more details, see Step-2 as mentioned below in this section.
For on-premises: You must manually install Python 3.12 on your machine to use Deep Dive Analysis.
Step 2: Post deployment Configurations for Deep Dive Analysis
You must perform the post deployment steps for Deep Dive Analysis to configure Dialogs Agent and Code Execution Agent.
Select your cloud environment:
Step 3: Configure the Kyvos Dialogs Agent Service
After deploying the Python Agents, configure the Kyvos Dialogs Agent Service in Kyvos Manager. This enables Kyvos Dialogs to communicate with the deployed agent service during query execution. For more information, see Configuring Dialogs Agent Service.
Step 4: Enable Deep Dive analysis and tool configurations
Configure the AI Space that will use Deep Dive Analysis. This includes:
Defining tool instructions. You can configure planner, summarizer, and tool-specific instructions.
Step 5: Start Using Deep Dive Analysis
After completing the configuration, Deep Dive Analysis becomes available in Kyvos Dialogs.
Users can:
Select Deep Dive Analysis from the analysis mode.
Configure the Thinking Effort, if enabled.
Submit complex analytical queries.
View execution activities while the query is processed.
Review the final AI-generated response.
For more details, see the Starting Conversational Analytics with Deep Dive section.
Step 4 Should be adding ENABLE_DIALOG_AGENT=yes
ENABLE_CODE_EXECUTOR_AGENT=yes - Set this as yes - in Clients Global . Properties - Rohit Anshik