Introduction

wolkvox Studio incorporates the new Deep Research Component, also known as Agentic Deep Research, an agentic artificial intelligence module designed to perform searches, tracking, and exhaustive analysis based on the instructions defined by the flow designer.

Unlike a conventional query, this component can interpret a complex objective, follow specific guidelines, evaluate the collected information, and consolidate the findings into a structured response. This way, it allows incorporating advanced research capabilities within a Routing Point without having to manually build multiple steps for querying, validation, and data organization.

The generated result is automatically stored in the predefined variable $ResDeepR, from where it can be used by other components in the flow to continue automation, make decisions, present information to the user, or execute new actions.

The component is available in the following types of Routing Points:

• Voice Flow
• Chat Flow
• Interaction Flow
• Agentic Engine

The Deep Research Component is not available in Agent Scripting.

Benefits

• Autonomous Contextual Analysis: Executes broad investigations based on detailed guidelines, interpreting the objective and applying the rules defined by the administrator.
• Reduced Design Complexity: Minimizes the need to create multiple connections and conditions to query, validate, organize, and consolidate information within the Routing Point.
• Business-Tailored Guidelines: Allows defining the AI's role, the research objective, expected data, constraints, accepted sources, and response format.
• Persistence Between Investigations: The Enable Session Memory option allows preserving findings from previous executions to maintain context during consecutive investigations.
• Integration with Other Components: The result is stored in $ResDeepR, allowing it to be reused in text components, TTS, conditions, databases, integrations, autonomous agents, or other flow actions.
• Pre-Production Validation: The Test Agent button allows reviewing the component's behavior and adjusting instructions before connecting it to the rest of the process.

What Can the Deep Research Component Be Used For?

The component can support processes that require gathering, evaluating, and organizing information from different sources. Some examples include:

• Prospecting potential companies or customers based on commercial criteria.
• Researching markets, sectors, competitors, or trends.
• Gathering public information about organizations and contacts.
• Comparing products, services, prices, or conditions.
• Preparing executive summaries for agents or internal departments.
• Analyzing backgrounds before sales, support, or collection management.
• Identifying opportunities and risks based on business-defined rules.
• Obtaining structured information to store in a CRM or database.
• Preparing context for another autonomous agent to continue the process.
• Validating whether the found information meets mandatory requirements before proceeding in the flow.

For example, the designer can configure the component as a commercial prospecting analyst and ask it to identify restaurants in a city, provide their contact details, explain why they could be potential customers, and attach the exact source used.

How to Configure the Deep Research Component

1. In the component sidebar, go to the "Cognitive" tab and look for the "Agents" section.
2. Drag the component to the part of the flow where it is needed.
3. Left-click the component to open the configuration panel on the right side.

In the "Name" tab, enter a clear identifier for the component. The name should make it easy to locate within the flow, especially when using multiple agents or research processes. Some examples include:

• Restaurant Prospecting
• Customer Research
• Competitor Analysis
• Supplier Validation

Switch to the "Description" tab and enter additional information about the component's purpose. This field can be used to document what it investigates, where in the process it executes, or what result is expected.

In the "Research Guidelines" field, write the general context that will guide the agent's behavior. These instructions may include:

• The role it should assume.
• The general objective.
• The fields it should return.
• Mandatory rules.
• Accepted sources.
• Criteria for discarding results.
• Maximum number of records.
• Language and response format.

Example:

# Role
You are a commercial prospecting analyst specializing in identifying potential companies.

# Objective
Identify restaurants that could be customers for natural beverage distribution.

# Required Fields
- Company
- City
- Phone
- Email
- Address
- Notes
- Source

# Rules
- Return a maximum of 20 results.
- Do not invent missing information.
- Prioritize official sources.
- Avoid duplicate records.
- Respond only in Spanish.

The guidelines function as the permanent framework for the research and must be clear enough to establish how the agent should reason and respond.

In the "Research Request" field, specify the concrete task to be executed. For example:

Search for commercial opportunities for restaurants located in Medellín.

While the guidelines establish the agent's general behavior, the request indicates what should be investigated in that specific execution.

Click on "Test Agent" to run a validation of the configuration.

The system processes the entered guidelines and request and generates a local HTML file with the test result. This file automatically opens in your default web browser once the test finishes loading. The path may have a structure similar to the following:

C:/ipdialbox/download/wolkvox_deepresearch_test_20260723_165400.html

Inside the file, you can check:

• The guidelines or prompt used.
• The research request.
• The generated result.
• The session memory status.
• The date and time of execution.
• The identifier of the tested component.

This test allows you to review whether the agent correctly interprets the rules, finds the expected information, and whether the output format is appropriate.

Enable the "Enable Session Memory" checkbox when you need consecutive investigations to retain context and previous findings.

• Enabled: The component can use information obtained in previous executions of the same session.
• Disabled: Each investigation starts without considering previous results.

This option is useful for progressive investigations, refinements, or cycles where each new request depends on what was previously found.

The "Research Result Variable" field displays the predefined variable:

$ResDeepR

This variable stores the complete response produced by the component.

The variable name is fixed and cannot be modified.

You can use $ResDeepR in the following flow components to:

• Display the result to the user.
• Convert it into audio using TTS.
• Evaluate it in a condition.
• Save it in a database.
• Send it to an external system.
• Deliver it as context to another agent.
• Continue automation based on the findings.

Click on "Save" to apply the component's configuration.