Overview
Create Agent is a feature that allows you to create domain-specific expert agents from Corvic Spaces created with the Embed feature. These agents can process any connected data, answer questions, and perform analysis on the fly, enabling intelligent interactions with your data through natural language queries. Build agents tailored to your business needs and use cases.Category
Corvic Tables - This feature is designed to work with Corvic Spaces (created from Corvic Tables using Embed), enabling you to create intelligent agents that interact with your embedded data.Input
Corvic Spaces - The Create Agent feature accepts Corvic Spaces created with the Embed feature as input. Select one or more Spaces from your data room that contain the embeddings you want your agent to work with.The input must be Corvic Spaces created using the Embed feature. These spaces contain the vector embeddings that enable the agent to understand and query your data.
- Embed - Generate embeddings from Corvic Tables to create Spaces
- Spaces - Learn about embedding spaces and how they work
Output
Data Analysis Agent - The Create Agent feature produces a data analysis agent that can process any connected data, answer questions, and perform analysis on the fly. The output agent includes:- Intelligent Query Processing: Understands natural language questions and queries
- Data Analysis Capabilities: Performs on-the-fly analysis of connected data
- Multi-Space Access: Can work with multiple Spaces simultaneously for comprehensive insights
- Dynamic Reasoning: Uses LLM-powered reasoning to provide intelligent responses
The output agent serves as an intelligent interface to your data, enabling users to interact with complex datasets through natural language. Agents can traverse multiple Spaces, combine information, and provide comprehensive answers and analysis.
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
agent_name | string | Yes | Name for the agent. Provide a descriptive name that reflects the agent’s purpose and functionality. |
model | string | Yes | LLM model selection. Choose from available models or use “bring your own” if your admin has added custom LLM endpoints. Options include GPT models, Claude, and custom models configured by administrators. |
persona_prompt | string | Yes | Prompt to define the agent’s tone and personality based on the task. Describes how the agent should communicate and behave (e.g., professional, friendly, technical, analytical). |
workflow_prompt | string | Yes | Natural language description of what the agent is supposed to do. Defines the agent’s primary function, tasks, and how it should approach queries and analysis. |
completion_prompt | string | Yes | Description of what the final response should look like. Specifies the format, structure, and style of the agent’s responses (e.g., detailed analysis, concise summary, structured report). |
thinking_mode | selection | Yes | The reasoning mode for the agent. Select from: standard for one-shot planning (faster for simple queries) or reflection for continuous planning with self-correction (advanced reasoning for complex queries). |
Model Selection: Learn how to configure custom LLM endpoints in Bring Your Own LLM Tutorial.
Thinking Modes
Standard Mode
One-shot planning approach that is faster and more efficient for simple queries. The agent makes a single plan and executes it, ideal for straightforward questions and basic analysis tasks.Reflection Mode
Continuous planning with self-correction for advanced reasoning. The agent iteratively refines its approach, making it ideal for complex queries, multi-step analysis, and tasks requiring deep reasoning.Usage Example
To use Create Agent in a Data App:- Ensure you have Corvic Spaces created using the Embed feature
- Add your Corvic Spaces to the Data App canvas
- Click the ”+” button next to one of the Spaces
- Select “Create Agent” from the actions menu
- Provide an agent name
- Select the LLM model to use (or configure “bring your own”)
- Define the persona prompt describing the agent’s tone and personality
- Specify the workflow prompt describing what the agent should do
- Set the completion prompt describing the response format
- Choose the thinking mode (Standard or Reflection)
- Run the Data App to create the agent
- Test the agent with sample queries to verify functionality