Docs

DataAgent (OnPremise)

Intelligent Analysis Agent User Guide

Overview of Intelligent Analysis Agent

The Intelligent Analysis Agent (hereinafter referred to as the Analysis Agent or Intelligent Agent) is an expert advisor based on AI deep reasoning and big data analysis, integrating the deep reasoning capabilities of large models and supporting both flexible intelligent data querying and in-depth research reports. This article provides a brief introduction to the feature overview, applicable scenarios, and general operation process of the Intelligent Analysis Agent.

Feature overview

Intelligent analysis

Intelligent query Agent	In-depth research Agent
As a flexibly configurable intelligent query tool, it supports queries across multiple datasets, leverages semantic model parsing and business knowledge, and accurately understands user requirements. It offers question recommendation and bookmarking, personalized notifications, supports multi-turn interactive querying and attribution analysis, and meets diverse data query needs through simple conversations.	An expert advisor for AI advanced reasoning and big data analysis, built on data querying as the foundation for interaction and integrating advanced reasoning capabilities of large models. It automatically parses natural language requirements, constructs a comprehensive analysis framework, generates SQL and Python code to gain insights from data, outputs in-depth research reports, assists enterprises in efficient decision-making, and drives the intelligent transformation from raw data to business insights.

Intelligent query Agent

In-depth research Agent

As a flexibly configurable intelligent query tool, it supports queries across multiple datasets, leverages semantic model parsing and business knowledge, and accurately understands user requirements.
It offers question recommendation and bookmarking, personalized notifications, supports multi-turn interactive querying and attribution analysis, and meets diverse data query needs through simple conversations.

An expert advisor for AI advanced reasoning and big data analysis, built on data querying as the foundation for interaction and integrating advanced reasoning capabilities of large models.
It automatically parses natural language requirements, constructs a comprehensive analysis framework, generates SQL and Python code to gain insights from data, outputs in-depth research reports, assists enterprises in efficient decision-making, and drives the intelligent transformation from raw data to business insights.

Model integration

Model integration refers to the capability of the Intelligent Analysis Agent to adapt to large language models, including integration with Volcano Ark's cloud-based models as well as private or on-premises models in customer environments.
The currently supported models, along with detailed descriptions of application scenarios and deployment resource requirements, are as follows.

Model types		Supported model names	Supported environments	Application scenarios and deployment resource requirements
Cloud-based models		The intelligent analysis Agent provides a finely tuned combination of multiple models (best performance)	SaaS Private deployment	Can be used for full-chain scenarios such as vectorization and inference. In SaaS environments, models are used directly without deployment; for on-premises environments, deployment is performed by BytePlus technical support personnel. Deployment resource requirements: GPU: NVIDIA T4 or NVIDIA cards with more than 16 GB of memory The machine hardware brand and operating system are the same as those required for on-premises product deployment.
Local models	Inference models	Use intelligent query and in-depth research: DeepSeek-V3.1 (latest version) DeepSeek-V3 (latest version)	Private deployment	Used for mainline scenarios such as code generation and reasoning, requiring strong capabilities to complete complex tasks. Model requirements: support function calls, enable stateless mode, and support a context length of 128k tokens. For deployment in a private environment, the customer is responsible for deployment. Resource recommendations: DeepSeek-V3: H20 * 8 or NVIDIA cards with equivalent memory configuration. Qwen3-235B: H20 * 4 or NVIDIA cards with equivalent memory configuration.
	Inference models	Use only the intelligent query function: DeepSeek-V3.1 (latest version) DeepSeek-V3 (latest version) Qwen-235b (latest version)	Private deployment
	Vectorization model	Customer self-deployed model: BGE-M3	Private deployment	Used for vectorization For deployment in a private environment, the customer is responsible for deployment. Resource recommendations: GPU: NVIDIA T4 or NVIDIA cards with more than 16GB of memory After deployment is completed, provide the URL of the deployed instance to BytePlus deployment personnel.
	Vectorization model	Fine-tuned model provided by the Intelligent Analysis Agent (Customer provides deployment resources, BytePlus provides the model and performs deployment)	Private deployment	Used for vectorization For deployment in a private environment, BytePlus technical support personnel are responsible for deployment. Resource requirements: GPU: NVIDIA T4 or NVIDIA cards with more than 16GB of memory The machine hardware brand, operating system, and resources for private product deployment must be consistent.

Data preparation

The Intelligent Analysis Agent base can connect to external data sources, offline data, public data, and data from the BytePlus data product matrix, allowing data from these sources to be added to the Agent base for data processing and integration, enabling intelligent data querying.
The Intelligent Analysis Agent base comes with ByteHouse, a self-developed high-speed engine, as the query acceleration engine, significantly improving data query and write speeds and delivering an excellent user experience with second-level queries on hundreds of billions of records.
The Intelligent Analysis Agent base also supports data processing, allowing data transformation through visual modeling. This prepares data for subsequent consumption, including data cleaning, transformation, integration, and pre-computation, while also supporting dataset management and maintenance.

Configuration management

The Intelligent Analysis Agent provides a comprehensive set of configuration tools to optimize intelligent data querying, covering semantic models, enterprise knowledge engine configuration, recall logic customization, prompt intervention, recommended question/metric summary management and operation, and more, helping the data querying tool integrate into business scenarios and become a professional assistant for business data querying.

Open integration

The Intelligent Analysis Agent supports integration with Feishu bots, embedding into customer application systems via H5, open integration for secondary development through OpenAPI, and embedding with BytePlus's HiAgent product to enable collaborative applications.

Applicable scenarios

Data integration and efficient analysis to effectively drive e-commerce GMV growth

Data insights drive business strategies, implementation, and forward-looking decision-making

Driven by both qualitative insights and quantitative analysis, the solution empowers intelligent incubation of new products

Caution and limitations

Intelligent query limitations

To achieve better intelligent query results and experience, ensure that your configuration and usage meet the following requirements.

Sub-items	Detailed description of restriction requirements
Number of datasets	The number of datasets associated with and bound to a single agent must not exceed 5.
Total number of fields	The total number of fields across all datasets associated with and bound to a single agent must not exceed 300.
General requirements for intelligent query	Each agent supports only one scenario. Do not mix multiple business scenarios involving data queries and in-depth research within a single agent. Having a large number of scattered business datasets usually results in poor data query and analysis outcomes. Indicators that cannot be processed through ETL are not supported during intelligent query; for indicators that can be processed through ETL, it is recommended to process them in advance. When asking questions using the intelligent analysis Agent, do not include multiple intents in a single interaction. That is, each question should query only one piece of data. Incorrect example: `Which are the top 10 products with the highest profit in 2024?` `What are the profit and sales revenue of these products, and what are their year-over-year and month-over-month changes, respectively?` `Which product has the largest year-over-year increase?` — In this example, multiple sub-intents are included in a single interaction.

In-depth analysis limitations

To achieve better in-depth analysis results and experience, ensure that your configuration and usage meet the following requirements.

tip

When using deep analysis, data querying capabilities are typically involved, so it is also necessary to pay attention to the restrictions described in the above "intelligent query limitations" section.

Subcategories	Detailed restriction requirements
Analysis template character count	The character count of an analysis template must not exceed 2,000.

Usage recommendations and best practices

When using the Intelligent Analysis Agent, you can review the current chart types supported by intelligent analysis and refer to best practices for each chart. For details, see Best Practices: Supported Chart Types and Usage Practices.
To achieve better usage results (query accuracy), you can review the current recommendations and best practices for intelligent queries. For details, see Best Practices: Recommendations and Examples for Intelligent Data Query Questions.

General operation process

Purchase and activate the Intelligent Analysis Agent

Starting from May 30, 2025, new customers can purchase the Intelligent Analysis Agent independently or combine it with DataWind as needed. For existing DataWind customers, the Intelligent Analysis Agent module can be added for a quick upgrade.

Create projects and grant permissions

Users can create projects based on business scenarios and set data access and feature operation permissions for different users or user groups according to internal roles and data security requirements. For details, see Project Information and User Management.

Data integration and processing

Through the data connection feature, the Intelligent Analysis Agent can connect to external data sources, offline data, public data, and data from the BytePlus data product matrix. It also supports data processing through visual modeling. For details, see Data Preparation Overview.

Create and configure the intelligent analysis Agent

Configure the intelligent analysis Agent in areas such as semantic models, business knowledge, recommended questions, fuzzy question decomposition, metric summaries, and prompts to ensure alignment with business requirements, precise understanding of user queries, and effective analysis. For details, refer to Overview of the Creation and Configuration Process.

Use the intelligent analysis Agent

Users initiate data queries or analysis requests in natural language. The intelligent analysis Agent receives the instructions, parses and processes them, provides query results or in-depth analysis reports, and supports multi-turn interactions to refine the analysis. For details, refer to Intelligent Data Query Agent and Deep Research Agent.

Last updated: 2026.07.03 13:52:03

DataAgent (OnPremise)

Intelligent analysis #

Model integration #

Data preparation #

Configuration management #

Open integration #

Data integration and efficient analysis to effectively drive e-commerce GMV growth #

Data insights drive business strategies, implementation, and forward-looking decision-making #

Driven by both qualitative insights and quantitative analysis, the solution empowers intelligent incubation of new products #

Intelligent query limitations #

In-depth analysis limitations #

Usage recommendations and best practices #

Purchase and activate the Intelligent Analysis Agent #

Create projects and grant permissions #

Data integration and processing #

Create and configure the intelligent analysis Agent #

Use the intelligent analysis Agent #