DataAgent (OnPremise)
DataAgent (OnPremise)
- Documentation
DataAgent (OnPremise)Intelligent Marketing Agent User GuideEnterprise Knowledge EngineOverview of the Enterprise Knowledge Engine
Copy page
Download PDFCopy pageDownload PDFEnterprise Knowledge Engine serves as the foundational capability of the Marketing Interaction Assistant. The Enterprise Knowledge Engine provides comprehensive knowledge construction, management capabilities, and intelligent AI application capabilities, enabling enterprise users to build a general-purpose enterprise knowledge engine, customize AI applications, and ensure data security through refined Authority Management. The Enterprise Knowledge Engine is suitable for various scenarios such as internal enterprise knowledge management, intelligent customer service, and intelligent Q&A. This document introduces the functional overview of the Enterprise Knowledge Engine.
- Intelligent Marketing Agent – Enterprise Knowledge Engine (hereinafter referred to as the “Enterprise Knowledge Engine”) is built upon BytePlus’s proprietary large model and big data technology system. It is designed for various enterprise intelligent Agent application scenarios, providing structured + unstructured multimodal knowledge with end-to-end knowledge management and application capabilities, covering integration, processing, modeling, mining, and application. It broadly supports business scenarios such as intelligent knowledge Q&A, product recommendation marketing, and knowledge summarization and analysis, helping enterprises deeply mine and transform the value of their knowledge assets in the era of large models.
- The product offers more than 40 knowledge integration methods, links high-quality Datasets to supplement third-party knowledge, supports multi-type knowledge base management, visualized knowledge processing, and diversified knowledge slicing strategies as foundational capabilities. It also features core capabilities such as GraphRAG knowledge graph, ontology modeling, and refined Authority Management. Multiple sales models are supported, with optional value-added features available for purchase as needed. It can be combined with Intelligent Analysis Agent, Intelligent Marketing Agent, and User Research Agent. At the large model layer, it is compatible with the full series of BytePlus Ark models and privatized open-source models. Special privatized models can be evaluated or customized for integration as required.
Key Features
Whether it is an internal company database, local Documents, cloud Documents, Web Pages, or industry public data, all can be integrated uniformly. Furthermore, we support various content formats including text, Images, Audio, and Videos, enabling unified processing without the need for separate management.
Equipped with multiple intelligent tools, the system can automatically parse text, Documents, Audio, Videos, and Images. It can also clean messy data, identify key information, translate multiple languages, and convert speech to text. Ultimately, it transforms incomprehensible and unusable 'raw content' into 'usable knowledge' that can be recognized and invoked by intelligent systems.
Supports the creation of different types of knowledge bases (such as general knowledge, Q&A repositories, and specialized terminology databases). It can flexibly split long texts based on content context and structural characteristics, facilitating subsequent retrieval and usage, and organizes scattered knowledge in an orderly manner for easy storage and access.
Build knowledge association graphs, GraphRAG graphs, and ontology repositories to connect knowledge dispersed across different documents and knowledge bases. Clarify relationships between knowledge, enable cross-document logical reasoning, and help you quickly find precise and comprehensive information.
By integrating intelligent retrieval and large model capabilities, the system enables multi-turn conversations with users, summarizes knowledge, provides precise answers, and can adjust Q&A strategies according to requirements. It can also be encapsulated as intelligent assistants for various scenarios (such as Intelligent Analysis Agent, Marketing Strategy Assistant, Marketing Interaction Assistant) for direct deployment and use.
Provides standardized interfaces for easy embedding into existing business systems. Whether it is HiAgent, Coze, Dify, or other ready-made Agent platforms, or a company’s self-built Agent system, all can be adapted and connected. Fine-grained permission and security governance capabilities are also supported.
Product Architecture Diagram
Core Advantages
The functional advantages of the enterprise knowledge engine are mainly reflected in 'comprehensive data, deep understanding, precise inquiry, broad applicability, and robust governance.' This solution integrates multi-source and multi-modal data, utilizes large models and knowledge graphs to deeply model enterprise knowledge, and, through open and controllable Q&A and Agent capabilities, rapidly delivers intelligent knowledge applications that are practical, scalable, and governable across diverse industry scenarios.
- [Extensive Coverage: Multi-source and Multi-modal Universal Access]Supports access to 40+ data sources and multi-modal content, enabling unified management of internal data, external knowledge, and public domain data, achieving 'one-time access, reuse across all scenarios.'
- [Deep Understanding: Strong AI Processing and Knowledge Modeling]Equipped with 50+ data processing operators and visual preprocessing capabilities, supports flexible slicing, GraphRAG, knowledge graphs, ontology modeling, and multi-hop reasoning, making knowledge interrelated, discoverable, and inferable.
- [Precise Inquiry: Configurable and Optimizable Intelligent Q&A]Supports knowledge classification management, recall testing, multi-turn dialogue, context memory, and Q&A strategy optimization, significantly reducing hallucinations and enhancing answer accuracy and business relevance.
- [Broad Applicability: Open and Easy Integration, Compatible with Various Agents]Provides full-chain API and plugin capabilities, enabling rapid integration with enterprise systems and various Agent platforms, achieving permission isolation and allowing multiple Agents to share the same knowledge base, thus avoiding redundant construction.
- [Robust Governance: Enterprise-level Security and Permission Control]Supports document-level, field-level, and row-level permissions, version management, lifecycle, operation auditing, and data masking, meeting enterprise compliance, confidentiality, and security management requirements.
Application Scenario
- Enterprise Intelligent Knowledge Center: Centralizes regulations, procedures, R&D documents, FAQs, etc., enabling employees to quickly access information through intelligent Q&A.
- Intelligent Customer Service and Service Assistant: Performs Q&A based on product manuals, work orders, maintenance documents, etc., for customer service or frontline support.
- Industry Knowledge and Business Assistant: For example, automotive maintenance manual Q&A, retail product guidance and recommendation, medical/policy/regulation Q&A, industry research report center.
- Internal Analysis and Decision Support: By integrating business data and documents, supports data insights, marketing strategy generation, user research analysis, and other agent-based applications.
Customer Cases
- Financial Industry Case: Knowledge Graph Mining of Potential Relationships
- A Leading B2B Enterprise Case: Knowledge Q&A and Knowledge Mining
- A Leading Airline Case: Intelligent Vehicle Book
Knowledge Base
The enterprise knowledge engine requires the construction of a corresponding knowledge base. For more information about the knowledge base, please refer to Create a New Knowledge Base.
Knowledge Types
The enterprise knowledge engine currently supports the following types of knowledge bases:
Type | General Content Type Knowledge Base | Query Cache Type Knowledge Base | QA Answer Type Knowledge Base | Terminology Type Knowledge Base | Customer Data Type Knowledge Base |
|---|---|---|---|---|---|
Application Scenario | Suitable for enterprise-wide knowledge management, new employee training, and various intelligent Q&A scenarios. | Supports maintenance of common queries for intelligent agents. Common queries will cache COT, improving the efficiency of responses to frequently asked questions. Suitable for scenarios with explicit queries. | Applicable to scenarios with clear questions and answers. If users ask such questions, answers can be directly retrieved. | Suitable for scenarios where users need to maintain industry-specific terminology independently. | Mainly used for customer service/sales guidance and conversation analysis scenarios. Only supports knowledge imported from the Customer Data Platform. |
tip
- Query Cache Type Knowledge Base is a value-added option and requires additional payment.
- Query Cache Type Knowledge Base, QA Answer Type Knowledge Base, and Terminology Type Knowledge Base are supported starting from VeCDP privatized version V1.28.
Comparison Description
comparison term | General Content Type Knowledge Base | Query Cache Type Knowledge Base | QA Answer Type Knowledge Base | Terminology Type Knowledge Base | Customer Data Type Knowledge Base |
|---|---|---|---|---|---|
Knowledge Import Methods |
| Supports xlsx, csv and other format documents, with each document not exceeding 100 MB. | Supports Document formats such as xlsx and csv, with each Document not exceeding 100M. | Supports Document formats such as xlsx and csv, with each Document not exceeding 100M. | Imports data such as Label, Subject attributes, behavioral events, business details, and conversation archives from VeCDP (Customer Data Platform). |
Imported Data Storage | The imported data will be stored in the Marketing Interaction Assistant. | The imported data will be stored in the Marketing Interaction Assistant. | The imported data will be stored in the Marketing Interaction Assistant. | The imported data will be stored in the Marketing Interaction Assistant. |
|
Notes | none | none | none | none | The BytePlus VeCDP service must be activated under the same account, and the user performing the import operation must have view permissions for the VeCDP data to be imported. |
Knowledge Parsing Model Capability | The Marketing Interaction Assistant provides the Doubao-embedding model for knowledge parsing (for example, vectorizing knowledge). If you are using a private environment, you may also choose to connect other models for knowledge parsing. tip When using the enterprise knowledge engine in various application scenarios, you can select supported inference models as needed. | Relies on VeCDP data capabilities and combines large models for analysis and application. | |||
Enterprise Knowledge Engine Management | Provides Document vectorized storage, intelligent retrieval, and knowledge maintenance functions. | ||||
System Monitoring | Tracks the number of AI applications associated with the enterprise knowledge engine, content usage effectiveness (such as recall count), and model performance metrics (call volume, token consumption, response speed), providing data support for resource optimization and troubleshooting abnormalities. | ||||
Authorization | Provides a multi-level account system and resource access control mechanism | ||||
Knowledge Modeling
Subject Modeling
Subject modeling combines the knowledge base and knowledge graph to automatically extract Subjects (such as people, vehicles, stores, etc.) and their related entity attributes (such as ID card number, geographic location, etc.) from original documents, and builds a Subject library. Applicable to scenarios such as multi-scene knowledge classification within the same document, aggregation of similar knowledge across multiple documents, and construction of multiple Subject libraries for enterprises. It significantly improves Data Management efficiency, enhances data support capabilities for upper-layer applications (such as AI Q&A, marketing Push), and meets various complex business requirements. This article introduces how to construct a Subject library. For more information, please refer to (不发布)Subject Modeling.
tip
This feature is supported starting from VeCDP Private Deployment V1.29
Subject Knowledge Graph
In the knowledge base, the Subject Knowledge Graph visually displays logical relationships between knowledge (such as conceptual hierarchy, attribute association, causal links) to achieve knowledge structuring and deep mining. It is suitable for scenarios requiring clarification of complex knowledge associations and efficient location of related information, especially for cross-domain knowledge integration, complex problem traceability analysis, and intelligent recommendation of related knowledge. This article introduces how to create a Subject Knowledge Graph in the enterprise knowledge engine. For more information about knowledge graphs, please refer to Knowledge Graph.
AI Applications
After creating a knowledge base, you can quickly build various AI conversational assistant applications with zero code, such as customer service applications and shopping guide applications. The enterprise knowledge engine offers two delivery modes for different scenarios: API and plugin development platform collaboration mode, and complete platform delivery mode. These are suitable for users with different development backgrounds and business status. You can choose based on your own business needs. Additionally, the enterprise knowledge engine provides a wide range of API application templates, such as customer 360 profiles and script suggestions. For more usage scenarios and operations related to AI applications, please refer to User Case: Customer Service / Sales Scenarios and User Case: Knowledge Q&A Scenario.
Project Center
Includes modules such as Authority Management, system monitoring, and project information, used to manage user permissions, knowledge permissions, and view the status of assets and models. For more information, please refer to Authorization Management.
Customer Pain Points | Specific Issues | Our Proposed Solutions |
|---|---|---|
Difficulties in Multimodal Knowledge Management | Agent applications lack a 'unified knowledge base foundation', with each Agent platform integrating only a portion of knowledge. Knowledge is scattered and cannot be centrally managed. | “Enterprise-level Unified Knowledge Engine Platform”
|
Difficulties in Knowledge Processing | With numerous internal departments and scattered, complex knowledge, how to centrally manage internal knowledge assets, improve employee efficiency, and ensure refined and flexible Authority Management for each department. | “Enterprise Internal Knowledge Management Platform”
|
Poor Knowledge Mining Capability | It is difficult to extract value from unstructured data. With the rapid development of large model technologies, there is a desire to unlock the value of unstructured data, such as extracting content from Image and Video. After extraction, the data can be processed and provided to data annotation platforms and model training platforms; it can also be used directly for knowledge Q&A, collectively supporting all aspects of enterprise production, marketing, and operations. | “Enterprise Knowledge Middle Platform”
|
Knowledge security is difficult to guarantee. | Due to the lack of Authority Management for data, sensitive information faces the risk of leakage. Version disorder in the knowledge base can adversely affect business continuity. | “Enterprise Internal Knowledge Authority Management”
|
Insufficient high-quality content. | Low internal content quality makes it difficult for business departments to effectively query data, such as lacking industry data or high-quality datasets for knowledge Q&A or model training. | 「High-quality Datasets + Knowledge Q&A」
|
The relevant operations for the enterprise knowledge engine are as follows:
- Knowledge Access:
- Create a knowledge base. For detailed operations, please refer to Create a New Knowledge Base.
- Import knowledge content. For detailed operations, please refer to Import Knowledge.
- Knowledge Processing:
- Set labels, validity period, and other operations for knowledge content. For detailed operations, please refer to Manage Knowledge.
- Knowledge Recall and Debugging:
- Enter keywords to recall paragraphs, evaluate recall accuracy based on various metrics. For detailed operations, please refer to Manage Knowledge.
- Subject modeling combines the knowledge base and knowledge graph to automatically extract subjects and their related entity attributes from original documents, and builds a subject repository to improve the accuracy of knowledge retrieval. For detailed operations, please refer to (不发布)Subject Modeling.
- Integrate knowledge graphs with vectorization technology to achieve enhanced retrieval and generation capabilities based on graph structures. For detailed operations, please refer to Knowledge Graph.
- Knowledge Application:
- Build AI Agents with zero code. For detailed operations, please refer to User Case: Customer Service / Sales Scenarios and User Case: Knowledge Q&A Scenario.
- Project Management:
- Set user permissions for modules such as knowledge base and functions. For detailed operations, please refer to Authorization Management.
- View the usage status of models and assets. For detailed operations, please refer to System Monitoring.
- Integrate with the workspace in HiAgent. For detailed instructions, please refer to Connect with HiAgent.
- All content output by this service is generated by artificial intelligence models.
- Intended users of this feature: only for customers' internal use, proprietary operations, and personal use. Customers are prohibited from providing generative AI services to the public via data intelligent agents.
- This product supports both SaaS and private deployment versions. To purchase or request a trial, you can log in to Intelligent Marketing Agent and contact the sales team for consultation.
