RAGFlow is an open source enterprise-grade RAG (Retrieval-Augmented Generation) engine. AI-Native Database Service integrates RAGFlow into the Agent Developer Platform to help you build RAG applications based on private knowledge bases.
What is RAGFlow
RAGFlow is an open source RAG engine built on deep document understanding. It intelligently parses, chunks, and vectorizes unstructured documents in various formats and provides reliable knowledge retrieval capabilities for large language models (LLMs). This reduces model hallucinations and improves response accuracy.
AI-Native Database Service integrates RAGFlow into the Custom Development module of the Agent Developer Platform and provides a one-stop enterprise-grade RAG development experience. You can create and manage RAGFlow instances in the AI-Native Database Service console without building or maintaining the RAGFlow runtime environment on your own.
This topic describes how to use RAGFlow as a development tool in the Agent Developer Platform. To create a RAGFlow knowledge base instance for knowledge data storage and retrieval, see the Multimodal Data Preparation documentation.
Core capabilities
RAGFlow provides the following core capabilities:
|
Capability |
Description |
|
Deep document parsing |
Supports intelligent parsing of various document formats including PDF, Word, Excel, PPT, Markdown, and HTML. RAGFlow recognizes complex structures such as tables, images, and formulas to ensure complete and accurate knowledge extraction. |
|
Intelligent chunking |
Performs intelligent chunking based on document structure and semantic understanding. This preserves contextual information and improves retrieval hit rates and recall quality. |
|
Hybrid retrieval |
Combines vector retrieval with keyword retrieval. By using multi-path recall and reranking strategies, RAGFlow delivers more precise knowledge retrieval results. |
|
Q&A and conversation |
Includes a built-in conversation engine that supports multi-turn conversations based on knowledge bases. You can trace the source of each answer to verify response reliability. |
Benefits of platform integration
Compared with self-deploying open source RAGFlow, using RAGFlow in AI-Native Database Service offers the following benefits:
-
Maintenance-free deployment: Create RAGFlow instances from the console without configuring servers, databases, or dependencies on your own. This reduces deployment and operations and maintenance (O&M) costs.
-
Proximity data access: RAGFlow instances can directly access database instances in AI-Native Database Service. This reduces data transfer between multiple systems.
-
Seamless integration with Dify: You can register RAGFlow instances as external knowledge bases for Dify. This way, Dify applications can call the knowledge retrieval capabilities of RAGFlow to build a complete RAG application pipeline.
-
Elastic resource management: Create and release RAGFlow instances on demand. Scale resource configurations based on your business requirements.
Workflow
The typical workflow for using RAGFlow in the Agent Developer Platform is as follows:
-
Create a RAGFlow instance: Create a RAGFlow instance in the console. The system automatically completes environment deployment.
-
Build a knowledge base: Log on to the RAGFlow instance. Upload business documents and complete parsing, chunking, and vectorization to build a proprietary knowledge base.
-
Develop RAG applications: Configure retrieval strategies and conversation parameters based on the knowledge base that you built. Develop RAG applications such as knowledge-based Q&A or intelligent customer service.
-
Integrate with Dify (optional): Register the RAGFlow knowledge base as an external knowledge base for Dify. Then, use the retrieval capabilities of RAGFlow in Dify applications.
Access RAGFlow
To access RAGFlow, perform the following steps:
-
Log on to the AI-Native Database Service console.
-
In the left-side navigation pane, choose Agent Developer Platform > Custom Development.
-
On the Custom Development page, click the RAGFlow tab to go to the RAGFlow management page.
On the RAGFlow management page, you can create instances, view instance statuses, log on to instances to manage knowledge bases, and develop applications.
Integration with Dify
RAGFlow supports integration with Dify through standard API interfaces. You can register knowledge bases that are built in RAGFlow as external knowledge bases for Dify. This way, Dify applications can automatically retrieve knowledge content from RAGFlow during conversations.
The integration process is as follows:
-
Create a knowledge base in a RAGFlow instance and upload and parse documents.
-
Obtain the API address and AccessKey of the RAGFlow instance.
-
In Dify, configure an external knowledge base connection by specifying the API address and AccessKey of the RAGFlow instance.
-
In the knowledge base configuration of your Dify application, select the RAGFlow external knowledge base that you registered.