All Products
Search

This Product

    AI-Native Database Service:Get started with AI Data Preparation

    Document Center

    AI-Native Database Service:Get started with AI Data Preparation

    Last Updated:Jul 13, 2026

    Multimodal Data Preparation provides enterprise-grade knowledge base capabilities powered by RAGFlow. It supports intelligent parsing and semantic retrieval across multiple document formats. This topic describes how to create a RAGFlow knowledge base instance, upload documents, and build an enterprise knowledge base for AI applications.

    Prerequisites

    • An Alibaba Cloud account is created and identity verification is complete.

    • Documents that you want to import into the knowledge base are prepared. Supported formats include PDF, Word, Markdown, and TXT.

    Note

    A 15-day free trial is available for first-time RAGFlow knowledge base purchases.

    Procedure

    Step 1: Go to Multimodal Data Preparation

    1. Log on to the AI-Native Database Service console.

    2. In the left-side navigation pane, in the AI Data Preparation section, click Multimodal Data Preparation.

    3. Make sure that the RAGFlow tab is selected at the top of the page, and then select the target region.

    Step 2: Create a RAGFlow knowledge base instance

    1. Click the Create RAGFlow Instance card.

    2. You are redirected to the instance creation page in the Data Transmission Service (DTS) console. Select the vector database type and specifications, confirm the configuration, and then click Create.

    3. Wait until the instance status changes to Running.

    Step 3: Upload documents and create a knowledge base

    1. Click the instance card. You are redirected to the instance details page in the DTS console. Click Log On to Knowledge Base to open the RAGFlow console.

    2. Create a knowledge base and select a parsing method, such as General or Enhanced.

    3. Upload document files. RAGFlow automatically chunks the documents, generates vector embeddings, and builds the index.

    4. Wait until the document parsing is complete and the status shows Completed.

    Step 4: Test knowledge base retrieval

    After document parsing is complete, test the knowledge base by using the chat feature in RAGFlow:

    1. In RAGFlow, create a chat assistant and associate it with the knowledge base that you created.

    2. Enter questions related to the document content and verify the accuracy of the retrieval results.

    What to do next

    • Register the RAGFlow knowledge base as an external knowledge base in Dify to build RAG applications.

    • For more information about data source integrations such as OSS, DingTalk, and Lark, see AI Data Preparation.

    • For more information about Enhanced Parsing and built-in models, see AI Data Preparation billing.