How to Build a Full-Stack RAG Knowledge QA App with Alibaba Cloud Low-Code Platform

Introduction

With the rapid rise of large‑model AI, developers need to transform powerful models into end‑to‑end applications quickly. Building such applications requires reliable database services, robust operations tools, flexible API management, monitoring, security, load balancing, and containerized deployment.

Alibaba Cloud’s ecosystem—RDS, SLS, OSS, Function Compute—and the Bailei model development platform together with the Mobi low‑code platform provide a cloud‑native solution for rapid AI app development.

Low‑Code Platform Capabilities

UI development with rich components and visual designers.

Account and permission management, supporting DingTalk and Enterprise WeChat integration.

Seamless integration with cloud services (RDS, SLS, OSS, etc.) and existing business systems.

Multi‑end publishing for Web, H5, mobile apps, and mini‑programs.

Solution Overview

The case study builds a classic RAG knowledge‑QA system using Bailei and Mobi. The workflow includes knowledge‑base creation, RAG agent configuration, integration into Mobi, UI design, database persistence, and analytics.

RAG Agent Development

In Bailei, create a knowledge base, import documents, and configure retrieval settings to display source references. After testing, publish the RAG agent.

CREATE TABLE conversation (
    id VARCHAR(255) PRIMARY KEY,
    parentId VARCHAR(255),
    createAt BIGINT,
    updateAt BIGINT,
    userId VARCHAR(255),
    role VARCHAR(50),
    content TEXT
);

CREATE TABLE behavior (
    id INT AUTO_INCREMENT PRIMARY KEY,
    conversation_id VARCHAR(255),
    user_id VARCHAR(255),
    behavior INT,
    FOREIGN KEY (conversation_id) REFERENCES conversation(id)
);

Integrating the Agent in Mobi

Create an integration resource of type “Bailei Smart Agent” in Mobi, fill in the API‑KEY from Bailei, and reference the appId in integration operations.

Chat UI Customization

Use Mobi’s visual designer to add a chat component, set its width and height to 100vw/100vh, and bind the component to the integration operation’s prompt field (e.g., chat1.currentMessage.content).

Database Integration (RDS)

Create a MySQL integration resource in Mobi with the RDS connection details (host, port, user, password). Build integration operations for querying conversation history and inserting new messages.

SELECT * FROM conversation ORDER BY createAt LIMIT 10;

INSERT INTO conversation (id, parentId, createAt, updateAt, userId, role, content)
VALUES ({{chat1.context[chat1.context.length-1].id}},
        {{chat1.context[chat1.context.length-1].parentId || ''}},
        {{chat1.context[chat1.context.length-1].createAt}},
        {{chat1.context[chat1.context.length-1].updateAt}},
        {{mobi.currentUser.id}},
        {{chat1.context[chat1.context.length-1].role}},
        {{chat1.context[chat1.context.length-1].content}});

INSERT INTO behavior (user_id, conversation_id, behavior)
VALUES ({{mobi.currentUser.id}}, {{conversation_id}}, {{behavior}});

Persisting Sessions and Feedback

Bind the “Insert Conversation” operation to the chat component’s send and answer‑complete events, and bind “Insert Behavior” to the like/dislike buttons using event handlers.

Analytics Backend

Design a simple admin page with layout, statistic, line‑chart, and table components. Create four integration operations to fetch conversation counts, user counts, daily conversation aggregates, and detailed conversation data with feedback.

Publishing

Publish the app to development, test, and then production environments. Mobi also offers pre‑built templates for quick start and supports multi‑channel deployment (Web, mobile app, mini‑program).

Future Extensions

Omnichannel publishing and embedding.

Fine‑grained permission groups.

Multiple identity sources (Alibaba Cloud account, DingTalk, Enterprise WeChat, custom IdaaS).