ai chat bot

[srijani-smsrc]

Want to build an Agentic AI BI System for my company. There is already SupersetMCP out there. Now, how to integrate a chatbot window into the main page so that people can use it to create visual, query data, analysis data

@dosu

[dosubot[bot]]

This exact question was asked and answered recently [1]. Here's the summary:

Superset supports this through two layers:

1. Frontend — Extensions Registry for the Chat UI

You can inject a custom React chatbot widget into the Superset home page using the Extensions Registry. Key extension points: 'welcome.banner', 'navbar.right', or 'welcome.message' [1].

import { getExtensionsRegistry } from '@superset-ui/core';
import ChatbotWidget from './components/ChatbotWidget';

export default function setupExtensions() {
  const extensionsRegistry = getExtensionsRegistry();
  extensionsRegistry.set('welcome.banner', ChatbotWidget);
}

There's also a newer chatbot extension framework with a dedicated superset.chatbot contribution point that supports floating window and panel modes with proper lifecycle management [2] [3].

2. Backend — MCP (Model Context Protocol) for AI Agent Operations

SupersetMCP is a standalone FastMCP service that exposes Superset functionality (chart creation, SQL querying, dashboard management, etc.) as standardized tools for AI agents [4]. It runs separately (superset mcp run --port 5008) and communicates via JSON-RPC 2.0 while respecting Superset's RBAC/security [5].

Recommended architecture:

1. Frontend: Custom React chat widget registered via Extensions Registry or the chatbot extension framework
2. AI Agent Backend: Your agent (LangChain, CrewAI, etc.) connected to SupersetMCP
3. Communication: Chat widget → your agent API → MCP tools → Superset APIs → results back to UI

There are also deployment and authentication guides for connecting AI clients like Claude Desktop and ChatGPT to Superset's MCP service [6], plus a user guide covering natural language workflows for data exploration and visualization [7].

For natural language SQL specifically, there's also SIP-166 which proposes an AI Assistant for text-to-SQL directly in SQL Lab [8].

To reply, just mention @dosu.

---

Docs are dead. Just use Dosu.