Magic Insight: ask SQL databases questions in natural language

Understand the project

Practical overview

Use Magic Insight when you want a self-hosted conversational interface for SQL data and can provision the documented runtime and model setup.

Teams that want to explore SQL databases through conversational questions.
Organizations seeking a privately or semi-privately deployed ChatBI service.
Developers who want to access the service through its web client or integrate it into another system using the documented REST API.

Ask questions about SQL database data in natural language.
Choose between a general-purpose-model-and-RAG agent mode and a fine-tuned-model mode.
Use a web interface or call the service from another system through its REST API.

Before you choose

The project is open source under AGPL-3.0 and documents self-hosted installation options including pip, Docker and source compilation.

Agent Mode is described as easier to start with, but with lower accuracy and higher inference cost.
Fine-Tuned Model Mode is described as more accurate and less costly to infer with, but requires training-data generation and fine-tuning.
The documented runtime is Ubuntu 22.04, RTX 4090, CUDA 12+ and PyTorch 2.0+; other environments are not rigorously tested.
Docker deployment requires Docker with GPU support enabled.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Magic-BI and start its web interface using the documented source-compilation path.

A Magic-BI source directory.
Docker and GPU support for Docker to start the documented dependencies.
The README lists Ubuntu 22.04, an RTX 4090, CUDA 12+ and PyTorch 2.0+ as its runtime environment.

From the Magic-BI directory, run the command below.
Documented command
```
pip3 install -e .
```
Start the dependencies with the command below.
Documented command
```
docker compose -f deployment/docker-compose.yml up -d
```
Start the additional components with the command below.
Documented command
```
docker compose -f deployment/docker-compose-component.yml up -d
```
Run the command below.
Documented command
```
python3 -m magic_bi.main --config config/system.yml
```
Open http://127.0.0.1:6688 in Chrome or Firefox.

The documented web client is available locally at http://127.0.0.1:6688.

The README identifies Ubuntu 22.04, RTX 4090, CUDA 12+ and PyTorch 2.0+ as its runtime environment; other environments are not rigorously tested.
Docker GPU support setup is not covered in the README.

Evidence and transparency

Source record details

Source availability: Active
Sources checked: 15 October 2026
Original website title: tianlan-ltd/Magic-Insight · Official homepage title · retained in its source language