English · 简体中文
Ask your data. Get SQL, charts, and business insights.
A conversational workspace for business analysis, from spreadsheets to SQL databases.
Sponsors · Demo · Quick start · Capabilities · Installation · Website · Releases
Discord ·
Telegram · QQ: 991636855
Thank you to the sponsors supporting this project!
Thank you to DolOffer for supporting this project! DolOffer is a platform focused on digital product recommendations and deal sharing, helping users quickly discover useful tools, services, and limited-time offers. The platform provides popular subscription services such as YouTube Premium, Claude, ChatGPT Plus, Spotify, Apple Music, and more, with prices as low as 30% of the official price or even lower. All services are genuine, stable, and backed by reliable after-sales support. Register through our exclusive link and use the coupon code AI8888 when topping up to enjoy an additional 10% discount.
Mozhenzhen — a consumer anti-pitfall guide for developers and AI enthusiasts. Whether you are looking for real-time API price comparison, latency benchmarking and authenticity verification (no watered-down or fake models), comparing after-sales warranties, risk prevention and hidden traps among account / subscription top-up providers, or even discovering free community relays and backup nodes, Mozhenzhen uses transparent data to help you avoid pitfalls and spend every cent wisely.
Thank you to Infistar.cc for sponsoring Data-Analysis-Agent! It provides stable and efficient multi-model API services that fit perfectly into data cleaning, code generation, chart plotting and automated analysis report workflows.
⚙️ Compatible integration: select the OpenAI-compatible interface in the project, configure the Base URL and API Key, and you are ready to go.
🧩 Multi-model support: brings together top models such as Claude, GPT, Gemini and DeepSeek, making it easy to switch between them and compare analysis and code generation results.
🏷️ Pricing & service: selected models from as low as 1% of official pricing, with public and transparent multipliers, RMB settlement, invoicing and model verification.
🎁 Exclusive perk: sign up via our exclusive promo link and get $5 in free credits!
👉 Register on Infistar · Claim $5 credits
Illustrated workflow with fictional sales data; not a live screen recording.
| Capability | What you can do |
|---|---|
| Ask in plain language | Turn a business question into SQL, a chart, and an explanation. |
| Bring your data | Upload Excel / CSV files or connect a SQL database. |
| Follow the analysis | Watch schema discovery, queries, and results stream over SSE. |
| Visualize and export | Choose from 41 chart types and export Excel, Word, or PowerPoint deliverables. |
| Choose your model | Connect OpenAI-compatible APIs or discover compatible local services. |
| Extend the workflow | Add business knowledge, analysis Skills, MCP tools, or Feishu Bitable integration. |
Requires Python 3.10+. Run in a dedicated virtual environment:
git clone https://github.com/Zafer-Liu/Data-Analysis-Agent.git
cd Data-Analysis-Agent
python -m venv .venvActivate the environment:
# Windows PowerShell
.\.venv\Scripts\Activate.ps1# macOS / Linux
source .venv/bin/activatepython -m pip install -r requirements.txt
python app.pyOpen localhost:5001, configure your model API key, upload Excel / CSV or connect a database, and ask:
Show monthly sales and plot the trend.
Running the source directly requires no cloud login. See installation options and model configuration for more details.
Optional machine-learning, deep-learning, packaging, and test dependencies live in requirements/. See the product specification and frontend build guide for development details.
Table of contents
No SQL required — just type in plain language:
Monthly order volume trend for this year
The system will automatically:
- Generate SQL
- Query the data
- Recommend a chart
- Summarize the analysis
Supports uploading and connecting to multiple data sources:
- Files: Excel / CSV
- Databases: SQLite, MySQL, PostgreSQL, SQL Server
- Planned: DuckDB, Spark
Based on your query results, the system automatically recommends the most appropriate chart from 6 major categories:
| Category | Chart Types |
|---|---|
| COMPARING | Marimekko_ABS, Marimekko_PCT, Bar_Chart, Grouped_Bar_Chart, Stacked_Bar_Chart, Diverging_Bar_Chart, Dot_Plot, Waffle, Bullet_Chart, Sankey_Chart, Heatmap, Waterfall |
| TIME | Line_Chart, Circular_Line_Chart, Slope_Chart, Sparkline, Bump_Chart, Cycle_Chart, Area_Chart, Stacked_Area_Chart, Horizon_Chart, Connected_Scatter |
| DISTRIBUTION | Histogram_Pareto_chart, Pyramid_Chart, Error_Bar_Chart, Box-and-Whisker_Plot, Violin_Chart, Ridgeline_Plot, Beeswarm_Plot, stem_leaf |
| GEOSPATIAL | Flow_Map, Dot_Density_Map, Choropleth_Map |
| RELATIONSHIP | Scatter_Plot, Bubble_Plot, Radar_Charts, Chord_Diagram, Arc_Chart, Network_Diagram, Parallel_Coordinates_Plot |
| PART-TO-WHOLE | Treemap, Sunburst_Diagram, Nightingale_Chart, Pie_Chart |
The analysis process is visible in real time:
[1/4] Reading data structure...
[2/4] Generating SQL...
[3/4] Executing query...
[4/4] Generating chart and insights...
More transparent and interactive than traditional BI tools.
Supports the following model providers:
- DeepSeek
- OpenAI
- AtlasCloud
- Any OpenAI SDK-compatible API
Supports custom base_url, model, and api_key. Default configurations:
| Provider | Default Model |
|---|---|
| DeepSeek | deepseek-v4-flash |
| OpenAI | gpt-4o-mini |
| AtlasCloud | deepseek-v4-pro |
Currently supported analysis features:
- Outlier handling (trimming and winsorizing)
- Decile grouping analysis
- K-Means clustering
- Decision tree modeling
- And more...
Supports exporting:
- Formatted Excel spreadsheets
- Reports in
.docxformat - Built-in styled PowerPoint presentations
Supports connecting to local or remote MCP servers to expand the Agent's capabilities.
- Tutorial: MCP_tutorial
Supports uploading business knowledge to help the Agent better understand your data.
- Tutorial: repository_tutorial
Download the latest version from GitHub Releases:
| Platform | File |
|---|---|
| Windows | Latest Windows .exe installer |
Prerequisites: Python 3.10+ and 64-bit Windows 10 / 11.
Run the installer and follow the setup wizard. After installation, launch Business Analytics Agent from the desktop or Start Menu.
Prerequisite: Python 3.10+ Don't have it? Download here (Windows: check "Add Python to PATH" during install)
Step 1: Download and extract
Step 2: Double-click to launch
| OS | Action |
|---|---|
| Windows | Double-click start.bat |
| macOS | ① Open Terminal (Command + Space → type Terminal → Enter) ② Run the following (replace the path with your actual extraction location): chmod +x ~/Downloads/Data-Analysis-Agent/start.command then xattr -d com.apple.quarantine ~/Downloads/Data-Analysis-Agent/start.command ③ Double-click start.command |
First launch will automatically create a virtual environment and install dependencies — this takes about 3–5 minutes. Subsequent launches are instant.
Step 3: Browser auto-opens at http://localhost:5001
Step 4: Configure your API Key
Step 5: Future updates
Windows (run in PowerShell):
iwr -useb https://raw.githubusercontent.com/Zafer-Liu/Data-Analysis-Agent/main/install.ps1 | iexAfter installation, double-click data-analysis-agent.bat on your desktop, or run:
cd $env:USERPROFILE\.data-analysis-agent\Data-Analysis-Agent
.\.venv\Scripts\activate
python app.pymacOS / Linux (run in Terminal):
curl -fsSL https://raw.githubusercontent.com/Zafer-Liu/Data-Analysis-Agent/main/install.sh | shAfter installation, run:
data-analysis-agentIf you see command not found, add the following to ~/.zshrc or ~/.bashrc, then restart Terminal:
export PATH="$HOME/.local/bin:$PATH"git clone https://github.com/Zafer-Liu/Data-Analysis-Agent.git
cd Data-Analysis-Agent
pip install -r requirements.txt
python app.pyOpen http://localhost:5001 in your browser and configure your API Key.
| Command | Status | Description |
|---|---|---|
/chart |
✅ | Force chart generation as the priority output |
/sql |
✅ | Execute SQL directly |
/analyze |
✅ | Deep statistical analysis |
/tree |
✅ | Decision tree analysis |
/kmeans |
✅ | K-Means clustering analysis |
/data |
✅ | Data exploration and preview |
/inset |
✅ | Missing value imputation |
/winsorize |
✅ | Winsorizing (replace extreme values) |
/trimming |
✅ | Trimming (remove extreme values) |
/export |
✅ | Export data file |
/report |
✅ | Export Word/PDF report |
/ppt |
✅ | Export PowerPoint presentation |
/status |
✅ | View task status |
User input:
Sales trend for the last 12 months
System output:
- SQL query
- Trend line chart
- Sales growth analysis
User input:
Which region has the highest profit?
System output:
- Regional profit rankings
- Bar chart
- Regional business insights
User input:
/chart User growth overview
The system will prioritize generating a visualization.
In the sidebar ⚙, fill in:
API Key
Base URL
Model
to switch between models.
Business Analytics Agent can run independently or with Agent Manager for a more convenient desktop management experience.
Learn how Agent Manager manages this project
Agent Manager is a unified control center for local AI Agent projects. After adding Business Analytics Agent, you can:
- Start or stop the Agent with one click
- View logs and port status in real time
- Open the web analysis interface inside the desktop app
- Ask the Manager Agent to start or open this project in natural language
- Create a temporary public sharing link for demos
Example:
Start Business Analytics Agent and open its interface.
This is especially useful when you frequently demonstrate, debug, or run multiple Agent projects.
👉 Project: Agent Manager
Current source version
v1.4.0· October 16, 2026 · Check Releases for published installers.
The public source now includes the existing private-edition features: the read-only administration console, desktop cloud authentication and model gateways, referral and quota ledgers, browser/table ingestion, browser-local persistence, PyTorch GRU/LSTM, chart themes and PNG export, in-memory SQL transformations and numeric provenance checks, canvas save recovery, Skill tags, and SSE reliability improvements.
Running python app.py directly still requires no cloud login. Hosted and desktop services require their corresponding configuration. See the sync inventory and hosted configuration.
Manage HTTP and Docker connections in Settings → Local services. Custom HTTP connections support health checks and manual JSON requests; OpenAI-compatible services can discover models and add them to the LLM list. TimesFM is a forecasting adapter with run_analysis, backtesting, result tables, and chart export. See the local services guide for configuration and capability boundaries. No models, model dependencies, or container images are downloaded automatically.
For existing model directories, explicit downloads, Docker templates, and a three-product example, see the TimesFM guide. The main application environment stays lightweight; TimesFM weights are limited to non-commercial, non-production research.
v1.3.0 enters the Feishu AI Skills Hackathon, adds a Feishu bot and built-in Bitable analysis, and delivers targeted feature improvements:
- Feishu bot and Bitable loop: Mention the bot in a group to read a Feishu Bitable, load records into DuckDB / SQL analysis, and reply in the same group. Creating result tables or changing records always requires an explicit request.
- More dependable long-term memory: Fixes extraction failures caused by thinking-model output, JSON formatting, and scope detection; adds result notices, 24-hour consolidation, safer session restore, and a global memory switch.
- Trustworthy knowledge-retrieval status: Repairs local Embedding model download/compatibility and makes cloud retrieval status an explicit real check. The former Hash mode is now labelled Basic in the UI.
- Clearer settings and reading experience: Unifies settings headers, cards, and spacing; fixes LLM provider-row inset, hover, spacing, and long-content clipping; wide chat layouts now use available space more effectively.
- Compatibility and delivery reliability: Clarifies SQL Server connection/error boundaries and workflow observability. GPU compute is labelled Planned until end-to-end validation is complete; existing implementation and configuration remain preserved.
📦 Installation & Startup
Network timeout while installing dependencies?
The script will automatically switch to the Tsinghua mirror and retry.
If it still fails, run manually:
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simplepip install error / dependency installation failed?
The script automatically retries using a domestic mirror (Tsinghua).
If it still fails, specify the mirror manually:
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simpleAlso ensure at least 2 GB of free disk space is available.
Wrong Python version (requires 3.10+)?
Check your current version:
python --versionIf below 3.10, download the latest version at:
Double-clicking start.bat does nothing or flashes briefly?
Python is not correctly added to the system PATH.
Reinstall Python and check "Add Python to PATH", then restart your computer and try again.
macOS blocks start.command from running?
Run this in Terminal to remove the restriction:
xattr -d com.apple.quarantine /your/path/to/start.commandIf you see:
"Cannot be opened because the developer cannot be verified"
You can:
- Right-click
start.command - Select "Open"
- Click "Open" again
🔑 API Configuration
Prompted that LLM is not configured?
Enter your API Key in the sidebar ⚙ and save.
How do I get an API Key?
Using DeepSeek as an example:
🗄️ Database Connection
How do I connect to a SQL database?
Use the following connection format:
mysql+pymysql://username:password@host:port/database_name
Example:
❌ Wrong:
mysql://user:pass@host:3306/dbname
✅ Correct:
mysql+pymysql://user:pass@host:3306/dbname
📊 Charts & Files
Chart links broken after restart?
Generated charts are saved locally at:
outputs/charts
You can open the corresponding HTML files directly in your browser.
We welcome every kind of contribution, whether you are fixing a bug, improving documentation, or adding a feature.
- Fork this repository
- Create a feature branch
git checkout -b feature/your-feature-name
- Commit your changes
We recommend following the Conventional Commits specification.
git add . git commit -m "feat: add your feature"
- Push your branch
git push origin feature/your-feature-name
- Open a Pull Request
- Describe your changes
- Include screenshots or a recording for new UI features
- Link related issues, if any
- Follow PEP 8 for Python code
- Run
flake8before submitting changes - Add relevant tests for new features
- Keep commit messages clear and concise
Report bugs through GitHub Issues and include:
- 📋 Steps to reproduce
- 🎯 Expected behavior
⚠️ Actual behavior- 📸 Screenshots, when applicable
- 💻 Environment details, including OS and Python version
A great open-source project is never a solo act. We're building a data tool that can genuinely handle complex business scenarios — one that races through massive datasets, navigates multi-table logic with ease, and surfaces insights on visual dashboards. We've hit a few deeply challenging, high-value problems. If you love solving hard technical problems, we need you:
- Multi-sheet inter-table logic optimization — How do you intelligently map dependencies and calculations across dozens of sheets?
- Visualization dashboard interactivity & performance — Making data stories flow more smoothly, intuitively, and powerfully.
- Model capability enhancement for specialized business scenarios — The edge cases that general-purpose tools can't handle.
- Remote server invocation — Building a framework for remote GPU calls.
- You'll tackle real, deep, non-toy technical challenges
- Your code will directly impact the productivity of front-line business users
- Contribute freely, collaborate flexibly — submit a PR or reach out directly, entirely up to you
- Outstanding contributors may be invited to become project Committers
- Open a Pull Request directly — we review within 24 hours
- Or contact us at:
[email protected](please note "Contributor + area of expertise") - Join the official community: QQ Group
991636855or Telegram Group or Discord
This software is licensed under CC BY-NC 4.0 and a Chinese software copyright registration has been applied for (Acceptance No. 2026R11S0817877, Software: Natural-Language Interactive Business Data Analysis Software V1.0). See the LICENSE file for the full terms.
Unauthorized commercial use is prohibited; contact the author for a separate commercial license.
Leave the process to Business Analytics Agent, and save your time for thinking.
⭐️ If this project helps you, please give it a star!
Made with ❤️ by Zafer-Liu
