🔎 Overview

Academic-Data-Agent helps you analyze spreadsheet data without assembling the whole workflow by hand. Upload a file, describe what you want to learn, and the system produces a structured report with charts and conclusions.

The report is designed for review, not just reading. Supported charts and conclusions can be opened to inspect the data used, the calculation process, and the evidence behind the result. Previous reports stay in the sidebar so you can reopen, continue asking questions, or export them later.

It is designed for:

✨ Features

🧭 Workflow

Configure and test a model connection once
      ↓
Choose general analysis or mathematical modeling
      ↓
Upload the requested files and describe your goal
      ↓
Review the detected data or suggested plan when prompted
      ↓
Start the analysis and wait for the report
      ↓
Check important results in the report and export when ready

⚡ Quick Start

1. Install dependencies

Clone the repository, then install the Python and Web workspace dependencies:

You need Python 3.10 or newer, plus Node.js and npm.

git clone https://github.com/healer-666/Academic-Data-Agent.git
cd Academic-Data-Agent
pip install -r requirements.txt
cd frontend
npm install
cd ..

2. Configure a model

If you plan to use the Web workspace, you can skip the .env file: complete Step 4, open Model Settings, and enter:

• the model name;
• the service Base URL;
• your API key;
• the request timeout.

Click Test connection before starting an analysis. You only need to do this again when the model connection changes.

If you prefer an environment file, create .env in the repository root:

LLM_MODEL_ID=your_model
LLM_BASE_URL=https://your-llm-endpoint
LLM_API_KEY=your_api_key
LLM_TIMEOUT=120

Do not commit .env or share your API key.

3. Run from CLI

This step is optional. Use it when you prefer the command line:

python main.py --data data/simple_data.xlsx --query "Summarize the main patterns and verify the important differences."

4. Launch the Web workspace

For most users, this is the recommended way to use the project:

cd frontend
npm run build
cd ..
python web_app.py --host 127.0.0.1 --port 8010

Open http://127.0.0.1:8010 in your browser.

After opening the page:

1. Check the model status in the top-right corner.
2. Choose General data analysis or Mathematical modeling.
3. Upload the requested files and enter your question.
4. Review any data package or analysis plan shown by the system.
5. Start the analysis, then inspect or export the generated report.

🧰 Common Options

🧪 Python API

Python users can call the analysis workflow directly:

import sys
from pathlib import Path

sys.path.insert(0, "src")
from data_analysis_agent.agent_runner import run_analysis

result = run_analysis(
    Path("data/simple_data.xlsx"),
    query="Summarize the main patterns and verify the important differences.",
    quality_mode="standard",
)

print(result.report_path)
print(result.workflow_complete)

Each run is saved under outputs/run_*/. This folder contains the report, cleaned data, generated charts, and the information required to review the results later.

📊 Evaluation

These are local reproduction results, not official leaderboard submissions, official rankings, or SOTA claims. You do not need these benchmark datasets for normal use.

Detailed reports:

• DABench public benchmark report
• DataSciBench formal comparison
• DataSciBench clean ablation report

🗂️ Structure

.
├── data/                      # Example data and bundled competition cases
├── docs/                      # User guides and evaluation reports
├── frontend/                  # Web workspace
├── memory/                    # Local materials and saved experience
├── outputs/                   # Generated reports, charts, and analysis files
├── main.py                    # Command-line start
└── web_app.py                 # Web workspace start

📚 Documentation

• Competition experience library
• Mathematical modeling problem workspace
• Modeling skills
• DABench public benchmark report
• DataSciBench formal comparison

---