DeepAnalyze

Autonomous Agents 💻 Python ⚖️ MIT 🟢 Actively maintained
4.6k stars

DeepAnalyze is an agentic large language model designed for autonomous data science. It solves the problem of automating end-to-end data-centric tasks, from data preparation to report generation, without human intervention. It is intended for data scientists, analysts, and developers who need to automate data analysis and research on diverse data sources.

✨ Key features

  • Autonomous end-to-end data science pipeline
  • Supports structured, semi-structured, and unstructured data
  • Open-ended data research with analyst-grade reports
  • Fully open-source model, code, and training data
  • Multiple interfaces: WebUI, CLI, JupyterUI
  • OpenAI-style API endpoint support

🎯 Use cases

  • Automate data preparation, analysis, and visualization
  • Generate comprehensive research reports from raw data
  • Perform deep research on databases, CSV, Excel, JSON, XML, and text files
  • Deploy a local data analysis assistant for any data science task

📦 Installation

🧰 Requirements: Requires Python 3.12, PyTorch, Transformers, and vLLM (>=0.8.5). For API access, an API key is required (apply via Google or Feishu form).

conda create -n deepanalyze python=3.12 -y
conda activate deepanalyze
pip install -r requirements.txt
For training
(cd ./deepanalyze/ms-swift/ && pip install -e .)
(cd ./deepanalyze/SkyRL/ && pip install -e .)

🚀 Usage

from deepanalyze import DeepAnalyzeVLLM

prompt = """# Instruction
Generate a data science report.

# Data
File 1: {"name": "bool.xlsx", "size": "4.8KB"}
File 2: {"name": "person.csv", "size": "10.6KB"}
..."""

workspace = "/path/to/workspace/"

deepanalyze = DeepAnalyzeVLLM(
    "/path/to/deepanalyze-8b/"
)
answer = deepanalyze.generate(prompt, workspace=workspace)
print(answer["reasoning"])

❓ FAQ

How do I get an API key for DeepAnalyze?

You can apply for an API key via the Google Form or Feishu Form linked in the README. Detailed usage instructions are in the API guide.

What are the hardware requirements for running DeepAnalyze locally?

The README provides a table with recommended GPU memory configurations. For example, 16GB can run a 4-bit quantized model with FP8 KV cache, while 80GB can run the original model with max context length.

Can I use DeepAnalyze through a command-line interface?

Yes, DeepAnalyze supports a CLI. You need to start the API server with python start_server.py and then run python api_cli.py (English) or python api_cli_ZH.py (Chinese).

What data formats does DeepAnalyze support?

It supports structured data (Databases, CSV, Excel), semi-structured data (JSON, XML, YAML), and unstructured data (TXT, Markdown).

📊 Repository

Stars★ 4,597
Forks🍴 729
Open issues🐛 24
Last commit🕒 Aug 30, 2026
Created📅 Oct 2025
Language💻 Python
License⚖️ MIT

🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.