Together Open Data Scientist开源ReAct代理，执行Python以探索数据、构建模型和生成分析报告

4.3 (4)

审阅者 Daniel Nikulshyn·更新 2026年6月

Open Source Data Science Python AI Agent Cloud Local Analytics Reporting

概览

Together Open Data Scientist 是由 Together AI 在 GitHub 上发布的开源、AI 驱动的数据分析代理。它遵循 ReAct（Reasoning + Acting）框架，在语言模型推理步骤和具体的 Python 代码执行之间交替，以完成端到端的数据科学任务，例如探索数据集、计算摘要统计、构建模型以及生成详细的书面分析报告。该代理可以以两种模式执行 Python。 “internal” 模式在本地的 Docker 容器中运行代码，适合单用户本地开发；而 “tci” 模式则将执行任务卸载到 Together Code Interpreter（TCI），通过 Together AI API 访问的云沙箱。用户可以上传数据目录以实现自动摄取，设置最大推理迭代次数，并选择驱动代理的底层模型——默认使用 DeepSeek‑V3，也可以指定 Together 平台上提供的 Llama 系列模型及其他模型。它以 pip 可安装的包（open-data-scientist）形式发布，并同时提供命令行界面（CLI）和 Python API。CLI 支持诸如 --write-report 用于生成 Markdown 分析报告、--save-trace 用于记录完整的查询和执行跟踪，以及通过会话 ID 进行会话复用等选项。Python API 以 ReActDataScienceAgent 类为核心，该类接受自然语言任务并返回结果。 Together Open Data Scientist 项目明确标注为实验性软件。由于所有代码和分析均由 AI 生成，输出可能包含错误或次优方案，最好将其视为探索和学习的起点，而非用于生产决策。维护者强调，尤其在关键业务或研究应用中，仍需人工监督和验证。相较于商业 AI 数据分析助手（如 ChatGPT 的 Advanced Data Analysis 或 notebook copilots），Together Open Data Scientist 的差异在于它完全开源、可自行托管、在 Together 生态系统内模型无关，并且能够自主串联多个代码执行步骤，生成完整报告，而不是仅提供一次性答案。

主要功能

ReAct 行为推理代理循环
本地 Docker 或 Together Code Interpreter 云执行两种模式
自动数据目录上传分析
Markdown 报告生成选项 --write-report
可配置模型和最大推理迭代数
命令行界面和可编程 Python API
SaaS API

价格

模型: Free
分类: AI
评分: 4.3 / 5 (4)

使用场景

自动数据集探索

在新的数据集上运行代理并执行探索数据分析 Python，接收详细的报告

模型构建辅助

使用代理来在本地或云环境中prototype 和建立机器学习模型

分析报告生成

生成详细的写入分析报告，概括数据集见解和模型结果

本地或云Python工作流

在计算需求中根据灵活性执行Python 基于数据科学任务

优点 & 缺点

优点

开源和可自主部署
在 Docker 或 TCI 云中运行真正的Python代码
在Together的生态系统中具有可配置的潜在LLM和迭代数
CLI 和 Python API，以及自动报告和跟踪生成
可扩展性

缺点

明确为实验软件；ai生成的代码可能包含错误
需要人类审阅，不适合生产决策
Docker模式有会话隔离和安全限制
绑定到一起 AI API 密钥以便于云执行
学习成本高

对决战绩

在万神殿中参与了 6 对决。

第1

第2

第3

Last 5 battles

评测

4.3

4 个评分的平均值。

登录以留下评测。

Victor Nguyen

Apr 23, 2026

Solid for our team

We rolled this out across the team last quarter and the value for money is strong. The integrations fits neatly into how we already work, and the automation removed a step we used to do by hand. A few rough edges remain, which is the main caveat, but it has held up under daily use.

Kwame Mensah

Apr 12, 2026

Years in this space

I've evaluated a lot of these over the years. What stands out here is the API — handled better than most — and it is genuinely easy to set up. A few rough edges remain is my one real gripe. Worth the time if this is your use case.

Liam O’Connor

Nov 5, 2025

Use it every day

Honestly didn't expect to like it this much. The API is exactly what I needed, and it saves real time. I do wish the docs could be deeper, but I reach for it almost every day now and it just clicks.

Diego Fernández

Sep 24, 2025

Does the job

Pretty happy overall. The core workflow just works and support is responsive. The docs could be deeper can be annoying, but no dealbreakers — I'd recommend it to a friend without hesitating.

问答

What output does the tool generate and how can I retrieve it?

After analysis, the agent can generate a markdown report using the `--write-report` flag, and it also produces a trace of the reasoning steps. Both the report and trace are saved to your working directory, making it easy to review the results and the code that was executed.

Asked by Julia Steiner · Aug 21, 2025

Can I customize the language model or limit the reasoning steps?

Yes. The agent is model‑agnostic; you can specify any compatible LLM for the reasoning part and set a maximum number of ReAct iterations, allowing you to control how many reasoning‑and‑acting cycles the agent performs on a task.

Asked by Sami Virtanen · Aug 6, 2025

What are the requirements for using the cloud (TCI) execution mode?

To use the "tci" mode you need a valid Together AI API key, which authenticates requests to the Together Code Interpreter service. The service runs your code in a managed cloud sandbox, so no local Docker installation is required.

Asked by Ines Fernandes · Jul 23, 2025

How does the execution mode choice affect where my Python code runs?

Together Open Data Scientist offers two modes: "internal" runs the generated Python inside a local Docker container for single‑user development, while "tci" sends the code to Together Code Interpreter, a cloud sandbox accessed via the Together AI API. Choose Docker for full local control; choose TCI for cloud execution without local Docker setup.

Asked by Hasan Demir · Jun 15, 2025