Auto-Analyst: Your Open-Source AI Data Scientist for Automated Workflows

By Firebird Technologies

Auto-Analyst is a fully open-source, modular AI system meticulously designed to automate comprehensive data science workflows. From initial data cleaning and statistical analysis to advanced machine learning and insightful visualizations, Auto-Analyst streamlines your entire process.

Experience it live at: https://www.autoanalyst.ai/chat

🚀 Key Highlights

• Open Source: Licensed under the highly permissive MIT License, fostering flexibility and collaboration.
• LLM Agnostic: Enjoy broad compatibility with any Large Language Model (LLM) API, including OpenAI, Anthropic, Deepseek (Groq), and more.
• Bring Your Own API Key: Avoid vendor lock-in by utilizing your own API keys, ensuring you only pay for what you use.
• User-Centric UI: Crafted with the data scientist in mind, offering an intuitive and efficient user experience.
• Reliable Outputs: Integrated guardrails ensure robust, interpretable, and trustworthy analytical responses.
• Modular Agent Architecture: Easily add or customize specialized agents using the powerful DSPy framework.

Live Application

Start your data analysis journey here: https://www.autoanalyst.ai/chat

How It Works: A Step-by-Step Walkthrough

1. Upload Your Dataset

Click the 📎 icon near the chat input to upload your data. Auto-Analyst currently supports .csv or .xlsx files. Additional connectors (for APIs, SQL databases, etc.) are available upon request.

2. Describe Your Dataset

Provide a brief textual description of your dataset's content. Auto-Analyst will then generate a cleaned, structured metadata summary, optimized for seamless integration into LLM workflows.

• Tip: Improve analysis quality by renaming generic columns like var_1 to descriptive names such as price or category.

3. Ask a Question

Interact with Auto-Analyst in two ways:

• Specify an Agent: Use @agent_name (e.g., @preprocessing_agent) to direct your query to a particular agent.
• Planner Mode: Omit an agent tag, and the intelligent planner will automatically route your query to the most suitable agent(s).

Built-in Agents

Auto-Analyst comes equipped with a suite of specialized agents:

• @preprocessing_agent: Cleans and prepares data using pandas and numpy. This includes fixing data types, handling null values, and computing aggregates.
• @statistical_analytics_agent: Performs various statistical tests such as regression, correlation, and ANOVA using statsmodels.
• @sk_learn_agent: Trains machine learning models like Random Forest, KMeans, and Logistic Regression leveraging scikit-learn.
• @data_viz_agent: Generates high-quality visualizations using plotly, featuring a retriever to select optimal chart formats automatically.

🌟 Modular and Extensible! You can easily add custom agents tailored for specific domains, such as:

• Marketing Analytics
• Quantitative Finance
• Web APIs (e.g., Slack, Notion)

💬 Planner Mode

For automated query routing, simply type your question without specifying an agent. The planner will intelligently:

• Select the appropriate agent(s).
• Generate detailed plan instructions.
• Coordinate workflows between multiple agents.
• Collect and display comprehensive results, including plots and summaries.

🧑‍💻 Developer Features

📁 Modular Agent System (DSPy)

Agents are implemented as dspy.Signature classes, making them highly customizable and easy to extend. Here's an example:

class google_ads_analyzer_agent(dspy.Signature):
    goal = dspy.InputField(desc="User goal")
    dataset = dspy.InputField(desc="DataFrame")
    plan_instructions = dspy.InputField(desc="Instructions")
    code = dspy.OutputField(desc="Python code")
    summary = dspy.OutputField(desc="Analysis summary")

You can integrate your own agents in minutes.

🔌 Built-in Dataset Connectors

Auto-Analyst offers robust integration with various data sources:

• Ads: Google Ads, Meta, LinkedIn Ads
• CRM: HubSpot, Salesforce
• SQL: Postgres, MySQL, Oracle, DuckDB

Looking for more? Submit a request via our Contact Us page.

🖼️ UI Feature Overview

Feature	Description
Chat Interface	Engage in natural language conversations to ask questions and receive answers.
Code Editor	Inspect and edit generated code, with AI-assisted edits and auto-fix capabilities for broken code.
Analytics Dashboard (Enterprise)	Monitor usage, set limits, allocate credits, and enforce roles & permissions for enterprise environments.

🛠 Backend Highlights

• Agent orchestration powered by DSPy.

• Model-agnostic LLM support for maximum flexibility.

• Built-in chart formatter for intelligent visualization type suggestions.

• Multi-agent workflows seamlessly managed by a centralized planner.

• Daily scheduled reports and auto-regeneration capabilities, ideal for enterprise deployments.

📅 Roadmap

🔜 Short-Term Goals

• Deep Analysis Mode: An LLM-equivalent of long-form research for in-depth insights.
• Multi-CSV / Multi-Sheet Excel Analysis: Enhanced capabilities for complex data inputs.
• User-Defined Analytics Agents via UI: Empower users to create custom agents directly through the interface.
• Improved Code-Editing and Auto-Debugging: Advanced tools for refining and troubleshooting generated code.

🔭 Long-Term Vision

• Usability-First: Continuous optimization of the user experience through iterative design and valuable user feedback.
• Community-Driven: The future of Auto-Analyst will be shaped by the global analyst community. Follow us on Substack and LinkedIn for updates.
• Open Collaboration: Fostering a collaborative environment to build and share new agents, retrievers, and datasets.

🧩 Contributing

We warmly welcome contributions from the community! You can contribute by:

• Adding new agents.
• Suggesting user experience improvements.
• Contributing templates or datasets.
• Submitting bug reports or pull requests.

📬 Contact Us

For collaboration opportunities or enterprise inquiries, please visit: https://www.autoanalyst.ai/contact

📄 License

Auto-Analyst is released under the permissive MIT License, encouraging free use, remixing, and building upon the project.

🐦 Follow Us

• Website: https://www.autoanalyst.ai/
• Substack: https://substack.com/
• LinkedIn: https://www.linkedin.com/

Built with ❤️ by Firebird Technologies AI. Tech. Fire.