Air Data Agent (airda) is a multi-agent system designed for data analysis, capable of understanding data development and analysis needs, data schemas, and generating SQL and Python code for querying, visualization, and machine learning tasks. It targets data analysts and developers seeking to streamline data exploration and insight generation.

How It Works

Airda employs a multi-agent architecture where specialized agents collaborate to fulfill user requests. This includes agents for data retrieval, SQL generation, Python code generation, and visualization. The system engages in multi-turn conversations to clarify user needs, plans tasks, and executes them through these specialized agents, which can also self-debug their code to improve accuracy and efficiency.

Quick Start & Requirements

• Install: pip install airda -i https://pypi.python.org/simple/
• Prerequisites: Python >= 3.10, MongoDB (Docker recommended: docker pull mongo, then docker run -itd --name mongo -v /{path_of_mongo_data}:/data/db -p 27017:27017 mongo).
• Configuration: Load environment variables from .env_template and optionally customize logging via log_config.yml.template.
• Data Sources: Add data sources using airda datasource add (currently supports MySQL). Sync schema with airda datasource sync.
• Usage: Start interaction with airda run cli -n {datasource_name}.
• Docs: https://github.com/hitz-ids/airda

Highlighted Details

• Precision data retrieval from numerous tables.
• Understanding of business logic and metrics beyond raw data.
• Multi-agent collaboration with self-debugging capabilities.
• Generates SQL, Python code, and supports data visualization.

Maintenance & Community

The project welcomes contributions via GitHub Issues and Pull Requests, with templates provided for both. Further details on contribution guidelines can be found in the repository.

Licensing & Compatibility

The repository does not explicitly state a license in the provided README. Users should verify licensing for commercial or closed-source integration.

Limitations & Caveats

The project is actively under development, with features like "语料库" (corpus) and "图表生成" (chart generation) marked as incomplete. The default embedding model requires manual download if not present in the Hugging Face cache. Data source support is currently limited to MySQL.