Installation

Installation

DataAgent requires Python >=3.11. We recommend using uv to manage the environment and dependencies.

1. Install uv

If uv is not installed yet, install it via the official method or pip:

python3 -m pip install uv

Verify installation:

uv --version

2. Install from Source

git clone https://github.com/datagallery-ai/DataAgent.git
cd dataagent
uv sync

To run tests or build documentation:

uv sync --extra test
uv sync --extra mkdoc

Development tools:

uv sync --group dev

3. Configure Environment Variables

Copy .env.example to .env and fill in model API keys, model service URLs, database endpoints, and other environment-specific values.

cp .env.example .env

Which variables are required depends on your YAML configuration. For example, when MODEL uses provider: "bailian", provide the corresponding provider API key.

4. Verify Installation

uv run -m dataagent quickstart

You can also run an example configuration:

import asyncio
from dataagent.interface.sdk.agent import DataAgent


async def main():
    agent = DataAgent.from_config("dataagent/core/flex/examples/ecommerce_agent.yaml")
    result = await agent.chat("What can you do?")
    print(result["messages"][-1].content)


asyncio.run(main())

5. Optional Database Services

Some scenarios use Elasticsearch, PostgreSQL, MySQL, or business databases. Deploy these external services only when your configuration enables MEMORY, DATASOURCE, NL2SQL databases, or specific business tools.

For database deployment and data import, see: Database Installation Guide.

6. Build Distribution Packages

To build wheel/sdist artifacts:

uv build

Output is written to dist/.