Introduction to Dynamic Data Access
Dynamic Data Access is available from version 1.1 onwards.
Sema4.ai Dynamic Data Access empowers AI agents with zero-copy access to enterprise data, enabling them to make informed decisions using past, present, and future data without creating additional data silos. This tutorial will guide you through the complete lifecycle of building, deploying, and managing data-enabled agents.
After completing this tutorial, you will be able to:
- Create Data Actions that connect agents to enterprise data sources
- Build agents that leverage real-time data access for decision-making
- Deploy data-enabled agents securely in production environments
- Monitor and manage data access across your agent ecosystem
Components Overview
Dynamic Data Access connects your enterprise data sources to AI agents through a seamless workflow that spans development, deployment, and production use. Data Action Packages serve as the bridge between your data sources and the Sema4.ai platform components, enabling different personas to work with data-enabled agents according to their roles.
| Component | Description |
|---|---|
| Sema4.ai Data Access extension | VS Code extension for working with data sources, queries, models and predictions. Allows you to write and run SQL queries against data sources, and create named queries that can be used by agents. |
| Sema4.ai Data Server | A component that provides the zero-copy access to enterprise data sources. On a developer laptop our SDK and Studio shares one Data Server, and automatically install and run it. In Control Room each deployed agent has its own Data Server instance. |
| Actions | Action packages contain Data Access components, such as named queries that enable agents to interact with enterprise data sources. |
| Data Source | A connection to a specific data source. The data sources are identified by a unique name and "engine" type, for example snowflake, redshift, file. You may use different Saved Connection Details for the Data Source when developing locally and in Control Room. In Control Room you can further use different Connection Details for e.g. Testing and Production environments. |
| Connection Details | Stored connection details for a Data Source - these are used to configure the Data Source in the Data Server either in Studio or in Control Room. They typicall contain details such as host, port, username, password, etc. |
What You'll Build
Throughout this tutorial, you'll create a fully functional enterprise agent that can:
- Connect to multiple data sources including PostgreSQL, and file-based storage
- Execute named queries for precise data retrieval
- Deploy safely with Control Room
- Scale in production with monitoring and auditing
We will provide all the necessary datasets and demo databases to follow this tutorial. We are using a small synthetic dataset that mimics sales orders and customer data in order to build a simple agent that helps a fictional sales team to make better decisions grounded in real-world data.
Let's begin by setting up your development environment for building data-enabled agents!