Skip to main content
Context Engine teaches Maia Team about your business. It builds knowledge graphs that capture what your data means, how it’s structured, and how it connects, so Maia Team can work with your data the way your team does. Your warehouse stores tables and columns, but it doesn’t say what they mean to your business. A column named order_total doesn’t tell Maia Team whether it holds revenue before or after tax, or which of several similar tables your analysts rely on. Without that context, Maia Team has to work out the answer from names and structure alone. With a knowledge graph, it already knows. This guide explains why Context Engine is useful, how knowledge graphs work, and how to use Context Engine. To get started with your first knowledge graph straight away, read Creating knowledge graphs. If you’ve already created a knowledge graph, read Managing knowledge graphs and Visualizing knowledge graphs to learn how to configure additional crawlers, manage access to knowledge graphs, and explore the data in your knowledge graphs.
Visualizing a knowledge graph for the first time? We’ve included the demo Ecommerce Knowledge Graph in Maia to help you learn about visualizing data in knowledge graphs. Open it up, ask Maia Team about it, and explore the data it contains. For more information, read Exploring knowledge graphs.
Expand this box to watch our video about Context Engine and how to use it to build and manage knowledge graphs.

​
How Context Engine helps you work more efficiently

When Maia Team has access to a knowledge graph, you can:
  • Ask in your own words: Use the terms your business uses, such as “revenue”, “active customer”, or “churn”, instead of naming exact tables and columns. Maia Team maps your terminology to the right data.
  • Build pipelines that fit your data landscape: Maia Team builds pipelines using the tables, columns, and relationships that exist in your warehouse, rather than starting from generic assumptions.
  • Get more relevant answers: Maia Team draws on your business data, semantics, and processes when it responds, so you spend less time adding background detail to your prompts and correcting its assumptions.
  • Understand your data landscape: Explore your knowledge graph visually to see how business concepts, tables, columns, and pipelines relate to one another.
  • Keep context up to date: Crawlers refresh your knowledge graph on a schedule you set, so it keeps pace with changes to your data.
  • Share context across your team: A knowledge graph is built once and reused by everyone who has access to it, so your team works from a shared understanding of the same data.

​
Knowledge graphs

A knowledge graph is a map of your data and what it means. It’s made up of nodes, which are the things in your data landscape, such as tables, columns, business concepts, and pipelines. It also contains edges, which are the relationships that connect them. Every knowledge graph has two connected layers:
  • Physical layer: The structure of your warehouse, including its tables, columns, and the relationships between them.
  • Semantic layer: The business meaning on top of that structure. This includes the business areas, concepts, and properties that your data represents, each linked to the physical tables and columns that hold the data.
For example, a business concept such as “total order revenue” can link to every column across your warehouse that holds that value. When you ask Maia Team about revenue, it follows those links to find the right data. To learn more about what knowledge graphs contain and how to explore them, read Visualizing knowledge graphs. There’s no set number of knowledge graphs you should have. As a general rule, we recommend that you add all your business information to one knowledge graph, as this will help Maia Team understand your business logic and workflows in more detail. For example, if you want Maia Team to know about a new schema that holds sales data, it’s best to add this schema to an existing knowledge graph, so Maia Team can see how the new data connects to its existing knowledge. If the data you want to add is subject to different data sovreignty requirements (i.e. it needs to be processed in a specific region), or if it needs different access requirements, create a new knowledge graph. Each knowledge graph has its own contributors and its own list of projects that can use it, so a separate knowledge graph lets you restrict sensitive data to specific projects and users.

​
Using Context Engine

  1. Create a knowledge graph and add crawlers to it. A crawler connects to your data warehouse or your pipeline executions, and discovers and maps the structure it finds. You don’t need to build the graph by hand.
  2. Choose a schedule for your crawlers to run on, so the knowledge graph stays up to date as your data changes.
  3. When creating a task in Mission Control, or chatting with Maia Team in Designer, Maia Team selects the most relevant knowledge graph for your prompt automatically. For more information, read Knowledge graphs in chats.
The richer your data’s metadata, the more Maia Team can learn about it. We recommend you add metadata, such as column descriptions and tags, to your data—that is, your databases, schemas, or datasets. This metadata helps Maia Team understand your data landscape and provide better responses to your prompts.

​
Accessing Context Engine

The Context Engine dashboard shows all the knowledge graphs you can access or request access to. Use the Projects drop-down to filter the displayed knowledge graphs by project, or use the search field to search for a knowledge graph by name or description. To open the Context Engine dashboard, in the left navigation, click , then Context Engine. Access to Context Engine and knowledge graphs depends on your account settings and role:
  • Context Engine is only visible in the left navigation if the Enable AI Agents permission is enabled for your account.
  • By default, only users who have an account role with the Create knowledge graph permission can create and edit knowledge graphs and crawlers. However, all users can request contributor access to edit specific knowledge graphs.
  • All users can select knowledge graphs when creating tasks in Mission Control, and Maia Team will automatically use knowledge graphs in chats with all users if they enable Context Engine in chats.

​
Data encryption

Context Engine encrypts data both at rest and during crawling. Data at rest is encrypted in the Context Engine backend, which uses Amazon Neptune to store the knowledge graph and a PostgreSQL database to store vector content. Data being crawled is also encrypted throughout the crawling process, including while data is staged during harvesting, before it’s added to the knowledge graph as nodes and edges.