Contexts

A context is what tells Beaker how to be helpful for a particular kind of work. When you set a context, you are choosing the language Beaker runs (its subkernel), the tools the agent has access to, any data or objects that should be pre-loaded into your environment, and the instructions that shape how the agent thinks about your requests.

Beaker ships with a default context that works well for general-purpose programming and data analysis. Custom contexts can be installed as Python packages — see Adding Contexts — and you can switch between installed contexts at any time from the context selector in the notebook UI.

Some examples of what a context might add:

• Domain knowledge — a context for working with a specific modeling library can pre-load helpful objects and teach the agent that library’s conventions.
• External data — a context can wire up access to a database, an API, or a private dataset, exposed as integrations the agent can use.
• Custom tools — a context can give the agent specialized tools (run a simulation, query a knowledge base, render a domain-specific visualization).
• Workflows — a context can ship workflows the agent can follow when accomplishing common tasks.

To build your own context, see Context Development.