Subkernels

A subkernel is the language runtime that actually executes the code in your Beaker notebook. When you run a code cell, Beaker passes the code through to its subkernel, which runs it the same way a stand-alone Jupyter kernel would.

The subkernel determines what language your notebook speaks and which libraries are available. Beaker ships with subkernels for Python, R, and Julia, and additional subkernels can be installed as Python packages — see Adding Subkernels.

You do not usually choose a subkernel directly. Instead, you choose a context, and the context decides which subkernel to use. Most contexts work with the default Python subkernel; some contexts are tied to a particular language because their tools or pre-loaded objects only make sense there.

The subkernel runs in its own process with its own environment, so:

• Variables you define in one cell persist into later cells, just like a regular Jupyter notebook.
• The agent can introspect the subkernel’s state (variables, imports) when deciding what code to generate.
• Switching contexts to one that uses a different language will start a fresh subkernel — your previous variables will not carry over.

To build your own subkernel for a new language, see Subkernel Development.