A component that takes input messages (e.g., those containing code blocks), performs execution, and outputs messages with the results.
- Inside an agent
- Executes the code → returns results
- Two built in code executors
- Command Line Code Executor: runs code in cmd env (Unix shell)
- Jupyter Executor: runs in a Jupyter env
- Execute code 2 ways
- Locally: Run directly on the host system
- Docker container: Runs code in an isolated conainer
- More secure when using production work
- Process
- The code executor will write to a code file, then execute that code file. Then it will read the console output of the code execution, and it’s sent as a reply message.
- Everything triggered by the message the agent receives