AiiDA PythonJob & PyFunction

PythonJob and pyfunction enable users to run Python functions—either locally or remotely—with automatic serialization, provenance tracking, and workflow integration.

These tools are designed to simplify the experience for users not familiar with AiiDA’s internal types, allowing them to write pure Python functions while still benefiting from AiiDA’s powerful infrastructure.

Key Features

1. Remote Execution with PythonJob Seamlessly run Python functions on a remote computer via PythonJob. A working directory is automatically created for the job, allowing full support for:

   • File upload and retrieval

   • Parent/remote folders

   • Integration with external codes (e.g. ASE + DFT engines)

2. Local Execution with PyFunction Use @pyfunction to run pure Python functions locally. All inputs and outputs are automatically serialized/deserialized, enabling users to work with native Python types. This decorator is ideal for functions that are:

   Note

   pyfunction does not create a dedicated working directory. It executes in the current local Python environment. File I/O or relative path access should be avoided unless explicitly handled.

3. User-Friendly Interface Designed for users unfamiliar with AiiDA’s internal Data classes. The decorators handle all conversion and provenance tracking behind the scenes.

4. Workflow Management with Checkpoints Combine pyfunction and pythonjob in WorkGraph workflows to build robust, restartable workflows with full checkpoint support.

5. Full Data Provenance All inputs, outputs, and intermediate results are stored in the AiiDA provenance graph, ensuring reproducibility and traceability.

When to Use What

Decorator	Execution	Input/Output Types	AiiDA Provenance	Recommended Use Case
@pyfunction	Local	Python native	✅ Yes	Pure functions
@pythonjob	Remote	Python native	✅ Yes	Remote jobs with file handling