Introduction to AiiDA

What is AiiDA?

AiiDA (Automated Interactive Infrastructure and Database for Computational Science) is an open-source workflow management engine for computational science. It runs, tracks, and stores calculations while automatically preserving their full provenance — a complete, queryable record of every input, output, calculation, and the connections between them.

At its core, AiiDA lets researchers describe complex, multi-step computations once and then run them reliably and reproducibly, whether that means a single calculation on a local machine or thousands of jobs dispatched to remote high-performance computing (HPC) clusters. Every piece of data that enters or leaves a calculation is captured as a node in a directed graph, so results can always be traced back to exactly how they were produced.

Aims of AiiDA

AiiDA is built around four guiding principles, often summarised as ADES — Automation, Data, Environment, and Sharing:

Automation

Complex workflows spanning many calculations and codes can be encoded once and executed with minimal manual intervention. The engine handles job submission, data transfer, scheduling, and error recovery, making high-throughput studies of thousands of calculations practical.

Data

All inputs, outputs, and the calculations connecting them are stored in a provenance graph. This makes the full history of a result explicit and allows the entire database to be searched and queried efficiently.

Environment

By recording the complete provenance of every calculation, AiiDA makes results reproducible — the exact steps and data behind any output can be inspected and re-run, supporting the FAIR (Findable, Accessible, Interoperable, Reusable) principles for research data.

Sharing

Provenance graphs, workflows, and results can be exported and shared with collaborators or the wider community, enabling others to reuse and build upon existing work.

How AiiDA is Used

A typical AiiDA workflow involves a few key concepts:

  • Computers and codes — AiiDA is told about the machines available to it (from localhost to remote HPC clusters reached over SSH) and the executables installed on them.

  • Plugins — code-specific plugins know how to prepare the input files for a given program and parse its outputs back into structured AiiDA data nodes.

  • The daemon — a long-running background process picks up submitted calculations and workflows, drives them through their steps, submits jobs to the appropriate computers, and records everything as it goes.

  • Querying results — because every calculation and data node is stored in the provenance graph, results can be retrieved, compared, and analysed long after the original run.

Traditionally these features are accessed through AiiDA’s Python API or command-line interface, which offers great power and flexibility but assumes familiarity with programming and with AiiDA’s concepts.

AiiDAlab builds on top of AiiDA to make these capabilities available through a browser-based graphical interface, so that users can configure resources, submit workflows, and inspect results without writing code. The next page, AiiDA & AiiDAlab: How They Work Together, looks at how AiiDAlab and the underlying AiiDA components fit together.