A powerful Python library for creating, managing, and executing scalable scientific workflows with automatic data provenance.

AiiDA-WorkGraph empowers researchers and developers to build complex, reproducible workflows with ease.

• 🎨 Pythonic Workflow Design: Define workflows using familiar Python functions and decorators.
• 🖥️ Interactive GUI: Visualize, monitor, and debug your workflows in real-time with a user-friendly web interface.
• 🔗 Automatic Data Provenance: Guarantee scientific reproducibility with zero effort. The complete history of all data and calculations is automatically tracked.
• 🚀 Remote & Parallel Execution: Seamlessly offload tasks to remote supercomputers and run them concurrently.
• 🧠 Dynamic Control Flow: Build adaptive workflows that respond to data at runtime using standard Python if/else statements and loops.
• 🛡️ Checkpointing & Error Handling: Protect long-running workflows from interruptions and build resilient logic to recover from failures.
• ⚡ High-Throughput Computing: Built to scale, AiiDA-WorkGraph can efficiently manage thousands of concurrent workflows.
• 🧩 Reusable Components: Encapsulate common routines as sub-workflows and easily reuse them in larger, more complex pipelines.

pip install aiida-workgraph

First, ensure you have a working AiiDA environment.

verdi presto  # Or 'verdi quicksetup' for a detailed setup

Let's create a simple workflow to calculate $(x + y) \times z$.

1️⃣ Define Tasks

Use the @task decorator to turn Python functions into workflow components.

from aiida_workgraph import task

@task
def add(x, y):
    """Adds two numbers."""
    return x + y

@task
def multiply(x, y):
    """Multiplies two numbers."""
    return x * y

2️⃣ Compose a Workflow

Use the @task.graph decorator to link tasks. Data flows naturally from one task's output to the next one's input.

@task.graph
def add_multiply(x, y, z):
    """A workflow to add two numbers and then multiply by a third."""
    sum_result = add(x, y).result
    product_result = multiply(x=sum_result, y=z).result
    return product_result

3️⃣ Run the Workflow

Build the workflow with your inputs and run it.

from aiida import load_profile

# Load your AiiDA profile
load_profile()

# Build and run the workflow
results = add_multiply.run(x=2, y=3, z=4)

# Print the final result
print(f"✅ Result: {results}")
# Expected output: ✅ Result: 20

4️⃣ Automatic Provenance Tracking

AiiDA-WorkGraph automatically generates a detailed provenance graph, tracking the full history of data and calculations to ensure full traceability and reproducibility. Here is an example of the provenance graph generated for the above workflow:

AiiDA-WorkGraph supports three complementary approaches to building workflows, letting you choose the best method for your needs.

• 🐍 Pythonic Workflows (Recommended): Use @task.graph for clean, readable, and powerful workflows, as shown in the Quick Start.

• 👁️ Visual Graph with Explicit Logic: Use zones like If, While, and Map to build a graph where the control flow is visually explicit.

• ⚙️ Low-Level Node-Graph Programming: Programmatically define each task and link them manually for maximum control and dynamic graph generation.

Visualize, monitor, and debug your workflows in real-time. To launch the GUI, first install the package and then run:

pip install aiida-gui-workgraph
aiida-gui start

Navigate to http://127.0.0.1:8000/workgraph in your web browser.

Note: The GUI is an experimental feature and is under active development.

• 📚 Full Documentation: Dive deep into all features and capabilities.
• 🧪 Demo & Examples Repository: See real-world examples with various computational codes.
• 🧑‍💻 Development & Contribution: Learn how to contribute to the project.
• 📄 License: AiiDA-WorkGraph is licensed under the MIT License.