FireWorks

Official Resources

• Homepage: https://materialsproject.github.io/fireworks/
• Documentation: https://materialsproject.github.io/fireworks/
• Source Repository: https://github.com/materialsproject/fireworks
• License: Modified BSD License

Overview

FireWorks is a workflow software for running, tracking, and managing high-throughput calculations. It is designed to handle complex workflows with dependencies, enabling the automation of large-scale computational tasks. FireWorks uses a centralized database (MongoDB) to store the state of all workflows, allowing for dynamic workflow modification and robust failure recovery.

Scientific domain: Workflow management, high-throughput computing Target user community: Computational scientists running large-scale simulations

Capabilities

• Centralized Management: Uses MongoDB to store workflow state, allowing queryable status of all jobs.
• Dynamic Workflows: Workflows can modify themselves during execution (e.g., adding new steps based on previous results).
• Duplicate Detection: Automatically detects and prevents duplicate runs.
• Failure Recovery: Rerun failed jobs, detect "lost" runs.
• Flexible Execution: Can run on local machines, supercomputers (SLURM, PBS, etc.), or cloud resources.
• Python & CLI: Full control via Python API or command-line interface.

Interfaces & Ecosystem

• Part of the Materials Project Stack: Designed to work with pymatgen (analysis), Custodian (error correction), and Atomate (recipes).
• Database: Requires MongoDB.

Workflow and Usage

A workflow in FireWorks consists of Fireworks (individual jobs) connected by dependencies.

1. Define Workflow:

   from fireworks import Firework, Workflow, LaunchPad
   
   # Define a simple task
   fw1 = Firework(ScriptTask(script="echo 'hello'"), name="hello")
   fw2 = Firework(ScriptTask(script="echo 'world'"), name="world")
   
   # Create workflow: fw2 depends on fw1
   wf = Workflow([fw1, fw2], {fw1: [fw2]}, name="hello_world")
   

2. Add to Database:

   launchpad = LaunchPad.auto_load()
   launchpad.add_wf(wf)
   

3. Execute: Run rlaunch (rapid launch) or qlaunch (queue launch) on computing resources to pull and execute jobs from the database.

Application Areas

• High-throughput materials discovery
• Complex simulation pipelines (e.g., relaxation -> static -> band structure)
• Distributed computing tasks

Verification & Sources

Primary sources:

1. Homepage: https://materialsproject.github.io/fireworks/
2. GitHub: https://github.com/materialsproject/fireworks

Confidence: VERIFIED

Verification status: ✅ VERIFIED

• Website: ACTIVE
• Documentation: COMPREHENSIVE
• Source: OPEN (GitHub)
• Development: ACTIVE
• Applications: Workflow management, Materials Project ecosystem