orcastrator 3.2.5

ORCA high-throughput pipeline orchestrator

Orcastrator

Orcastrator is a Python toolkit and CLI for automating ORCA quantum-chemistry calculations.

Features:

• Simple TOML-based configuration
• Automatic directory and scratch management
• Multi-stage pipelines with caching and chaining (opt → freq → sp)
• Parallel execution across molecules
• SLURM batch script generation

Version: 3.0.0 License: MIT Requirements: Python ≥3.11

Installation

uv tool install orcastrator

Global Configuration (Optional)

You can create a global configuration file at ~/.orcastrator_config to set default paths:

# Path to ORCA executable (if not in PATH)
orca_path = "/opt/orca/orca"

# Path to OpenMPI installation directory
openmpi_path = "/opt/openmpi"

# Default scratch directory
scratch_dir = "/scratch"

Settings are optional - if not specified, Orcastrator will:

• Search PATH for ORCA executable
• Use /scratch (or /tmp if /scratch doesn't exist) for scratch files

All paths support ~ for home directory expansion.

See .orcastrator_config.example in the repository for a template.

Quickstart

1. Generate a template configuration:

   orcastrator init
   

2. Edit orcastrator.toml:

   # Core paths (relative to this file)
   output_dir = "output"
   molecules_dir = "molecules"
   
   # Resources
   cpus = 8
   mem_per_cpu_gb = 4
   workers = 2
   scratch_dir = "/scratch"
   
   # Workflow settings
   overwrite = false
   debug = false
   
   # Molecule filtering (optional)
   # include = ["mol1", "mol2"]
   # exclude = ["mol3"]
   rerun_failed = false
   
   # Pipeline stages
   [[stages]]
   name = "opt"
   simple_keywords = ["D4", "TPSS", "def2-SVP", "OPT"]
   
   [[stages]]
   name = "freq"
   simple_keywords = ["D4", "TPSS", "def2-SVP", "FREQ"]
   keep = ["*.gbw"]  # Copy wavefunction from previous stage
   
   [[stages]]
   name = "sp"
   simple_keywords = ["D4", "TPSSh", "def2-TZVP"]
   keep = ["*.gbw"]
   

3. Run the pipeline:

   orcastrator run orcastrator.toml
   

4. (Optional) Submit to SLURM:

   orcastrator slurm orcastrator.toml
   orcastrator slurm --no-submit orcastrator.toml  # just generate script
   

Configuration Reference

Core Settings

• output_dir - Output directory for results (relative to config file)
• molecules_dir - Directory containing .xyz files (relative to config file)
• cpus - Total CPU cores available
• mem_per_cpu_gb - Memory per CPU core in GB
• workers - Number of parallel workers (molecules processed simultaneously)
• scratch_dir - Scratch directory for temporary files (default: /scratch)

Workflow Settings

• overwrite - Rerun all calculations (default: false)
• debug - Enable debug logging (default: false)
• include - List of molecule names to include (optional)
• exclude - List of molecule names to exclude (optional)
• rerun_failed - Only rerun previously failed molecules (default: false)

SLURM Settings (Optional)

• nodelist - SLURM node list (e.g., ["node001", "node002"])
• exclude_nodes - SLURM nodes to exclude
• timelimit - SLURM time limit (format: HH:MM:SS)

Stage Configuration

Each [[stages]] block defines a calculation step:

• name - Unique stage identifier
• simple_keywords - ORCA keywords (e.g., ["OPT", "TPSS", "def2-SVP"])
• blocks - Additional ORCA % blocks (e.g., ["%scf maxiter 150 end"])
• mult - Override multiplicity for this stage (optional)
• charge - Override charge for this stage (optional)
• keep - File patterns to copy from previous stage (e.g., ["*.gbw", "*.xyz"])

Variable Substitution

You can use {variable_name} placeholders in both simple_keywords and blocks to make molecule-specific calculations:

Global defaults:

[keywords]
casscf_roots = 2
active_electrons = 6

[[stages]]
name = "casscf"
simple_keywords = ["CASSCF({active_electrons},{casscf_roots})", "def2-SVP"]
blocks = ["%casscf nroot {casscf_roots} end"]

Molecule-specific overrides in XYZ files:

5
{"charge": 0, "mult": 1, "casscf_roots": 4, "active_electrons": 8}
C  0.0  0.0  0.0
...

Resolution order:

1. Molecule-specific metadata (from XYZ file)
2. Global keywords (from [keywords] section in config)
3. Error if variable not found

This allows you to set sensible defaults while overriding parameters for specific molecules as needed.

Architecture

Orcastrator 3.0.0 uses a streamlined architecture for better maintainability:

src/
├── config.py    # Pydantic-based configuration
├── molecule.py  # Molecule representation
├── engine.py    # ORCA execution and scratch management
├── runner.py    # Parallel workflow execution
└── cli.py       # Command-line interface

Key improvements in 3.0.0:

• 27% code reduction (1,775 → 1,303 lines)
• Flat configuration structure (no unnecessary nesting)
• ProcessPoolExecutor for parallel execution
• Content hashing for cache validation
• Loguru for structured logging

Migration from 2.0.x

Version 3.0.0 is backward compatible. Legacy config format is automatically converted:

Old format (still works):

[main]
cpus = 4

[molecules]
directory = "molecules"

[[step]]
name = "opt"

New format (recommended):

cpus = 4
molecules_dir = "molecules"

[[stages]]
name = "opt"

No changes required to existing workflows.

Examples

Basic Optimization and Frequency

output_dir = "results"
molecules_dir = "molecules"
cpus = 4
mem_per_cpu_gb = 2

[[stages]]
name = "opt"
simple_keywords = ["OPT", "B3LYP", "def2-SVP"]

[[stages]]
name = "freq"
simple_keywords = ["FREQ", "B3LYP", "def2-SVP"]
keep = ["*.gbw"]

High-Level Single Point

output_dir = "results"
molecules_dir = "molecules"
cpus = 8
mem_per_cpu_gb = 4
workers = 2

[[stages]]
name = "opt"
simple_keywords = ["OPT", "TPSS", "def2-SVP", "D4"]

[[stages]]
name = "sp"
simple_keywords = ["DLPNO-CCSD(T)", "def2-TZVPP", "def2-TZVPP/C"]
blocks = ["%scf maxiter 200 end"]
keep = ["*.gbw"]

Selective Processing

output_dir = "results"
molecules_dir = "molecules"
cpus = 4

# Only process specific molecules
include = ["benzene", "toluene", "phenol"]

[[stages]]
name = "opt"
simple_keywords = ["OPT", "PBE0", "def2-TZVP"]

XYZ File Format

Molecule files should include charge and multiplicity in the comment line:

JSON format (recommended):

12
{"charge": 0, "mult": 1}
C    0.000000    0.000000    0.000000
H    1.089000    0.000000    0.000000
...

Simple format:

12
charge=0 mult=1
C    0.000000    0.000000    0.000000
H    1.089000    0.000000    0.000000
...

Development

Dependencies:

• click - CLI framework
• pydantic - Configuration validation
• loguru - Structured logging
• jinja2 - SLURM template rendering
• toml - Configuration parsing

Testing:

uv run python -m src.cli --version
uv run python -m src.cli init --slim

Project structure:

• src/ - Main codebase (3.0.0)
• orcastrator/ - Legacy codebase (2.0.x, deprecated)
• test/ - Integration tests

License

MIT License - see LICENSE file for details.