PyVAL — Pure Python PDDL Plan Validator

A pure-Python PDDL plan validator — a drop-in alternative to the compiled VAL binary. Built on unified-planning, zero compiled dependencies, pip install-able everywhere.

PyVAL produces rich, structured diagnostics designed for both humans and LLM consumption: per-step state changes, precondition deficit reporting, and repair-oriented messages.

Install

pip install pddl-pyvalidator

Quick Start

# Validate a plan against a domain and problem
pyval domain.pddl problem.pddl plan.txt

# JSON output for tool integration
pyval domain.pddl problem.pddl plan.txt --json

# Numeric fluent trajectory (one row per step)
pyval domain.pddl problem.pddl plan.txt --trajectory --track fuel --track cost

# Syntax check only (domain, or domain + problem)
pyval domain.pddl
pyval domain.pddl problem.pddl

Exit code is 0 on a valid plan, 1 otherwise.

Python API

from pyval import PDDLValidator

validator = PDDLValidator()
result = validator.validate(
    domain_path="domain.pddl",
    problem_path="problem.pddl",
    plan_path="plan.txt",
)

if result.is_valid:
    print("Plan is valid")
else:
    for err in result.errors:
        print(err)

What Gets Validated

PyVAL runs a three-phase pipeline and halts on the first fatal error:

1. Syntax & semantic — PDDL parses, types/predicates/functions are well-formed, no duplicates or arity mismatches.
2. Plan structure — every action name exists in the domain, parameters are declared objects with type-compatible (subtype-aware) bindings.
3. Plan execution — each step's preconditions hold, effects apply, and goals are satisfied in the final state. Numeric fluents are tracked across steps; quality metrics (minimize / maximize) are evaluated on the final state.

Unsupported PDDL features (durative actions, PDDL+ processes/events) are reported as warnings rather than silent failures.

Output Modes

• Plain text (default, VAL-like, optionally -v for per-step detail)
• JSON (--json) — structured for programmatic consumption
• Trajectory (--trajectory) — numeric fluent values per step, optionally filtered via --track

Development

python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
pytest

See CLAUDE.md for architecture notes and the validation pipeline overview.

License

MIT — see LICENSE.