README.md

kairos-planner

Kairos in Python: a temporal and numeric PDDL planner (PDDL 2.1 to 3.1 and PDDL+), with its validator. The wheel also installs the kairos command.

pip install kairos-planner

import kairos

domain = kairos.load_domain("domain.pddl")
problem = kairos.load_problem("problem.pddl", domain)

result = kairos.Planner().solve(problem, timeout=10)
if result.status == kairos.Status.SOLVED:
    print(result.plan)                     # IPC format; result.plan.steps for the steps
    print(kairos.validate(problem, result.plan).valid)
elif result.status == kairos.Status.UNSOLVABLE:
    print("no plan exists (proved)")

• Anytime: Planner().plans(problem, timeout=30) yields each better plan as it comes; Planner().start(problem, on_plan=...) runs the search in its own thread (wait_first, next_plan, best, stop, wait).
• Options as keywords (timeout, memory_limit_mb, anytime, engine="m3", threads, epsilon, stop_after_no_improvement, target_cost, ...) or as a SolveOptions object.
• Statuses: SOLVED, UNSOLVABLE (proved), NO_PLAN_FOUND (no engine that applies is complete for the task: likely none exists, but nothing proves it), TIMEOUT, OUT_OF_MEMORY, UNSUPPORTED, ... result.diagnostics says why there is no plan.
• PDDL errors raise kairos.PddlError; each of its diagnostics has a stable code, a message and a location (file, line, column).
• Checking and validation: check_domain, check_problem, validate, validate_plan, parse_plan, pddl_support().
• Planning releases the GIL (several threads can plan at once) and Ctrl-C stops it.