Skip to content
planning-and-learningPublic

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

88 Commits

Folders and files

Repository files navigation

pyrunir-mcp

pyrunir-mcp is a typed Python API for planning-and-learning agents. Callers keep domains, tasks, candidates, validation results, and history in memory; filesystem artifacts are written only at the dump boundary.

See docs/api.md for the API and docs/index.md for workflow/output tables.

Basic Flow

from pyrunir_mcp import (
    ValidationHistory,
    create_domain_context,
    create_policy,
    create_task_context,
    dump_result,
    find_solution,
    generate_tasks,
)

domain = create_domain_context("domain.pddl")
task = create_task_context(domain, "problem.pddl")
policy = create_policy(domain, "policy.txt")
history = ValidationHistory()

result = find_solution(task, policy)
feedback = history.fold(result.observation)
dump_result(result, "artifacts/find-solution")

Entry Points

Context/candidates:

  • create_domain_context(domain_file)
  • create_task_context(domain_context, problem_file, num_threads=1)
  • create_policy(domain_context, policy_file); pass None for an empty policy
  • write_empty_policy(domain_context, policy_file) for canonical empty policy text
  • create_module_program(domain_context, module_program_file); pass None for the built-in empty module program
  • create_classifier(domain_context, classifier_file); pass None for an empty classifier

runir.task_generation:

  • describe_generator(domain_name)
  • get_generator_path(domain_name) / get_generator_domain_path(domain_name)
  • generate_tasks(domain_name, output_dir, batch_name, configs, allow_invalid=False)

Validation:

  • find_solution(task_context, policy_or_module_program, classifier=None, universal=False, ...)
  • prove_termination(domain_context, policy_or_module_program, ...)
  • prove_classifier(task_context, classifier, ...)

Dumping:

  • dump_result(result, output_dir, formats=(DumpFormat.JSON,), include_witness=True, include_witness_trace=True, include_plan_trace=True, include_successors=True)
  • dump_validation_history(history, output_dir, formats=(DumpFormat.JSON,))

DumpFormat.PSV, DumpFormat.MD, and DumpFormat.JSON request LLM-readable PSV, human-readable Markdown, and machine-readable JSON. Validation calls do not write files.

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages