Checkpoint (runtime/executor.py)

Executor orchestrates scientific workflow nodes, handling execution, caching, and feedback.

# Source: https://github.com/beita6969/ScienceClaw | License: MIT | Copyright 2025 Peter Steinberger
# Provided via justclaw.it.com; keep this notice with the code.
"""Execution-guided orchestration runtime: Execute_{D_E}(G, chi, a) -> (G', chi', f)  (paper Eq. 7).

* :class:`Checkpoint` chi maps node fingerprints to completed :class:`NodeRecord` s whose output values
  are pickled under ``<values_dir>/<fingerprint>/<port>.pkl``. After an edit only nodes whose
  fingerprint is not in chi -- i.e. edited nodes and their descendants -- are (re)executed.
* :class:`Executor` executes a :class:`WorkflowGraph` in topological waves (independent nodes of a
  wave run concurrently, up to ``max_parallel_subprocs``):

  - ``tool``     trusted D_E adapter function, in-process, with a timeout guard thre

... (truncated -- full source via MCP)