One state, two futures
SimPy resumes the original schedule and the HiGHS candidate from the same snapshot. Completed operations stay frozen; work in progress keeps its machine and remaining duration.
01 · DECISION WORKFLOW
A machine outage becomes a proposed schedule, a fair comparison and an explicit operator decision. The solver and simulation carry the numbers; the workflow carries the state.
At hour 2, a machine fails while a job is partly fabricated. Replacing the entire schedule would erase completed work and move an operation that has already started. A useful planner must preserve both commitments before improving the remaining horizon.
SimPy resumes the original schedule and the HiGHS candidate from the same snapshot. Completed operations stay frozen; work in progress keeps its machine and remaining duration.
Compare completion, unfinished units, overtime and the timeline before confirming. Advancing consumes the selected proposal, then a second outage starts a new comparison from the resulting state.
The solver runs in an isolated process. A failed or timed-out solve publishes no proposal and does not advance execution. Infeasible or worse candidates keep the original recommendation; stale confirmations cannot apply a proposal to a different snapshot.
The public page replays a recorded, authored scenario. To edit an outage and run the actual solver, start the local demo after installing the repository requirements:
python -m mpa.demo --port 8840This is a four-unit synthetic exercise on a bounded time grid. It establishes the state transition and comparison contract, not factory throughput gains or LLM planning quality. There is no connection to shop-floor equipment.
Source and CI links let you inspect the implementation and reproducibility. Passing checks do not establish deployment success or business gains.