m/market labResearch, with a record.
Source on GitHub ↗

Ordered experiment execution

This package runs the shared background flow for Live Human/Rule and Agent experiments. It does not contain a Market process or the Agent decision loop.

File Executes where Responsibility
session.py App owner thread and command ingress ExperimentSession admits commands, tracks lifecycle and processes the financial timeline.
timeline/ App owner and ingress Sequencer fixes receipt time/order; ExecutionTimeline merges market events and commands.
market.py App owner LocalMarketExecution borrows the financial session; ExecutionCommands delegates the existing transaction.
controls.py App caller Coordinate clock, Runtime and participant Play/Pause/Stop acknowledgements.
delivery.py App output thread Resolve deferred results, record JSON and deliver ordered replies.
results.py Shared value/async boundary DeferredResult and result-slot behavior.
metrics.py App Measurements of processing lag and admitted work.

ExperimentSession is held by the method-specific runner. Its owner thread starts on Play; output starts during preparation. Financial effects use the existing Coordinator → Service → ledger path. No alternative submit/cancel implementation or tool-receipt SQL is introduced.

The financial ExecutionSession remains in execution/session.py. It owns the run's financial execution, while ExperimentSession here coordinates the background event flow. Market visibility comes from the shared clock and immutable prepared reader. Market reads bypass the financial queue; account/order reads use the owner for consistent mutable state. Tools formatting runs in its child.

Maintained source: sim_platform/experiments/execution/README.md