Recursim is Recurgent's simulation framework for measuring and shaping agent behavior with repeatable evidence.
Use this folder when you need to answer:
- what a simulation pack means,
- why deterministic fixture/replay runs exist,
- what live-shadow actually does,
- how to evolve an agent (for example calculator) in disciplined loops.
docs/recursim/glossary.mddocs/recursim/deterministic-replay-live-shadow.mddocs/recursim/agent-shaping-mechanisms.mddocs/recursim/tutorial-evolving-scientific-calculator.mddocs/simulation-readiness.mdfor full operator command reference and gate diagnostics.
This documentation is about simulation mechanics and learning loops. It complements but does not replace:
- ADR intent in
docs/adrs/0027-simulation-preparedness-and-readiness-gates.md, - live-shadow design in
docs/adrs/0028-live-shadow-simulation-and-dual-lane-evidence.md, - implementation detail in
docs/plans/live-shadow-simulation-dual-lane-implementation-plan.md.