Task routing layer for assigning work to agents, workers, or tools based on task type, context, rules, and execution constraints.
- classifies incoming requests into routing lanes such as implementation, audit, integration, documentation, and research
- matches requests against worker strengths, aliases, specialties, and project affinity
- applies cheap-first execution policy hints where appropriate, while allowing direct premium routing for higher-synthesis research/spec work
- supports overflow rebalancing when preferred workers are busy or overloaded
- produces a structured task brief with routing rationale and execution guidance
router.py- routing engine and brief generatorworkers_map.json- worker registry, affinities, aliases, and specialtiestest_router.py- deterministic smoke-style routing tests
This staging copy is standalone for classification, ranking, overflow selection, and brief generation.
Optional runtime-aware load checks are supported, but they degrade safely when an external runtime.py is not present. That means the public repo still works as a self-contained routing component.
python -m py_compile router.py test_router.py
python test_router.py
python router.py- The project currently uses only Python standard library modules.
- Runtime integration with a larger swarm/control-panel environment is optional, not required for local testing.