feat(acms): add strategy coordinator and fusion engine #671

Merged

freemo merged 1 commit from feature/m6-acms-fusion-engine into master

2026-03-10 15:15:32 +00:00

freemo commented

2026-03-10 07:42:31 +00:00

Owner

Summary

Add StrategyCoordinator facade over Phase 1 pipeline components (ConfidenceWeightedSelector, ProportionalBudgetAllocator, ParallelStrategyExecutor) with per-strategy max cap enforcement and circuit-breaker error handling
Add FusionEngine facade over Phase 2 pipeline components (ContentHashDeduplicator, MaxDepthResolver, WeightedCompositeScorer, GreedyKnapsackPacker) with budget overage guard
Add BDD tests (15 scenarios, all passing), Robot integration tests (9 tests, all passing), ASV benchmarks, and reference documentation

Key Changes

StrategyCoordinator.coordinate() — single API for strategy selection, proportional budget allocation, and parallel execution
FusionEngine.fuse() — single API for dedup by UKO URI + content hash, max-depth conflict resolution, weighted composite scoring, greedy knapsack packing with deterministic tie-breakers
Budget overage guard drops lowest-relevance fragments when total exceeds budget
Per-strategy max caps clamp allocations and redistribute excess
Coverage: 98.5% (threshold: 97%)

Files

File	Purpose
`src/.../strategy_coordinator.py`	StrategyCoordinator facade
`src/.../fusion_engine.py`	FusionEngine facade
`features/acms_fusion.feature`	BDD scenarios
`features/steps/acms_fusion_steps.py`	Step definitions
`robot/acms_fusion.robot`	Robot integration tests
`robot/helper_acms_fusion.py`	Robot helper
`benchmarks/acms_fusion_bench.py`	ASV benchmarks
`docs/reference/acms_fusion.md`	Reference documentation

ISSUES CLOSED: #192

## Summary - Add **StrategyCoordinator** facade over Phase 1 pipeline components (ConfidenceWeightedSelector, ProportionalBudgetAllocator, ParallelStrategyExecutor) with per-strategy max cap enforcement and circuit-breaker error handling - Add **FusionEngine** facade over Phase 2 pipeline components (ContentHashDeduplicator, MaxDepthResolver, WeightedCompositeScorer, GreedyKnapsackPacker) with budget overage guard - Add BDD tests (15 scenarios, all passing), Robot integration tests (9 tests, all passing), ASV benchmarks, and reference documentation ## Key Changes - `StrategyCoordinator.coordinate()` — single API for strategy selection, proportional budget allocation, and parallel execution - `FusionEngine.fuse()` — single API for dedup by UKO URI + content hash, max-depth conflict resolution, weighted composite scoring, greedy knapsack packing with deterministic tie-breakers - Budget overage guard drops lowest-relevance fragments when total exceeds budget - Per-strategy max caps clamp allocations and redistribute excess - Coverage: 98.5% (threshold: 97%) ## Files | File | Purpose | |---|---| | `src/.../strategy_coordinator.py` | StrategyCoordinator facade | | `src/.../fusion_engine.py` | FusionEngine facade | | `features/acms_fusion.feature` | BDD scenarios | | `features/steps/acms_fusion_steps.py` | Step definitions | | `robot/acms_fusion.robot` | Robot integration tests | | `robot/helper_acms_fusion.py` | Robot helper | | `benchmarks/acms_fusion_bench.py` | ASV benchmarks | | `docs/reference/acms_fusion.md` | Reference documentation | ISSUES CLOSED: #192

Rows
Columns