sequentialCrossingHorizon is declared in dist/sequential-*.d.ts and documented as the best-case reachability preflight for pairedEvalueSequence, but it is not re-exported from the package root: on 0.183.0, import { sequentialCrossingHorizon } from '@tangle-network/agent-eval' is undefined, while pairedEvalueSequence, minimumPairsForPairedDeltaTest, requiredPairedSampleSize and pairedMde are exported.
A decidability gate (discovery docs/09-failure-modes.md mode 22) has to simulate the all-at-bound best case to find the decision horizon before a sequential design is allowed to spend. Without the export, every caller re-implements the simulation on top of pairedEvalueSequence, which is exactly the drift the function's own doc comment says it exists to prevent.
Reproduce:
npm i @tangle-network/agent-eval@0.183.0
node -e 'import("@tangle-network/agent-eval").then(m => console.log(typeof m.sequentialCrossingHorizon, typeof m.pairedEvalueSequence))'
# undefined function
Expected: function function. Re-export it from src/index.ts beside pairedEvalueSequence.
Observed by simulating it by hand on 0.183.0 with alpha 0.05: evidence at 22 pairs, decision at 24, for any bound, matching the mode-22 record.
sequentialCrossingHorizonis declared indist/sequential-*.d.tsand documented as the best-case reachability preflight forpairedEvalueSequence, but it is not re-exported from the package root: on 0.183.0,import { sequentialCrossingHorizon } from '@tangle-network/agent-eval'isundefined, whilepairedEvalueSequence,minimumPairsForPairedDeltaTest,requiredPairedSampleSizeandpairedMdeare exported.A decidability gate (discovery
docs/09-failure-modes.mdmode 22) has to simulate the all-at-bound best case to find the decision horizon before a sequential design is allowed to spend. Without the export, every caller re-implements the simulation on top ofpairedEvalueSequence, which is exactly the drift the function's own doc comment says it exists to prevent.Reproduce:
Expected:
function function. Re-export it fromsrc/index.tsbesidepairedEvalueSequence.Observed by simulating it by hand on 0.183.0 with alpha 0.05: evidence at 22 pairs, decision at 24, for any bound, matching the mode-22 record.