wevibe-sim (internal research tool)

Known policy drift — read first: before using this harness as evidence, read KNOWN-DRIFT.md. The sim's policy constants and duplicated standing/decay logic have drifted from production. Simulation output must not be read as evidence about the shipped product policy until those recorded drift items are fixed.

wevibe-sim is an internal, local-only simulation workspace used to test retrieval/decay behavior before chain + hub changes. It contains two independent systems:

1) Legacy Decay/Ranking Sim (D-4.2 calibration)

• Lives at the repo root (sim-runner.js, ranking-fix.js, iterate.js, final-analysis.js, suppress-idle-analysis.js, unique-floor.js) plus the Next.js visualization shell (app/, next.config.js).
• Purpose: calibrate and compare decay/ranking dynamics for D-4.2-era mechanisms.
• Load-bearing note: ranking-fix.js is the source for the ported ranking logic referenced by wevibe-server/wevibe-hub/internal/retrieval/retrieval.go.

Quick start:

npm run dev
node sim-runner.js
node ranking-fix.js

See DECAY-SIM.md for script-by-script notes.

2) Recall-Alignment Simulation Suite (recall-sim/)

• Purpose: measure D-RECALL-ALIGNMENT behavior across model tiers and ablation cells.
• Lives entirely under recall-sim/ (lib/, pipeline/, corpus/, eval/, config.mjs, data/, results/).
• Includes hard stage watchdogs, per-call timeouts, and spawn budgets so sim stages do not run unbounded.

See recall-sim/README.md for architecture, config knobs, and run instructions.