fix(#40+voice+upload): SSOT scorecard parity + Sarvam 30s-truncation + uploaded-PDF retrieval gate

fix(pricing+scoring): rebuild premium engine to 0 logical violations + classification, faithfulness, links, writing

fix(reco): cap 3, exclude non-extracted policies, unify premium, ask PED

chore(cleanup): purge stale narrative/tombstones/dead code — codebase reads as the current standard

fix(retrieval): structural prose⇄card 1:1 gate (KI-PROSE-CARD-1to1)

fix(retrieval): quality-seeded candidates + wide recall pool — surface the genuinely-best policy (KI-QUALITY-SEED / KI-RECALL-POOL)

fix(scorecard): rec-path scorecard == marketplace (KI-FULLFACTS) — restores recommendation cards

recovery: integrate stalled-session work + de-stale/cleanup (pytest 215 green) [build-fix]

deploy#2: unified reco-fit gate + session-privacy fix + profile-completeness + full UI rebuild

data+scoring: verbatim-source all policy_facts, recalibrate scorecard, fix recommendation

feat(profile+pricing+voice): KI-275 — F-bundle: smoker + recap-verify + premium scope + price consolidation

feat(profile+pricing): KI-269 — D1+D2 bundle: chip merge + panel pre-fill + copay + family history

refactor: KI-268 — Path B cleanup: -5,219 LOC legacy + nim_fallback + chain collapse + modal wiring

feat(profile+pricing): KI-259..KI-266 — B-bundle (compare modal + per-policy + scorecard + premium band + budget/SI + union + persistence + recall)

fix(closure+hydration+voice+admin): KI-248..KI-251 — V1-V4 bundle from Playwright smoke

fix(single_brain): KI-244 — Z4 hardening — max_tokens, mark_recommendation gates, sticky-graceful retry

fix(arch+wiring): KI-237..KI-241 — Phase Y bundle (Y1+Y2+Y3+Y4 + profile-gate)

fix(arch+robustness): KI-225..KI-235 — Path B single-LLM rewrite + audit bundle