Changelog
MavenFin uses calendar-anchored semantic versioning. Breaking changes ship under a new namespace
(isrestapi19 → isrestapi20); non-breaking additions land in-place and are
documented below.
v4.19 — current
Released 2026-05-12. Production. SLO: 99.9% availability, P95 projection < 800 ms.
- Added.
POST /households/{id}/plans/{planId}/roth-optimize— three strategies (thresholdRoth,bracketFill,dynamic), IRMAA-tier aware. - Added.
seedparameter on the Monte Carlo endpoint for deterministic reruns. - Added.
granularity=monthlyon the projection endpoint (default remains yearly). - Improved. Federal tax engine re-validated against parity reference across 7,968 synthetic households — max delta < $0.01.
- Improved. Social Security WEP/GPO reductions now applied before the family-maximum cap, matching SSA POMS RS 00605.
- Improved. Refresh-token TTL extended from 24h to 7 days.
- Fixed. RMD timing for plan-end years (off-by-one when birthday fell after RMD start).
- Fixed. Pension cash-flow for survivors with simultaneous WEP-affected own benefits.
v4.18
Released 2026-02-04. Maintenance.
- Fixed. Off-by-one error in tax-year boundary handling for plans with January retirement dates.
- Improved. State-tax module now covers all 50 + DC; previously missing OR, NM, AK.
- Deprecated.
/api/v0/*legacy endpoints — sunset planned for v5.0.
v4.17
Released 2025-11-20. Feature drop.
- Added. AMT and QBI deduction fields on the tax response.
- Added. Asset-allocation strategies expanded from 5 → 12 presets.
- Improved. Monte Carlo runtime cut roughly 4× via vectorized random-walk generation.
v4.16 and earlier
Pre-v4.17 changes are tracked in our internal release notes. If you're integrating against a major older than 4.17, reach out to info@mavenfin.tech for the migration guide.
Versioning policy
Major bumps reserve the right to break wire compatibility — schema renames, removed
fields, semantic shifts. They ship under a new path namespace
(isrestapi19 → isrestapi20) so existing integrations keep working until you
opt in.
Minor bumps add fields, add endpoints, or add new optional parameters. They never remove anything. Your existing requests keep working untouched.
Patch bumps fix bugs and improve performance. We treat parity-reference bug fixes as breaking-by-numeric-output even though the schema is stable; those land at minor or major boundaries.