Changelog

June 2026

lexarna.ai — new home domain

Lexarna's primary domain is now lexarna.ai. All lexarna.com URLs 308-redirect with path preserved, so existing bookmarks and shared links land in the right place. Clerk auth re-issued on the new domain; existing users sign in once and continue.

• Primary domain switched to lexarna.ai across Vercel, Railway, and Clerk
• lexarna.com → lexarna.ai 308 redirect on every path and subdomain
• Email DKIM / DMARC re-set on the .ai zone for transactional deliverability

June 2026

Document Intelligence v2 — strategy memo + contradictions + cross-corpus

The biggest single shipping moment in Lexarna's history. Smart folders now run a full case-intelligence pipeline at ingest: per-doc map-reduce summaries on a cheap model, two-layer contradiction detection (cheap-pass on every upload + on-demand LLM verifier), entity + timeline extraction, and a synthesised strategy memo with five structured sections.

• Strategy memo synthesis — case theory, strongest arguments, weakest arguments, predicted opponent attacks, open questions
• Two-layer contradiction detection: instant cheap-pass on ingest + LLM-verified on click
• Cross-corpus research — matter docs and binding precedent ranked into the same answer, with [MATTER] vs [CASELAW] provenance tags
• Per-doc map-reduce summaries replace re-retrieval for routine 'what does Ex-A say' questions
• GenericDocChunker handles arbitrary matter-uploaded case files (FIRs, bail apps, charge sheets) instead of forcing them through the caselaw chunker

June 2026

MAX thinking mode — agentic decomposition

MAX mode in research replaces the old 'Brief' depth with genuine agentic decomposition. Query analysis splits the question into 4–5 sub-angles, retrieval runs for each in parallel, and the answer addresses every angle explicitly under structured headings.

• Old single-pass deep research → new multi-angle synthesis
• Sub-question retrieval in parallel with deduplication across the result sets
• Structured output: # Statutory position / # Case law / # Procedural posture / # Vulnerabilities / # Strong hooks

June 2026

Firm Memory — Lexarna learns across matters

Every cited research turn is distilled into a one-line memory and stored at user, org, or matter scope. Future questions automatically pull relevant prior insights — preferred citation style, judge-specific framing, recurring counter-arguments — into the system prompt by semantic similarity. Auditable and revocable.

• Three scopes: user (personal), org (firm playbook), matter (case-specific)
• Matter scope outranks the others when surfacing on a scoped query
• PII redacted before storage; short / uncited turns skipped
• Settings → Privacy lists every entry with single-row delete + scope-wide clear

June 2026

Library v2 — smart folders + default folders

Matters and the old document library are unified under a single Library. Smart folders run the full intelligence pipeline; default folders are plain storage with semantic search. Promote a default folder to smart any time without re-uploading.

• Sidebar rename: 'Matters' → 'Library' (shortcut M preserved for muscle memory)
• Browse is now a top-level nav: keyword-only search over the canonical corpus, no AI generation
• Tour + checklist rewritten to cover the new structure and Intelligence tab

June 2026

Onboarding refinements — working toggles, profile-aware prompts

Loop closure on the May onboarding overhaul. The consent acknowledgement modal's toggles are now actually interactive (they were read-only previously). The 3-question personalisation step now wires into research + draft system prompts, so practice area and primary court actually shape the LLM's framing.

• Consent modal: pill-style switches wired to PUT /privacy/consent/:key with optimistic UI
• Onboarding profile (practice area, role, primary court) read by useResearchStream and useDraftStream, passed in every request, rendered as a soft-hint block in the system prompt
• Profile data stays in Clerk metadata — no DB sync, no admin-API roundtrip

June 2026

Onboarding overhaul + auth resilience

First-run experience rebuilt as a single chain: a minimal welcome moment, DPDP consent acknowledgement, a 3-question personalisation step, a spotlight tour over the workspace, and a demo matter to explore. A persistent get-started checklist tracks the next five actions.

• New welcome overlay, consent ack, personalisation step, guided spotlight tour, demo matter, get-started checklist
• API client resilience: token cache properly resets on clear, transparent 401 retry with token refresh, on-visibility re-sync — no more random 'failed to load' after a tab idles

May 2026

Pricing v2 — Solo tier, credit packs, rollover, Razorpay

Pricing rebuilt to remove the seat-tier mismatch that pushed solo practitioners into Starter. New Solo plan at ₹1,499, credit packs from ₹499, 50% credit rollover on every paid plan, Razorpay integration wired end-to-end, and auto-recharge on Pro / Max.

• New Solo tier (₹1,499 / 400 credits) for light users and students
• Credit packs: 100 / 500 / 2,000 / 10,000 — pay-as-you-go top-up at marginal cost
• 50% credit rollover at cycle renewal — no more burnt unused credits
• Auto-recharge: optionally buy a top-up pack when balance drops below threshold
• FEATURE_COSTS rebalanced for May 2026 LLM prices — research drops from 1/3/15/50 to 1/3/8/20
• Team plans: 3-seat minimum, flat per-seat pricing — no penalty for being small

April 2026

Cross-matter recall + hearing badges + takeaway docs

Hearing Prep got working memory across matters. Research you ran two months ago on the same fact pattern resurfaces when you open a new matter; upcoming hearings carry status badges; one-pager takeaways generate automatically.

• Cross-matter recent sessions surface in the new matter's research panel
• Hearing badges at a glance: research ready / draft pending / evidence missing
• Takeaway documents auto-generated from session history for next hearing

April 2026

Courtroom Simulator — argument practice

Practice arguments against a critical AI bench tuned for Indian procedure. Soft, Standard, or Hostile modes; counter-argument library maps each objection to its motivating precedent; argument scoring gives a weakness map you can act on before the actual hearing.

March 2026

Privacy infrastructure rebuild — DPDP-ready

Comprehensive privacy infrastructure built end-to-end for DPDP Act 2023: explicit consent management, per-category toggles, granular audit log, self-service data export and erasure, sub-processor registry, Sentry PII scrubbing, and per-user storage scoping.

• DPDP consent acknowledgement modal — informed consent on first login
• Self-service erasure: complete account + content deletion from settings
• Sub-processor registry published; Sentry events scrubbed of PII
• Per-user localStorage isolation — switching Clerk users wipes prior content from the browser

February 2026

Three-layer document intelligence

Document intelligence rebuilt around three layers: DocTree (structural), semantic (embeddings), and cross-doc (contradiction detection). Surfaces inconsistencies between, e.g., a plaint and the supporting sale deed.

• OCR + structured extraction preserving tables and clause structure
• Cross-document contradiction detection across an entire matter
• Hearing pack assembly — one-click court-ready bundle with index and pagination
• Clause-level redlining and comparison between document versions

January 2026

Reasoning modes — pick your depth

Every AI operation now exposes four reasoning modes (Quick / Standard / Advanced / MAX). Default mode tuned per feature so casual queries don't burn deep-research credit.

December 2025

BNS + IPC dual-era support

Every retrieved passage and every citation is now tagged by legal era — BNS / BNSS / BSA versus IPC / CrPC / IEA — so you can't accidentally cite a repealed provision in current proceedings. Era filtering is one click.

November 2025

Matter workspaces

Lexarna moved from a search bar to a workspace. Every case lives in its own matter with isolated documents, research sessions, drafts, hearings, and team — and quick context-switching between them.