Changelog

Phase D — Pricing + Stripe + QuickBooks

• 3 V1 pricing tiers ported (Solo $149 / Boutique $899 / Enterprise custom)
• 4 billing endpoints (POST /v1/billing/checkout-session, /portal-session, /me, /v1/stripe/webhook)
• 5 QuickBooks endpoints (OAuth + P&L + Balance Sheet ingest into engagement_facts)
• 7 FE components (PricingCard, CheckoutButton, ManageBilling, QuickBooksConnect, QuickBooksSyncHistory, use-billing + use-quickbooks hooks)
• 3 smoke batteries (18 mock-mode Stripe + 31 mock-mode QuickBooks + 14 real-Stripe signed webhook = 63/63 new assertions)
• 7 pre-existing bugs caught + fixed (mock-Session attribute, ON CONFLICT clause, list_for_workspace uuid/text cast, etc.)

Phase C — PDF + Share + 7 client-facing deliverables

• 7 PDF generators lifted from V1 (ClarityMapPDF, BottleneckPDF, MarketSignalPDF, CompetitorBriefPDF, OperatingMemoPDF, RiskRegisterPDF, StakeholderBriefPDF)
• 3 public share-token routes (/share/:token, /s/:token, /summary/:token)
• public.shares table + RLS + 21/21 backend smoke assertions
• useExportPdf + useCreateShare + useRevokeShare + useListShares + useShareByToken hooks
• ShareButton + PdfDownloadButton + CreateShareModal + ViewHeaderActions

Tier 1 #9 — EM/Partner roles + QC inbox

• workspace_role enum (owner/partner/em/member) with rank-based permission helpers
• GET/POST/PATCH/DELETE /v1/workspaces/{id}/users
• GET /v1/qc-inbox + POST /resolve + POST /override + GET /overrides
• qbinbox_overrides audit log for the Partner-can-override-EM semantics
• 4 critique-gate graphs (4.1.A, 4.2.A, 4.1.B, 5.4.A) wired to write QC inbox rows on advisory warnings
• 35/35 role-helper smoke assertions

Phase O — Outbound transmission abstraction + 5 action wirings

• notification_service (notify_email + notify_slack + notify_calendar_invite + preview_notification)
• MockProvider that captures every payload to ~/.clarovo/notifications_mock/
• Importable Resend/SES/Slack-webhook/Ics stubs (raise NotImplementedError until real creds arrive)
• notifications audit table with ws_isolation RLS + 3 read endpoints
• 5 action wirings (0.1.A-OUTREACH, 1.2.B, 1.1.E, 5.1.A, 6.2.A)
• OutboundTransmissionsCard mounted on the engagement page

Phase B — All 11 V1 result views lifted to V2

• Clarity Map + Matrix + Stakeholder Sim + Pre-Mortem (B.1)
• Bottleneck + Market Signal + Competitor Watch + Benchmark + Scenario Engine + Burn Rate (B.2)
• Execution + FirmBrain (B.3 addendum)
• Each view: 1 read endpoint resolving view-slug -> Z.8 sub-agent, 1 reshape() function, 1 lifted V1 component, 1 route
• Mounted on the engagement page's DeepViewsCard (12 entries)
• 88 vitest reshape assertions + 15/15 backend smoke green

Phase A — V1 cognitive method restored

• 15 V1 session-type question sequences (PVT/EXT/TME/HIR/SNK/FND/BRN/REV/PRT/WAR/BOT/CHK/MKT/STK/SCN)
• 10 MBB-grounded questions per session type + a 1-line MBB principle
• Deterministic Consequence Engine Python scoring (regret normalization, decision-window parsing, sensitivity ranking, blind-spot fallback, pairwise winner)
• answer_interrogation + consequence_engine sub-agents post-process through this layer
• /method marketing page lifted from V1's MethodPage.tsx
• 25/25 A.* smoke assertions pass

Phase Z — Harness + 14 orchestrator-driven sub-agents

• 7-principle harness wrapper (~750 lines) — input validation, output validation, audit trail, rollback, cost/token budget, supervised gates, methodology coverage
• 14 sub-agents registered: situation_assessor, answer_interrogation, consequence_engine, avoidance_question, bottleneck_diagnostician, market_signal_analyzer, pre_mortem_runner, stakeholder_simulator, competitor_watcher, benchmark_puller, scenario_engineer, burn_rate_calculator, decision_memo_writer, clarity_mapper
• 5 orchestrator HTTP endpoints (run-next + sign-off + methodology-coverage + audit-trail + agent-runs/{run_id})
• Per-stage state machine + 7-stage cascade-advance
• 10 Langfuse observability events
• 25/25 Z.* smoke assertions pass

Phase E — Marketing surface (this work)

6 new public pages (case studies, changelog, docs, responsible AI, terms, privacy) + the shared MarketingLayout.

Worker / scheduler

Lightweight worker (cron / Celery beat / arq / Supabase pg_cron) for 1.3.A weekly hypothesis-tracker digest + 1.5.B DRL escalations + 6.2.A 30-60-90 cadence touchpoints. The system actually runs engagements on a schedule.

Langfuse dashboards

Per-action latency / token cost / retry rates. Cheap win for trust during demos; needs production traffic to be useful.