What's New in RemitRix

Switch on Agent Mode in the chat and type a plain-English command — the platform plans it into UI steps and executes them on the real screens, spanning multiple pages in one instruction.

• One natural-language command becomes a client-executed plan against the live interface
• Auto-continue loop: a single command can select a grid row, apply it, and complete the dialog that opens
• Executing a scenario or downloading a report runs end to end, through modals included
• Confirmation gate on destructive steps, with a full audit trail of every action
• Live step-by-step progress with an 'Agent working' status across the whole run

A new Scenario Generator popup compares a portfolio's risk-factor coverage against a chosen report date and scenario, so gaps are visible before a run — not after.

• Three views: By Assets, By Risk Factors, and By IR & Spreads
• Colour-flag cells mark each factor as Exists, Not relevant, or Missing
• Drill from any market-value cell straight into the underlying assets
• Grid filters mirror the Pivot report; Excel export built in
• Full Hebrew/RTL support with column labels driven by the data source

The application now carries the new RemitRix wordmark and a single brand font stack — IBM Plex Sans for UI and IBM Plex Mono for numbers — matching the design system across every screen.

• New wordmark and favicon applied across sidebar, header, and sign-in surfaces
• IBM Plex Sans as the default UI font, with a Hebrew companion for RTL
• IBM Plex Mono standardized for all numeric and data display

Record a real multi-screen flow as you work and the platform turns it into a reusable Skill that replays by driving the actual application.

• One-click recorder captures your clicks, inputs, and navigation as you work
• Recorded flows are saved as Skills alongside the rest of your skill library
• Replay drives the real screens step by step, with live progress

A new Personalization screen in the AI Hub runs a full adaptation loop: the platform captures genuine usage signals, proposes preference changes, and applies the ones you approve.

• Learns from real behaviour — for example, reshaping a card from chart to table is remembered
• Proposals surface automatically when you open the panel; approving one persists the preference
• Editable preferences with confidence meters and usage KPIs
• Backed by a dedicated adaptation engine and memory substrate with a full audit trail

Every VaR and scenario report is now scanned by a rule-based anomaly engine — detection runs entirely in code, never guessed by an AI — with findings surfaced in the report UI and explorable in chat.

• Rule catalog covering reconciliation breaks, stale runs, masking, coverage gaps, and dominance patterns
• Severity-coloured badge on the report's AI Insights button with a detailed findings view
• Anything that cannot be checked is reported as a data gap — never silently passed
• Thresholds are client-editable from a new admin tab, with an audit trail and no restart needed
• AI chat interprets the deterministic findings and offers grounded follow-up questions

Five premium in-app guides — IC Pipeline, Liquidity Risk, Skills, Workflows, and Risk Calibration — now share one polished reading experience.

• Searchable table of contents with scroll-spy that tracks where you are
• Draggable guide window, reading-progress bar, and print / save-as-PDF
• English and Hebrew editions with full RTL support

AI-generated stress scenarios now pass through a deterministic reasonableness check before use: each shock is banded against calibrated severity bounds and cross-factor consistency rules, and the verdict is shown with the scenario.

• Each shock banded Normal / Stress / Tail / Implausible against per-family bounds
• Cross-factor consistency rules catch contradictory combinations of shocks
• One reconciled verdict — Defensible, Needs Review, or Not Defensible — surfaced in chat
• Severity thresholds are client-editable from the Risk Configuration screen, with audit history

VaR runs replace the generic spinner with a staged progress tracker driven by the real calculation job, so you can see exactly where a computation stands.

• Stages from Queued through Loading assets, Market data, Scenarios, and Saving
• Progress polled live from the backend job from the moment the run starts
• Percent climbs through each stage instead of jumping at the end

Share a full dashboard — all its cards — with selected users directly from the dashboard tab menu, with a sender-chosen name for the copy the recipient gets.

• Share any dashboard tab from its context menu
• Recipient's dashboard name is editable at share time
• Access hardened with recipient validation and per-user authorization checks

A visual drag-and-drop builder in the AI Hub chains Skills into multi-step pipelines that run on a trigger — manually, from a chat keyword, or on a schedule.

• Node types include skills, prompts, conditions, parallel branches, and merges
• Triggers: manual run, chat keyword match, or scheduled cron
• Runs execute wave-by-wave across the pipeline graph with a live run panel
• Undo/redo, validation, dark mode, and an in-app user guide (EN/HE)
• Run history is scoped to the requesting user

A new top-level AI Hub groups the platform's AI surfaces — an embedded full-page Chat, the Skills library, and Usage — behind a single landing screen.

• The AI chat gains a docked full-page mode alongside the floating window — same chat, no duplication
• Skills moved into the hub with consistent role-based access
• Landing grid with direct tiles into each AI screen

The Solvency chat now answers regulation questions from a retrieval-grounded knowledge base, while SCR portfolio analysis moves behind dedicated commands and official skills.

• Regulation Q&A grounded in the Solvency knowledge base, with cross-lingual retrieval (ask in English, retrieve in Hebrew)
• /scr analysis command for a single-date SCR breakdown by module with diversification benefit
• Official SCR Analysis and SCR Comparison skills, reusable from chat and workflows
• Auto-orchestration applies the detected portfolio filter (e.g. Nostro) and shows it before running

Interest-rate and spread curves in Market Data moved to the platform's standard charting engine, with more reliable rendering and consistent behaviour under rapid selection changes.

• IR/Spread curves rendered on the same engine as the rest of the app's charts
• Out-of-order responses guarded so fast selections never show mismatched data
• Smoother hover and add-series interactions