v1.09 — CRM Integration Core & Notification-Driven Eva

Sprint 22 · May 4 – May 10, 2026

Sprint 22 delivered the foundational CRM integration layer — the CRM connection domain entity and lifecycle model, field mapping configuration, and role-scoped visibility across Advisor, Firm, and Enterprise tiers. Eva gained a new notification-triggered chat capability ("Ask Eva About This"), letting advisors launch a contextual AI conversation directly from any platform alert. The sprint also flattened the left-nav to section labels for faster navigation, expanded host administration with Tenant ID user management and case-insensitive search, and resolved five platform issues.

Shipped

• CRM Connection Domain Entity & Lifecycle — the core CRM connection model is now a first-class platform entity with a full create/update/delete lifecycle and persistent state management.

• CRM Field Mapping Configuration — advisors and firm admins can now configure how platform fields map to their connected CRM, enabling flexible data synchronization without custom code.

• Role-Scoped CRM Visibility — Advisor / Firm / Enterprise — CRM data is now surfaced according to the three-tier authority model, ensuring each role sees exactly the records within their permitted scope.

• Eva: Notification-Triggered Chat ("Ask Eva About This") — any in-app notification now surfaces an "Ask Eva About This" action, opening a pre-seeded Eva conversation with full context from the triggering event.

• Left-Nav Modernization — Flattened Tier-1 Section Labels — the primary navigation was redesigned with flat section labels replacing expandable parent nodes, reducing click depth and improving at-a-glance orientation.

• Host Administration — Tenant ID, Users Page & Case-Insensitive Search — host-level admin pages now include a dedicated Tenant ID view, an expanded Users management page, and case-insensitive search across all tenant records.

• LLM Performance Evaluation Framework — the platform now integrates Windmill and Volt Agent evaluation frameworks to systematically benchmark AI model performance, enabling data-driven model selection and regression detection.

• Platform Code Quality Pass — CodeRabbit Review Resolution — 74 code-quality findings across 10 blocks from the CodeRabbit automated review were addressed, improving platform stability and reducing technical surface area.

Fixes

• Platform Launch Error Detail Lost Before AI Interpretation — error payloads are now captured in full and correctly passed to the AI interpretation layer, eliminating silent truncation.

• Google Search Smart-Bidding Fallback Not Firing — the smart-bidding fallback path now correctly activates when conversion tracking is not enabled.

• Meta Lead Form Rejecting Canadian & Australian Locales — en_CA, en_AU, and fr_CA locale codes are now accepted on Meta lead form creation, unblocking campaigns targeting Canadian and Australian audiences.

• [Marker.io] Predis Generation Error for Meta Single Image Campaign — image generation via Predis no longer errors when a Meta Single Image campaign format is selected.

• [Marker.io] Robertson Stephenson Tennant Account Issue — account-specific rendering and data issue reported via Marker.io has been resolved.