[Unreleased]
Changed
- Typography: Default UI font is Geist Sans (
--font-geist-sans) with system fallbacks (replaces Arial onbody); Geist Mono forcode/kbd/samp/pre; Tailwindfont-sansuses the same stack. Landing (body.landing-page) uses 16px base and relaxed line-height for marketing copy;.text-bodyon the home page matches 16px there. - Configurations UI (liquid glass):
/configurationspanels use translucent glass surfaces (backdrop-filter+ semantic tokens); tree row hover vs selected states are distinct; destructive deletes use confirm modals; execution preview empty state has a primary CTA; typography usestext-balance/text-prettywhere appropriate. - Frontend animation stack:
motionpackage withmotion/reactimports; simple marketing section entrances use CSS +useIntersectionRevealOnceinstead of Motion; heatmap / SA modal row hovers use CSS; directframer-motiondependency removed (still supplied viamotion). - Configurations workspace:
/configurationsuses a VS Code–style sidebar tree (virtualized) for field configs (parent/child lazy expand) and checker configs (flat); execution count badges per field config; editor uses Collapse per country; Executions tab previews recent runs; API:GET .../sdu-extraction-executionsandrecords/countaccept optionalfieldConfigId; Results page honors?fieldConfigId=for deep links. - Configurations & extraction UX (previous iteration): library-first layout; safer deletes (confirm), clone name suffix
(copy); extraction selects show Base/Variant with parent hint; “Merge with parent configuration” copy; results tables show Inheritance column; automation panel documentsuse_parent_config.
Added
- Parent/child field configs: Mongo
parent_id/child_idson field configs; create child configs under a parent; list roots viaroots_only=true;GET .../sdu-field-configs/:id/children. Extraction acceptsuse_parent_config; execution service resolvesparent_config_idfor SDUvalidate_and_extractwhen enabled. - Admin Permanence KPI dashboard at
/admin/kpis— platform-wide analytics for extraction and check content executions - All-time totals and monthly usage trends (last 12 months default)
- Distribution views: doc types, configs (field/checker), users, statuses
- Cross-breakdowns: user by field config, user by doc type
- Admin KPI API endpoints:
permanence-kpis/summary,permanence-kpis/monthly,permanence-kpis/distributions - Health at a glance cards: failure rate, queue pressure, tokens per completed, active users (Phase 1)
- Health metrics endpoint
permanence-kpis/health-metrics: latency p50/p95, stale counts, failure reason categories - KPI dictionary and threshold policy in
setup/docs/guides/KPI_DICTIONARY.md
[0.3.0] - March 5, 2026
Added
- Changelog tracking and in-app changelog page at
/changelog - "What's New" section on home page surfacing recent releases
- Version display in header and changelog navigation link
Changed
- Home page refactored: streamlined section order, removed redundant differentiators block
- Announcements updated to reflect 2026 release timeline
[0.2.0] - February 14, 2026
Added
- Deep Agent Architecture for complex extraction tasks — handles 500-row PO tables with grouped calculations
- Smart-Cropping for image-based documents — reduces token usage by 20–30%
- Auto-Rotation: documents automatically corrected to proper orientation before extraction
- Config Agent streaming: real-time feedback during extraction configuration
Changed
- Zero-downtime Kubernetes rolling updates with readiness probes and pre-stop lifecycle hooks
- Streaming validation improvements for more reliable long-running extractions
Fixed
- 502 Bad Gateway transient errors during AKS node scale-down events
- Import cache issue in Starlight service causing timeout on cold start
[0.1.0] - January 13, 2026
Added
- Initial release of DocuAgent platform
- Document extraction pipeline with Azure OpenAI Vision Language Model
- Multi-document order processing (invoices, bills of lading, purchase orders)
- Config Agent: conversational AI-guided extraction setup
- Cross-document validation with 13 business rules
- Azure Service Bus integration for async document processing
- Multi-language document support
- Role-based access control via Azure AD B2C
- Real-time streaming extraction results
- Statistics and token usage dashboards
- Company search and entity enrichment service
- Alchemist agent service for specialized document tasks