ArcaQ feature
§1 — How versioning works
Versioning is automatic and transparent — no manual commits needed.
Automatic snapshot on save
Every note save writes a versioned copy to <arcaq:versions/> with timestamp.
Timeline view
Chronological list of all snapshots with version number, date and content preview.
One-click restore
Restoring a version replaces current note content and creates a new version entry.
Content diff
Side-by-side diff highlights additions, deletions and rewrites between any two versions.
Linked to annotations
Annotations from the time of a save are reflected in that version's metadata.
Audit-ready
Every version record includes user ID, timestamp and note type — audit-ready out of the box.
§2 — API reference
One endpoint retrieves the full version history for any note.
| Method | Endpoint | Description |
|---|---|---|
| GET | /second-brain/notes/{note_id}/history | Retrieve all version snapshots for a note |
§3 - Runtime stabilization update
Recent hardening fixes removed transient gateway errors and restored stable module loading across the dashboard.
SSE proxy reliability
Notifications stream is now proxied in streaming mode to prevent chunked-read interruptions and intermittent 500 errors.
Ontology API resilience
Ontology stats/products now degrade gracefully on backend faults, eliminating prior 502 regressions in UI workflows.
Dashboard module routing fix
The expert_queue route alias is handled explicitly, preventing unknown-module rendering failures during HTMX loads.
Cluster stability verified
Post-fix checks on server01 confirmed running core pods with no CrashLoopBackOff on active platform services.
§4 — Release 1.1.12 (August 2026)
A stabilization and capability release: bulletproof SSO sessions, the OLM Enrichment Dashboard, full OOC internationalization, Kimi K3 as enterprise LLM (in-house or sovereign provider), automated backups and platform-wide observability.
SSO session stability
Keycloak realm (tour-operator) and token audience alignment across API and UI removed all session-expired errors. Tokens now last 8 hours with silent 24-hour refresh.
OLM Enrichment Dashboard
Live pipeline view (Qdrant, LLM, SHACL, Jena) with causal state machine SLA tracking, HGT consensus tiers (P1/P2/P3) and expandable PROV-O lineage per event.
OOC fully translated
Operational Ontology Contracts are now available in 21 languages with high-quality translations, consistent with the rest of the platform UI.
Kimi K3 enterprise LLM
Enrichment, classification and reasoning now run on the Kimi K3 enterprise LLM, operated in-house or by a sovereign provider, improving latency and answer quality.
Backups & OOM safeguards
Scheduled PostgreSQL backups protect platform state, and memory limits with eviction policies keep pods stable under image and data load.
Observability & pagination
Prometheus metrics across all microservices, Grafana dashboards, and paginated API responses for large result sets.
§5 — Release 2.0.0 (September 2026)
A major release: full OpenMetadata 2.0 native integration, the OLM Enrichment Dashboard, causal HGT with SHACL certification, the Decision Intelligence Loop, the Query Builder virtual layer and a unified design system.
OpenMetadata 2.0 native integration
Native OM 2.0 MCP connector (streamable HTTP /mcp) with full bidirectional catalog sync — business domains, expert profiles and glossary terms flow into PostgreSQL and OpenFGA, with lineage migrated across Postgres schemas and memory push via create_context_memory.
OLM Enrichment Dashboard
Multi-agent enrichment observability (Patent Claims #27–#59): live pipeline view with causal workflow states, HGT consensus tiers and PROV-O lineage per event.
Causal HGT with SHACL certification
Heterogeneous Graph Transformer reasoning with quarantine and SHACL certification gates — only validated knowledge reaches the production graph.
Query Builder & virtual layer
Visual query builder over the virtual semantic layer — no SPARQL required for business users.
Unified design system
A single design system across the entire platform UI for a consistent, accessible experience.
Decision Intelligence Loop
Governed decision log with risk-appetite evaluation, action dispatch to agents, outcome measurement, SME meaning challenges and PROV-O invalidation — the full Data-to-Learning loop.