Changelog

Published public product pages for Regenemm Link and Regenemm Connect. Link is described as the patient-controlled sharing surface; Connect is described as the governed interoperability surface for FHIR, HL7, EMR/EHR, My Health Record-related pathways, and audit-aware exchange.

Added a public guide explaining how to read the API documentation, introduced endpoint maturity states, and added a spoke-to-Hub lifecycle model covering context capture, evidence binding, artifact generation, review gates, and approved release.

Expanded the public Security page with healthcare-specific trust posture sections and added a dedicated Clinical Safety page covering human review, artifact lifecycle, source provenance, patient-facing release, fail-closed behaviour, and Hub audit.

Aligned the Platform page to the public API taxonomy and added visible Resource groupings for Governance, Clinical AI, Security, Patient Communication, Interoperability, and Templates. Added Agent Role Governance as a public resource supporting the AI-assists-clinicians-decide position.

Expanded the developer-preview API documentation into Regenemm's public taxonomy: Runtime API, Templates API, Voice Audio API, Observability API, Billing API, and named Spoke APIs. The page keeps Hub governance, consent, release gates, audit, and PHI-safe telemetry visible across the integration surface.

Published a draft /api-docs surface for Regenemm Hub APIs. The page introduces required headers, authentication posture, RFC7807-style errors, artifact lifecycle states, and the Hub-and-spoke contract for developer-preview integrations.

Added and consolidated public governance materials covering agentic workflow governance, audit provenance, data processing, edge connector security, service providers, and clinical governance. Duplicate resource versions were removed in favour of canonical standalone pages where those pages already existed.

Updated public trust pages with clearer language around vulnerability reporting, Australia-first data residency posture, hosted clinical workflows, cross-border handling, and non-negotiable governance boundaries.

Created this draft public changelog model for review. The public version is intentionally curated: it records shipped product, governance, documentation, and platform improvements while keeping sensitive security, infrastructure, and clinical safety implementation detail in internal release evidence.

Added an architecture decision record describing the production launch posture, content strategy, security checks, dependency management, deployment readiness, and public trust surfaces for the Regenemm Healthcare website.

Accepted the platform contract that keeps governed spoke data, audit permanence, spoke identity, and fail-closed behaviour under Hub control. The public changelog summarises the architectural direction only; detailed persistence and audit evidence remains internal.

Approved a governed telemetry model for internal agent operations, covering structured event logs, low-cardinality metrics, distributed traces, dashboarding, and alerting. The policy separates operational observability from clinician-facing product content and prohibits patient-identifiable content in telemetry by default.

Documented the proposed integration-layer direction for sending approved clinical documents to hospital systems and reading structured patient context back through governed interfaces. This remains a draft architecture item until implementation, conformance, and clinical governance gates are complete.

Proposed a Hub-and-spoke terminology model for clinical code validation, terminology lookup, and spoke-local cache behaviour. The goal is consistent terminology use across spokes while preserving Hub authority over canonical terminology sources.

Updated the website deployment baseline with current Node and TypeScript support, tightened build configuration, remediated blocking static-analysis findings, and aligned dependency lockfiles for Netlify deployment.

Added and updated public-facing policy pages including privacy, acceptable use, accessibility, responsible disclosure, and data residency. These pages are intended to make Regenemm's clinical, privacy, and operational posture easier to evaluate.

Added public writing on patient-facing summaries, consent-first sharing, AI documentation in 2026, and moving from consultation to coordinated care. The series frames Regenemm around governed clinical outputs rather than generic note generation.

Accepted an internal bootstrap constitution for AI-assisted development roles. The decision formalises shared context, role boundaries, escalation expectations, uncertainty handling, and architecture constraints so agents operate from a consistent governance baseline.

Proposed the strategic doctrine that Regenemm should be understood as a voice-first clinical operating system built around encoded clinical domain logic, canonical clinical state, adjudication, and audience-specific outputs rather than generic note generation.

Accepted a transcription presentation policy to improve metadata consistency and reduce repeated unknown speaker labels in clinical transcript views. The change is presentation-layer focused and does not alter PHI storage, consent flow, or document-generation business logic.

Accepted the first RCDA retrieval pipeline architecture for patient readiness, future-window review, deterministic edge document mapping, event outbox behaviour, and auditable retrieval jobs. Public wording avoids partner-specific integration detail and implementation internals.

Accepted the end-to-end wiring pattern connecting patient-context enrichment and clinical document generation across frontend, backend, and Hub service tiers. The milestone reduced silent fallback risk and standardised request correlation across the workflow.

Accepted guardrails for consultation route contracts, CI reliability, pagination continuity, route-level monitoring, and stuck-run triage. The work improves release confidence and operator visibility for consultation workflows.

Accepted the end-to-end audio-to-document workflow architecture: voice capture, streaming transcription, interim summaries, multi-type clinical document generation, formatting, editor routing, and silent FHIR storage for interoperability.

Accepted the Regenemm risk playbook as a first-class governance artefact. The playbook classifies high-consequence clinical, identity, consent, FHIR, PHI, logging, and infrastructure work so review depth matches the real-world risk of the change.

Reaffirmed the migration away from legacy persistence paths toward Domain Node and Drizzle-backed data access. This strengthened typed schema ownership, reduced direct data access patterns, and clarified the Hub boundary for clinical records.

Accepted RCDA as a first-class clinical drafting vertical. The architecture separates transient drafting state from final approved clinical documents and uses Domain Node and Drizzle as the governed persistence path.

Accepted the two-phase Patient Context model: preload relevant context before transcription begins, then use cached context during live consultation insight generation. This reduced hot-path latency and separated continuity context from current clinical authority.

Piloted the three-rail voice architecture: live provisional coaching, parallel patient-context enrichment, and final structured synthesis. The design established that live assistance is transient while final clinical outputs require governed persistence and review.

Catalogued canonical Domain-Driven Design documentation from the broader Regenemm healthcare ecosystem. This created a reference baseline for bounded contexts, domain entities, events, schema validation, tenant isolation, and cross-application alignment.

Accepted FHIR integration hardening patterns using explicit data contracts, standards-aligned error responses, orchestration services, retry behaviour, and contract testing. This improved interoperability without leaking external data-shape complexity into core domain models.

Accepted safety patterns for explicit patient context, audited override flows, provenance display, masking for screen-sharing, and confirmation for high-risk actions. These patterns helped move clinical workflow design from prototype behaviour toward governed product behaviour.

Accepted common frontend data and state-management patterns for server state, view state, form validation, loading states, retries, and user-facing error taxonomy. This reduced brittle one-off data flows across the clinical workspace.

Accepted the Domain-Driven Design structure map for API, application, domain, infrastructure, observability, and storage boundaries. This became the architectural map for aligning clinical aggregates, domain services, events, and persistence handoffs.

Accepted patient-context governance rules covering provenance, quarantine-over-reject behaviour, async enrichment, redacted local context, controlled cloud persistence, conflict lineage, and correlation-aware observability.

Accepted a decoupled audio-to-LLM-to-patient-context workflow that separates card UI behaviour from recording, transcription, redaction, model calls, PHI reattachment, editor population, and persistence side effects.

Defined dual-mode audio processing so file-based recordings and realtime streaming could coexist. This preserved a reliable batch path while allowing live audio chunks to support more responsive clinical documentation workflows.

Accepted the enterprise clinical document management architecture, later verified with RCDA integration, clinician review before finalisation, governed document metadata, controlled storage, and durable audit expectations.

Accepted the clinical document approval workflow with explicit draft, review, approval, finalisation, and immutability concepts. This established early human-review and audit expectations for AI-generated clinical documents.

Accepted the first agent orchestration architecture using event-sourced memory, specialised clinical agents, FHIR anti-corruption boundaries, and domain-service coordination. Later governance work refined these ideas into bounded agent roles and Hub-controlled workflows.

Accepted a compliance-by-design framework for security, availability, processing integrity, confidentiality, privacy, and AI management. This gave early structure to how Regenemm would treat high-risk clinical AI work.

Proposed a terminology platform for concept validation, coding support, FHIR terminology operations, semantic search, and offline bundles. Later Hub-and-spoke terminology work refined this into a more governed model.

Proposed the earliest patient-context integration architecture: structured context ingestion, UUID-based patient mapping, FHIR-aligned patient summaries, and context-aware clinical documentation generation.