Resources
Changelog
What changed in each version. Newest first. Dates are the app’s own release dates in Pacific time.
v0.5.0 October 1, 2026 · Wave 4 · Phase 1 complete
Added
- Clinical decision support: drug-interaction, drug-allergy and duplicate-therapy checks against a synthetic rule set, wired into orders and prescribing; audited overrides with a reason scaled to severity and a second-person review.
- E-prescribing: draft, sign and send by the named prescriber, a simulated pharmacy network, pharmacist verify, clarify, reject and dispense steps, cancel, and refill decisions that create the next draft. Controlled substances need a password step-up and a simulated monitoring check. It is not DEA EPCS.
- FHIR R4-style read-only interface and HL7 ADT messages out, both simulated; inbound messages are validated and not applied.
- Patient portal: a separate patient identity activated in person, with appointments, released results, medications, allergies, secure messages and refill requests; staff screens to invite and manage accounts.
- A regression pass and a re-score of the specification catalog (47 requirements upgraded with a reason, none downgraded).
Changed
- The permission matrix grew to 103 actions. The inbox refill placeholder was replaced by the real refill decision.
- A message-escaping bug in HL7 ADT messages was fixed. A race in a lab-result insert that failed the deploy gate was fixed in the product.
Known limits
- Everything external is simulated: pharmacy network, monitoring check, FHIR consumers and HL7 partners. The decision-support rules are synthetic and small.
- Portal identity proofing is in person only; no text-message codes, proxy access or payments.
- One live click-through step is red (an on-screen message for a too-short override reason); it passes locally and the server refuses a short reason. The cause is unresolved.
- Windows builds are unsigned and have no auto-update. The 0.5.0 desktop app was built but has not been installed on the developer’s PC.
- Developer tests on synthetic data: 137 unit tests (re-run October 2, 2026); end-to-end suites of 108, 201, 64, 114, 93 and 85 checks on the live hub; a permission-matrix run of 8,565 checks; security probes of 34, 105, 96 and 58 checks with no breach; a 427-screen scan with no problems. Internal tests, not an audit.
v0.4.0 October 1, 2026 · Wave 3 · superseded by 0.5.0 the same day
Added
- Scheduling, check-in and rooming with conflict checks, waitlist and no-show handling.
- Coverage and a simulated eligibility check; billing with charge capture, claim build and scrub, a simulated clearinghouse, remittances, denials and statements (synthetic fee schedule and code sets).
- On-site hub (Node 22 and SQLite from the same worker code) with signed, encrypted, hash-chained replication, and a read limiter.
- Downtime mode with a six-level ladder and sealed downtime reports, a read-only in-memory offline read cache, and back-entry of paper records.
- Backups with per-table digests and a signed manifest, and an isolated restore drill with dual authorisation to promote.
- Printing: Code 128 wristbands, labels and chart prints in a logged, rate-limited queue.
- IM and Memo Desk completion: drafts, due dates and escalation, retention and holds, size limits, patient-linked threads. Alert hardening: drills with timings, reports, a stale-alert watch and a setting for how an imminent-threat alert can be cancelled.
Changed
- Four live deploys failed the circuit gate on timing races between self-tests and the live-update timer; the self-tests were fixed and the gate was left as it was.
- Closes defects on offline operation, IM and memo limits and the alert cancel setting; the owner’s default for the cancel setting is still pending.
Known limits
- No real printer driver and no screen to resolve a replication conflict; promoting a restored backup records the decision only.
- Site secrets and the backup signing key are demo-grade.
- Developer tests on synthetic data: 131 unit tests at that release, a 93-check end-to-end suite, an on-site hub suite of 37 steps, a permission-matrix run of 7,091 checks, security probes of 96 checks, a 28-step click-through and a 378-screen scan. Internal tests.
v0.2.0 September 30, 2026 · Wave 1 · verified on synthetic data
Added
- Sign-in hardening: forced first-login change, one-time temporary passwords, TOTP MFA with recovery codes, lockout with alert, idle and manual lock, session list and revoke.
- Administration console: create, disable and enable users, reset password or MFA, unlock, role change with two-person approval for privileged roles, facilities, units, locations, MRN prefix and care teams.
- Registration: check-digit MRNs, live duplicate warning, unknown-patient registration, edit history, visits, two-person virtual merge and unmerge, masked national IDs with audited reveal, wristband preview, data-quality tab.
- Coded lists: problems (ICD-10-CM subset), allergies, medications (RxNorm subset), immunizations (CVX), history, type-ahead, and medication reconciliation that must decide every active medication.
- Audit v2: per-clinic hash chain with verify button, filters, read events, per-patient access report.
- VIP and restricted flags, break-glass with review queue, disclosure log.
- AuroraMed Packaging: a 13-module capability catalog (35 modules from v0.5.0), manifest editor, live validation with locked safety floors, signed packages, web bundle and portable app builds, two-operator publishing.
- Four synthetic tenants on one hub; Small Clinic and Hospital presets.
Changed
- Hub code reorganised into separate route modules; additive-only migrations.
- Static pages now carry security headers and a content security policy (security finding F1, fixed).
- Several interface bugs found by viewing screenshots were fixed (for example placeholder text showing “null”).
Known limits at this release
- Cloud-only: a network outage stops alerts and messaging. No offline chart (later closed in part by the on-site hub in v0.4.0).
- No real integrations. Results enter through a simulated internal interface.
- Open security findings F2, F3, F4, F8, F11 and F12 (see the whitepaper).
- Buttons are 36 px, below the 48 px design target. Builds are unsigned and have no auto-update.
v0.1.1 September 29 to 30, 2026 · Baseline pilot
Added
- Facility messaging with receipts, Memo Desk, code alerts with escalation and two-person cancel, live patient charts, code catalog with dual control, username and password sign-in.
- Two synthetic tenants and a small set of synthetic users.
v0.3.0 September 30, 2026 · Wave 2 · tested on synthetic data
Added
- Server-enforced permission matrix: 23 role classes by 53 actions at release (103 from v0.5.0), editable only through the signed manifest within floors and ceilings.
- Server circuit: every piece is registered and proven by a live self-test after each deploy; the deploy script refuses to finish and no installer is built when it fails.
- Clinical notes with templates, smart phrases, signing, co-signing, addenda and amendments.
- Structured vitals with range flags, BMI, BSA, MAP and eGFR calculators, NEWS2 and intake and output.
- Orders (CPOE core): catalog, order sets, verbal orders with read-back, a duplicate and allergy hook.
- Lab and imaging through simulated connectors (HL7 v2), with LOINC-coded results and a critical-result chain.
- Clinical inbox and tasks; forms, consents and e-signature with 36 curated mock templates.
Changed
- The first live deploy of this wave was refused by the circuit gate on the hosting plan used then; the hub stayed on 0.2.0 until the hosting plan was changed and the gate passed on every tenant. Several later deploys failed the gate for real bugs, which were fixed in the product without weakening the gate.
- Interface bugs found by a live screen scan (placeholder text showing “null”) were fixed.
Known limits at this release
- Lab and imaging are simulated. A drawn signature is not a certified e-signature.
- Still cloud-only; buttons 36 px.
- Developer tests on the live hub: end-to-end suites of 108, 201, 64 and 105 checks, a 37 and a 25-step click-through, 266 screens scanned with no problems, and security probes of 34 and 105 checks with no breach. All internal and synthetic.
Dates are Pacific Time. Test results for v0.5.0 are on the status page and are the developer’s own.