release: v1.3.0 — ModelForge Medical, a clinical decision-support workspace

This release moves the project from a general-purpose local AI chat/agent
client to a medical-focused clinical workspace. See the notice at the top
of README.md and its Product boundary section for what that does and
doesn't mean (never an autonomous diagnostician or prescriber, no
regulatory certification claimed).

Highlights:
- Clinical layer: Patient Cases, Evidence Library, per-case Knowledge
  Graph, and an Audit & Privacy log, plus a structured 8-section response
  contract and deterministic safety checks (emergency red-flag detection,
  allergy/medication conflict warnings, transmission preview, redaction)
  that run independently of the model.
- MCP client rework on the official SDK: protocol version negotiation,
  AJV schema validation, resources/prompts, OAuth 2.1+PKCE, per-tool trust
  profiles, and a DICOM MCP integration with an enforced tool denylist.
- Privacy/security: passphrase-based encryption at rest for patient cases
  and chat sessions, automatic session locking, configurable audit
  retention, and a SHA-256 hash-chained tamper-evident audit log.
- Approved model registry, clinical-note review sign-off, structured
  consent records, and a swappable MedicationSafetyProvider.
- Rust-backed data store I/O (lib/, napi-rs): atomic JSON read/write and
  audit-log hashing/appends moved off the O(n) Node path, fixing an O(n^2)
  audit-log growth pattern (~20s -> ~0.3s at 5,000 events) — every native
  call has a verified pure-TypeScript fallback. See
  docs/RUST_DATASTORE_TEST_REPORT.md and docs/RUST_MIGRATION_ASSESSMENT.md.
- SBOM generation and Dependabot coverage across every workspace; CI's
  Rust job now also builds and load-verifies the native addon on Windows
  and macOS, not just Linux.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Author: voidstackloop

.github/dependabot.yml

@@ -0,0 +1,68 @@
+version: 2
+updates:
+  # Each of these is an independent package.json/package-lock.json — kept as
+  # separate entries (rather than one root entry) so an update PR for, say,
+  # frontend never bundles unrelated app or e2e dependency bumps.
+  - package-ecosystem: npm
+    directory: /frontend
+    schedule:
+      interval: weekly
+    open-pull-requests-limit: 10
+    groups:
+      # Security patches land immediately rather than waiting in a batched
+      # group PR that might sit unreviewed.
+      production-dependencies:
+        applies-to: version-updates
+        dependency-type: production
+      development-dependencies:
+        applies-to: version-updates
+        dependency-type: development
+
+  - package-ecosystem: npm
+    directory: /app
+    schedule:
+      interval: weekly
+    open-pull-requests-limit: 10
+    groups:
+      production-dependencies:
+        applies-to: version-updates
+        dependency-type: production
+      development-dependencies:
+        applies-to: version-updates
+        dependency-type: development
+
+  - package-ecosystem: npm
+    directory: /e2e
+    schedule:
+      interval: weekly
+    open-pull-requests-limit: 5
+
+  - package-ecosystem: npm
+    directory: /mastervault-mcp-server
+    schedule:
+      interval: weekly
+    open-pull-requests-limit: 5
+
+  # lib/ is the napi-rs (Rust) download engine bound into the Electron app.
+  - package-ecosystem: cargo
+    directory: /lib
+    schedule:
+      interval: weekly
+    open-pull-requests-limit: 5
+
+  # ml/hardware-recommender is a standalone Python project (see ci.yml's
+  # python-recommender job) — its own dependency surface, tracked separately
+  # from the shipped app since only its exported ONNX artifact ships.
+  - package-ecosystem: pip
+    directory: /ml/hardware-recommender
+    schedule:
+      interval: weekly
+    open-pull-requests-limit: 5
+
+  # The CI/release workflows' own actions (actions/checkout, setup-node, …)
+  # — easy to forget since they're not in any package.json.
+  - package-ecosystem: github-actions
+    directory: /
+    schedule:
+      interval: weekly
+    open-pull-requests-limit: 5

.github/workflows/ci.yml

@@ -136,11 +136,23 @@ jobs:
           retention-days: 7
 
   rust:
-    # lib/ (modelforge-native) is the napi-rs download engine used by
-    # download-queue.ts. GitHub's ubuntu-latest runners ship rustup with a
-    # default stable toolchain already installed, so this only needs to add
-    # the two components fmt/clippy actually use.
-    runs-on: ubuntu-latest
+    # lib/ (modelforge-native) is the napi-rs addon backing GGUF downloads
+    # and, more recently, the JSON datastore/audit-hashing primitives (see
+    # docs/ARCHITECTURE.md). fmt/clippy/cargo test only prove the Rust side
+    # is correct on whichever OS runs them — they say nothing about whether
+    # the resulting `.node` binary actually loads under Node on that OS,
+    # which is a materially different failure mode (wrong ABI, missing
+    # platform target, packaging mismatch). Previously this job only ran on
+    # ubuntu-latest, so nothing in CI ever built or loaded the addon on
+    # Windows or macOS despite installers shipping for both — see
+    # docs/RUST_MIGRATION_ASSESSMENT.md's platform-matrix section. The
+    # matrix below and the "napi build + require() smoke test" step close
+    # that gap.
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, windows-latest, macos-latest]
+    runs-on: ${{ matrix.os }}
     defaults:
       run:
         working-directory: lib
@@ -161,10 +173,15 @@ jobs:
           restore-keys: |
             ${{ runner.os }}-cargo-
 
+      # fmt/clippy only need to run once — Rust source and formatting are
+      # identical across OSes, so repeating this on all three runners would
+      # just burn CI minutes for the same answer three times.
       - name: Format check
+        if: matrix.os == 'ubuntu-latest'
         run: cargo fmt --check
 
       - name: Clippy
+        if: matrix.os == 'ubuntu-latest'
         run: cargo clippy --all-targets -- -D warnings
 
       - name: Build
@@ -173,6 +190,89 @@ jobs:
       - name: Test
         run: cargo test
 
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: npm
+          cache-dependency-path: lib/package-lock.json
+
+      - name: Install napi CLI
+        run: npm ci
+
+      - name: napi build (produces the real platform-specific .node binary)
+        run: npm run build:debug
+
+      # The actual gap this job exists to close: proves Node on *this* OS
+      # can load the binary napi build just produced, not just that Cargo
+      # compiled it. A wrong target triple, an ABI mismatch, or a packaging
+      # error would fail cargo build's cross-compile silently-succeed case
+      # but fail right here.
+      - name: Load the built addon under Node and verify its exports
+        shell: bash
+        run: |
+          node -e "
+            const addon = require('../app/native');
+            const expected = ['downloadGgufFile', 'DownloadManager', 'readJsonFileNative', 'writeJsonFileAtomicNative', 'sha256HexNative', 'appendJsonArrayElementNative'];
+            const missing = expected.filter((name) => !(name in addon));
+            if (missing.length > 0) {
+              console.error('Native addon loaded but is missing expected exports:', missing);
+              process.exit(1);
+            }
+            console.log('Native addon loaded successfully on ${{ matrix.os }} with all expected exports.');
+          "
+
+  sbom:
+    # Generates a CycloneDX SBOM per npm workspace so every release has an
+    # auditable, machine-readable dependency manifest — required groundwork
+    # for the supply-chain/provenance controls flagged in
+    # docs/ENTERPRISE_READINESS_ASSESSMENT.md, and useful on its own for
+    # answering "are we affected by CVE-X" without re-deriving the dependency
+    # tree by hand. Runs independently of the test job so a flaky SBOM
+    # generation never blocks the build/test signal.
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: npm
+          cache-dependency-path: |
+            frontend/package-lock.json
+            app/package-lock.json
+            mastervault-mcp-server/package-lock.json
+
+      - name: Install frontend dependencies
+        working-directory: frontend
+        run: npm ci
+
+      - name: Install app dependencies
+        working-directory: app
+        run: npm ci
+
+      - name: Install mastervault-mcp-server dependencies
+        working-directory: mastervault-mcp-server
+        run: npm ci
+
+      - name: Generate SBOM (frontend)
+        working-directory: frontend
+        run: npx --yes @cyclonedx/cyclonedx-npm --output-file ../sbom-frontend.cdx.json
+
+      - name: Generate SBOM (app)
+        working-directory: app
+        run: npx --yes @cyclonedx/cyclonedx-npm --output-file ../sbom-app.cdx.json
+
+      - name: Generate SBOM (mastervault-mcp-server)
+        working-directory: mastervault-mcp-server
+        run: npx --yes @cyclonedx/cyclonedx-npm --output-file ../sbom-mastervault-mcp-server.cdx.json
+
+      - uses: actions/upload-artifact@v4
+        with:
+          name: sbom
+          path: sbom-*.cdx.json
+          retention-days: 90
+
   python-recommender:
     # ml/hardware-recommender/ is a standalone project (its own venv, own
     # train/download-dataset scripts) that isn't part of the app's build —