draft: Make Tool Steps Seperate Pages

## What & why

Production reports of policy enforcement "hanging" traced to
large/many-page documents: the watermark step's flatten-to-image
(`convertPDFToImage`) on a 500+ page PDF takes minutes, exceeding both
the client poll cap and the backend per-step timeout. This makes the
slow case graceful instead of looking broken, and makes load-shedding
non-fatal.

### Poll runs to completion (no false "hang")
The client poll loop used a flat ~150s cap that was **shorter than the
backend's 300s per-step timeout**, so it abandoned long-but-healthy runs
mid-flight. The budget is now sized to the backend's real worst case —
`stepCount × per-step timeout + grace`, learned from the first status
report — so the client always polls long enough to surface the run's
**actual** terminal state (success or the backend's real error) rather
than a misleading client-side timeout.

### Per-step progress
The activity feed now shows `Enforcing… · step n/m` (from
`currentStep`/`stepCount`), so a slow step shows movement instead of a
dead spinner.

### Soft-retry on queue rejection
Under load the shared `JobQueue` rejects runs ("queue full"), which
previously surfaced as a hard failure needing a manual Retry. The
backend now tags that rejection with a stable `POLICY_QUEUE_FULL`
errorCode; the client treats it as transient backpressure and
**auto-retries the file in place** with exponential backoff (≈4s→64s, ~2
min), shown as a soft "Busy — retrying…" row, falling back to the manual
Retry only once the retry budget is spent.

## Testing
- **Frontend unit tests** (30 pass across the policies suite), including
a new `usePolicyAutoRun.retry.test.tsx` that drives the real controller
orchestration (poll → `POLICY_QUEUE_FULL` → relabel → backoff → in-place
re-dispatch), plus poll-budget, step-progress, and activity-feed relabel
cases.
- **Backend** `PolicyEngineTest` case asserting a queue-rejected run
carries the `POLICY_QUEUE_FULL` code.
- Typecheck clean on all three flavors (proprietary/saas/core); prettier
+ spotless clean.
- Poll-budget + progress + real-error surfacing were also verified live
end-to-end against a 599-page run (survived past the old cap, showed
step progress, reported the backend's real 300s-timeout failure,
recovered after a simulated network drop).

## Not included (follow-ups)
- The underlying flatten-to-image cost itself (bounded-memory/streaming
flatten, revisiting `convertPDFToImage` default and the 300s timeout) —
the real perf fix, deliberately out of scope here.

.dockerignore

@@ -26,8 +26,9 @@ version_builds/
 node_modules/
 **/node_modules/
 frontend/node_modules/
-frontend/dist/
-frontend/playwright-report/
+frontend/editor/dist/
+frontend/dist-portal/
+frontend/editor/playwright-report/
 .npm/
 .yarn/
 

.github/actions/setup-bot/action.yml

@@ -24,7 +24,7 @@ runs:
       id: generate-token
       uses: actions/create-github-app-token@bcd2ba49218906704ab6c1aa796996da409d3eb1 # v3.2.0
       with:
-        app-id: ${{ inputs.app-id }}
+        client-id: ${{ inputs.app-id }}
         private-key: ${{ inputs.private-key }}
     - name: Configure Git
       run: |

.github/aur/stirling-pdf-desktop/PKGBUILD

@@ -1,6 +1,6 @@
 # Maintainer: Stirling PDF Inc <contact@stirlingpdf.com>
 pkgname=stirling-pdf-desktop
-pkgver=2.11.0
+pkgver=2.12.0
 pkgrel=1
 pkgdesc="Locally hosted, web-based PDF manipulation tool (Tauri desktop app, official Stirling PDF Inc build)"
 arch=('x86_64')

.github/aur/stirling-pdf-server-bin/PKGBUILD

@@ -1,6 +1,6 @@
 # Maintainer: Stirling PDF Inc <contact@stirlingpdf.com>
 pkgname=stirling-pdf-server-bin
-pkgver=2.11.0
+pkgver=2.12.0
 pkgrel=1
 pkgdesc="Locally hosted, web-based PDF manipulation tool (server JAR, prebuilt)"
 arch=('any')

.github/config/.files.yaml

@@ -1,6 +1,8 @@
 build: &build
   - build.gradle
   - app/(common|core|proprietary)/build.gradle
+  - Taskfile.yml
+  - .taskfiles/backend.yml
 
 openapi: &openapi
   - *build
@@ -38,6 +40,11 @@ project: &project
   - frontend/**
   - docker/**
   - scripts/RestartHelper.java
+  - Taskfile.yml
+  - .taskfiles/backend.yml
+  - .taskfiles/docker.yml
+  - scripts/db-migration/**
+  - .github/workflows/db-migration-test.yml
 
 frontend: &frontend
   - frontend/**
@@ -53,6 +60,9 @@ frontend: &frontend
   - scripts/summarize_type3_signatures.py
   - scripts/type3_to_cff.py
   - scripts/update_type3_library.py
+  - Taskfile.yml
+  - .taskfiles/frontend.yml
+  - .taskfiles/e2e.yml
 
 # Files that affect the Tauri desktop bundle. Gate the multi-OS Tauri build
 # job on changes to any of these.
@@ -64,6 +74,8 @@ tauri: &tauri
   - frontend/package-lock.json
   - frontend/editor/vite.config.ts
   - .github/workflows/tauri-build.yml
+  - Taskfile.yml
+  - .taskfiles/desktop.yml
 
 # Files that affect the AI engine (Python tool models, fixers, tests). Gate
 # the engine validation job on changes to engine sources or to the Java
@@ -72,6 +84,8 @@ engine: &engine
   - engine/**
   - app/(common|core|proprietary)/src/main/java/**
   - .github/workflows/ai-engine.yml
+  - Taskfile.yml
+  - .taskfiles/engine.yml
 
 licenses-frontend: &licenses-frontend
   - ".github/workflows/frontend-backend-licenses-update.yml"
@@ -100,4 +114,4 @@ proprietary: &proprietary
   - configs/settings.yml.template
   - build.gradle
   - app/proprietary/build.gradle
-  - .github/workflows/build-enterprise.yml
+  - .github/workflows/build-enterprise.yml

.github/scripts/check_language_toml.py

@@ -13,7 +13,7 @@ Usage:
 """
 
 # Sample for Windows:
-# python .github/scripts/check_language_toml.py --reference-file frontend/public/locales/en-GB/translation.toml --branch "" --files frontend/public/locales/de-DE/translation.toml frontend/public/locales/fr-FR/translation.toml
+# python .github/scripts/check_language_toml.py --reference-file frontend/editor/public/locales/en-US/translation.toml --branch "" --files frontend/editor/public/locales/de-DE/translation.toml frontend/editor/public/locales/fr-FR/translation.toml
 
 import argparse
 import glob
@@ -184,7 +184,7 @@ def check_for_differences(reference_file, file_list, branch, actor):
     if len(file_list) == 1:
         file_arr = file_list[0].split()
 
-    base_dir = Path.cwd() / "frontend" / "public" / "locales"
+    base_dir = Path.cwd() / "frontend" / "editor" / "public" / "locales"
 
     for file_path in file_arr:
         file_path = Path(file_path)
@@ -211,7 +211,7 @@ def check_for_differences(reference_file, file_list, branch, actor):
             )
             continue
 
-        if basename_current_file == basename_reference_file and locale_dir == "en-GB":
+        if basename_current_file == basename_reference_file and locale_dir == "en-US":
             continue
 
         if (
@@ -308,7 +308,7 @@ def check_for_differences(reference_file, file_list, branch, actor):
         report.append("## ❌ Overall Check Status: **_Failed_**")
         report.append("")
         report.append(
-            f"@{actor} please check your translation if it conforms to the standard. Follow the format of [en-GB/translation.toml](https://github.com/Stirling-Tools/Stirling-PDF/blob/main/frontend/public/locales/en-GB/translation.toml)"
+            f"@{actor} please check your translation if it conforms to the standard. Follow the format of [en-US/translation.toml](https://github.com/Stirling-Tools/Stirling-PDF/blob/main/frontend/editor/public/locales/en-US/translation.toml)"
         )
     else:
         report.append("## ✅ Overall Check Status: **_Success_**")
@@ -372,6 +372,7 @@ if __name__ == "__main__":
                 os.path.join(
                     os.getcwd(),
                     "frontend",
+                    "editor",
                     "public",
                     "locales",
                     "*",

.github/scripts/verify-updater-signatures.py

@@ -0,0 +1,83 @@
+#!/usr/bin/env python3
+"""Verify Tauri updater .sig files against plugins.updater.pubkey in tauri.conf.json.
+
+Usage: verify-updater-signatures.py <dir-to-scan> [tauri.conf.json]
+"""
+
+import binascii
+import sys
+import json
+import base64
+import hashlib
+from pathlib import Path
+from cryptography.hazmat.primitives.asymmetric.ed25519 import Ed25519PublicKey
+from cryptography.exceptions import InvalidSignature
+
+ART_ROOT = Path(sys.argv[1])
+CONF = Path(
+    sys.argv[2] if len(sys.argv) > 2 else "frontend/editor/src-tauri/tauri.conf.json"
+)
+
+
+def load_pubkey():
+    # tauri pubkey = base64 of a minisign .pub file; last line is base64 of
+    # [2 algo][8 key-id][32 ed25519 public key].
+    raw = json.loads(CONF.read_text())["plugins"]["updater"]["pubkey"]
+    blob = base64.b64decode(base64.b64decode(raw).decode().splitlines()[-1])
+    return blob[2:10], Ed25519PublicKey.from_public_bytes(blob[10:])
+
+
+def hash_file(path: Path) -> bytes:
+    h = hashlib.blake2b(digest_size=64)
+    with path.open("rb") as f:
+        for chunk in iter(lambda: f.read(1 << 16), b""):
+            h.update(chunk)
+    return h.digest()
+
+
+def verify(artifact: Path, sig_file: Path, keyid_pub, pub) -> str:
+    # tauri .sig = base64 of a minisign signature file (4 lines).
+    try:
+        lines = base64.b64decode(sig_file.read_text()).decode().splitlines()
+        sig_blob = base64.b64decode(lines[1])
+    except (binascii.Error, IndexError, UnicodeDecodeError) as e:
+        return f"FAIL malformed sig ({type(e).__name__})"
+    algo, keyid, sig = sig_blob[:2], sig_blob[2:10], sig_blob[10:74]
+    if keyid != keyid_pub:
+        return f"FAIL key-id mismatch (sig {keyid.hex()} vs pub {keyid_pub.hex()})"
+    # 'ED' = prehashed (BLAKE2b-512), 'Ed' = legacy (raw message).
+    msg = hash_file(artifact) if algo == b"ED" else artifact.read_bytes()
+    try:
+        pub.verify(sig, msg)
+    except InvalidSignature:
+        return f"FAIL signature invalid (algo={algo.decode()})"
+    # Global signature covers sig + trusted_comment.
+    gc = "global-sig FAIL"
+    try:
+        tc = lines[2].split("trusted comment: ", 1)[1]
+        pub.verify(base64.b64decode(lines[3]), sig + tc.encode())
+        gc = "global-sig OK"
+    except (InvalidSignature, IndexError, binascii.Error):
+        pass
+    return f"VALID (algo={algo.decode()}, keyid={keyid.hex()}, {gc})"
+
+
+keyid_pub, pub = load_pubkey()
+print(f"updater pubkey keyid={keyid_pub.hex()}\n")
+sigs = sorted(ART_ROOT.rglob("*.sig"))
+if not sigs:
+    print(f"WARN: no .sig files under {ART_ROOT} - nothing to verify")
+    sys.exit(0)
+bad = 0
+for sig_file in sigs:
+    artifact = sig_file.with_suffix("")
+    if not artifact.exists():
+        print(f"  ? {sig_file.name}: artifact missing")
+        bad += 1
+        continue
+    res = verify(artifact, sig_file, keyid_pub, pub)
+    print(f"  {artifact.name}: {res}")
+    if not res.startswith("VALID") or "global-sig FAIL" in res:
+        bad += 1
+print(f"\n{'ALL SIGNATURES VALID' if bad == 0 else f'{bad} SIGNATURE(S) FAILED'}")
+sys.exit(1 if bad else 0)

.github/workflows/PR-Auto-Deploy-V2.yml

@@ -239,7 +239,7 @@ jobs:
 
       - name: Build and push V2 image (Depot)
         if: env.USE_DEPOT == 'true' && steps.check-image.outputs.exists == 'false'
-        uses: depot/build-push-action@5f3b3c2e5a00f0093de47f657aeaefcedff27d18 # v1.16.0
+        uses: depot/build-push-action@98e78adca7817480b8185f474a400b451d74e287 # v1.16.0
         with:
           project: ${{ vars.DEPOT_PROJECT_ID }}
           context: .
@@ -287,12 +287,13 @@ jobs:
                 - /stirling/V2-PR-${{ needs.check-pr.outputs.pr_number }}/data:/usr/share/tessdata:rw
                 - /stirling/V2-PR-${{ needs.check-pr.outputs.pr_number }}/config:/configs:rw
                 - /stirling/V2-PR-${{ needs.check-pr.outputs.pr_number }}/logs:/logs:rw
+                - /stirling/V2-PR-${{ needs.check-pr.outputs.pr_number }}/storage:/storage:rw
               environment:
                 DISABLE_ADDITIONAL_FEATURES: "false"
                 SECURITY_ENABLELOGIN: "true"
                 SECURITY_INITIALLOGIN_USERNAME: "${{ secrets.TEST_LOGIN_USERNAME }}"
                 SECURITY_INITIALLOGIN_PASSWORD: "${{ secrets.TEST_LOGIN_PASSWORD }}"
-                SYSTEM_DEFAULTLOCALE: en-GB
+                SYSTEM_DEFAULTLOCALE: en-US
                 UI_APPNAME: "Stirling-PDF V2 PR#${{ needs.check-pr.outputs.pr_number }}"
                 UI_HOMEDESCRIPTION: "V2 PR#${{ needs.check-pr.outputs.pr_number }} - Embedded Architecture"
                 UI_APPNAMENAVBAR: "V2 PR#${{ needs.check-pr.outputs.pr_number }}"
@@ -309,7 +310,7 @@ jobs:
 
           ssh -i ../private.key -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -T ${{ secrets.NEW_VPS_USERNAME }}@${{ secrets.NEW_VPS_HOST }} << ENDSSH
             # Create V2 PR-specific directories
-            mkdir -p /stirling/V2-PR-${{ needs.check-pr.outputs.pr_number }}/{data,config,logs}
+            mkdir -p /stirling/V2-PR-${{ needs.check-pr.outputs.pr_number }}/{data,config,logs,storage}
 
             # Move docker-compose file to correct location
             mv /tmp/docker-compose-v2.yml /stirling/V2-PR-${{ needs.check-pr.outputs.pr_number }}/docker-compose.yml

.github/workflows/PR-Demo-Comment-with-react.yml

@@ -222,10 +222,10 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Run Gradle Command
         run: |
           if [ "${{ needs.check-comment.outputs.disable_security }}" == "true" ]; then
@@ -256,7 +256,7 @@ jobs:
 
       - name: Build and push PR-specific image (Depot)
         if: env.USE_DEPOT == 'true'
-        uses: depot/build-push-action@5f3b3c2e5a00f0093de47f657aeaefcedff27d18 # v1.16.0
+        uses: depot/build-push-action@98e78adca7817480b8185f474a400b451d74e287 # v1.16.0
         with:
           project: ${{ vars.DEPOT_PROJECT_ID }}
           context: .
@@ -285,7 +285,7 @@ jobs:
 
       - name: Build and push engine image (Depot)
         if: env.USE_DEPOT == 'true' && needs.check-comment.outputs.enable_prototypes == 'true'
-        uses: depot/build-push-action@5f3b3c2e5a00f0093de47f657aeaefcedff27d18 # v1.16.0
+        uses: depot/build-push-action@98e78adca7817480b8185f474a400b451d74e287 # v1.16.0
         with:
           project: ${{ vars.DEPOT_PROJECT_ID }}
           context: ./engine
@@ -388,7 +388,7 @@ jobs:
               environment:
                 DISABLE_ADDITIONAL_FEATURES: "${DISABLE_ADDITIONAL_FEATURES}"
                 SECURITY_ENABLELOGIN: "${LOGIN_SECURITY}"
-                SYSTEM_DEFAULTLOCALE: en-GB
+                SYSTEM_DEFAULTLOCALE: en-US
                 UI_APPNAME: "Stirling-PDF PR#${PR_NUMBER}"
                 UI_HOMEDESCRIPTION: "PR#${PR_NUMBER} for Stirling-PDF Latest"
                 UI_APPNAMENAVBAR: "PR#${PR_NUMBER}"

.github/workflows/ai-engine.yml

@@ -1,8 +1,9 @@
 name: AI Engine CI
 
-# Validates the Python AI engine: regenerates tool models, runs fixers,
-# lint, type-check, and tests. Called from build.yml on PRs and merge_group;
-# also runs directly on push to main as a post-merge safety net.
+# Validates the Python AI engine: regenerates tool models and runs the
+# engine quality gate (lint, type-check, format-check, tests). Called from
+# build.yml on PRs and merge_group; also runs directly on push to main as
+# a post-merge safety net.
 on:
   workflow_call:
   push:
@@ -42,36 +43,104 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
 
       - name: Regenerate tool models
         run: task engine:tool-models
 
       - name: Verify tool models are up to date
+        id: tool-models-check
+        continue-on-error: true
+        run: git diff --exit-code engine/src/stirling/models/tool_models.py
+
+      - name: Comment on tool models check failure
+        # Only post a comment on PRs. github-script's PR helpers need an
+        # issue/PR number, which doesn't exist on merge_group runs.
+        if: steps.tool-models-check.outcome == 'failure' && github.event_name == 'pull_request'
+        continue-on-error: true
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        with:
+          script: |
+            const marker = '<!-- tool-models-check -->';
+            const body = [
+              marker,
+              '### Tool Models Check Failed',
+              '',
+              'The generated `engine/src/stirling/models/tool_models.py` is out of date with the Java OpenAPI spec and will need to be regenerated before it can be merged in.',
+              '',
+              'Run `task engine:tool-models` to regenerate, then commit the updated file.',
+            ].join('\n');
+            const { data: comments } = await github.rest.issues.listComments({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.issue.number,
+            });
+            const existing = comments.find(c => c.body.includes(marker));
+            if (existing) {
+              await github.rest.issues.updateComment({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                comment_id: existing.id,
+                body,
+              });
+            } else {
+              await github.rest.issues.createComment({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                issue_number: context.issue.number,
+                body,
+              });
+            }
+
+      - name: Fail if tool models check failed
+        if: steps.tool-models-check.outcome == 'failure'
         run: |
-          if ! git diff --exit-code engine/src/stirling/models/tool_models.py; then
-            echo "tool_models.py is out of date."
-            echo "Run 'task engine:tool-models' locally and commit the updated file."
-            exit 1
-          fi
+          echo "============================================"
+          echo "  Tool Models Check Failed"
+          echo "============================================"
+          echo ""
+          echo "The generated engine/src/stirling/models/tool_models.py"
+          echo "is out of date with the Java OpenAPI spec and will"
+          echo "need to be regenerated before it can be merged in."
+          echo ""
+          echo "Run 'task engine:tool-models' to regenerate, then"
+          echo "commit the updated file."
+          echo "============================================"
+          exit 1
 
-      - name: Run fixers
-        run: task engine:fix
+      - name: Remove tool models check comment on success
+        if: steps.tool-models-check.outcome == 'success' && github.event_name == 'pull_request'
+        continue-on-error: true
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        with:
+          script: |
+            const marker = '<!-- tool-models-check -->';
+            const { data: comments } = await github.rest.issues.listComments({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.issue.number,
+            });
+            const existing = comments.find(c => c.body.includes(marker));
+            if (existing) {
+              await github.rest.issues.deleteComment({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                comment_id: existing.id,
+              });
+            }
 
-      - name: Verify fixes are committed
-        id: fixer_changes
-        run: |
-          if ! git diff --quiet; then
-            git --no-pager diff --stat
-            echo "::error::There are issues with your Python code that will need to be fixed before they can be merged in. Run 'task engine:fix' to auto-fix what can be fixed automatically, then run 'task engine:check' to see what still needs fixing manually."
-            exit 1
-          fi
+      - name: Quality-check engine
+        id: engine-check
+        run: task engine:check
+        continue-on-error: true
 
-      - name: Comment on fixer failures
-        if: steps.fixer_changes.outcome == 'failure' && github.event_name == 'pull_request'
+      - name: Comment on engine check failure
+        # Only post a comment on PRs. github-script's PR helpers need an
+        # issue/PR number, which doesn't exist on merge_group runs.
+        if: steps.engine-check.outcome == 'failure' && github.event_name == 'pull_request'
         continue-on-error: true
         uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
         with:
@@ -107,11 +176,39 @@ jobs:
               });
             }
 
-      - name: Run linting
-        run: task engine:lint
+      - name: Fail if engine check failed
+        if: steps.engine-check.outcome == 'failure'
+        run: |
+          echo "============================================"
+          echo "  Engine Check Failed"
+          echo "============================================"
+          echo ""
+          echo "There are issues with your Python code that"
+          echo "will need to be fixed before they can be merged in."
+          echo ""
+          echo "Run 'task engine:fix' to auto-fix what can be"
+          echo "fixed automatically, then run 'task engine:check'"
+          echo "to see what still needs fixing manually."
+          echo "============================================"
+          exit 1
 
-      - name: Run type checking
-        run: task engine:typecheck
-
-      - name: Run tests
-        run: task engine:test
+      - name: Remove engine check comment on success
+        if: steps.engine-check.outcome == 'success' && github.event_name == 'pull_request'
+        continue-on-error: true
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        with:
+          script: |
+            const marker = '<!-- engine-check -->';
+            const { data: comments } = await github.rest.issues.listComments({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.issue.number,
+            });
+            const existing = comments.find(c => c.body.includes(marker));
+            if (existing) {
+              await github.rest.issues.deleteComment({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                comment_id: existing.id,
+              });
+            }

.github/workflows/backend-build.yml

@@ -1,8 +1,14 @@
 name: Backend build, format check, and coverage
 
 # Reusable workflow called from build.yml. Runs the backend build matrix
-# (JDK 25 × spring-security on/off), Spotless formatting check, JUnit, and
+# (JDK 25 × every flavor), Spotless formatting check, JUnit, and
 # posts Jacoco coverage to PRs.
+#
+# Flavor axis (maps to STIRLING_FLAVOR in settings.gradle):
+#   core         - DISABLE_ADDITIONAL_FEATURES=true, no proprietary, no saas
+#   proprietary  - default build, no saas
+#   saas         - proprietary + the saas subproject (build + JUnit only,
+#                  never any runtime/integration testing)
 on:
   workflow_call:
 
@@ -25,7 +31,7 @@ jobs:
       fail-fast: false
       matrix:
         jdk-version: [25]
-        spring-security: [true, false]
+        flavor: [core, proprietary, saas]
     steps:
       - name: Harden Runner
         uses: step-security/harden-runner@ab7a9404c0f3da075243ca237b5fac12c98deaa5 # v2.19.3
@@ -52,13 +58,16 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
           cache-disabled: true
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Check Java formatting (Spotless)
-        if: matrix.jdk-version == 25 && matrix.spring-security == false
+        # Runs once per matrix combination - pick the cheapest leg
+        # (core - no proprietary, no saas) so we don't wait for the
+        # heavier flavors just to fail formatting.
+        if: matrix.jdk-version == 25 && matrix.flavor == 'core'
         id: spotless-check
         run: task backend:format:check
         continue-on-error: true
@@ -67,7 +76,7 @@ jobs:
           MAVEN_PASSWORD: ${{ secrets.MAVEN_PASSWORD }}
           MAVEN_PUBLIC_URL: ${{ secrets.MAVEN_PUBLIC_URL }}
 
-      - name: Comment on Java formatting failure
+      - name: Comment on backend format check failure
         # Only post a comment on PRs. github-script's PR helpers need an
         # issue/PR number, which doesn't exist on merge_group runs.
         if: steps.spotless-check.outcome == 'failure' && github.event_name == 'pull_request'
@@ -78,15 +87,11 @@ jobs:
             const marker = '<!-- java-formatting-check -->';
             const body = [
               marker,
-              '### Java Formatting Check Failed',
+              '### Backend Format Check Failed',
               '',
-              'Your code has formatting issues. Run the following command to fix them:',
+              'There are formatting issues in your Java code that will need to be fixed before they can be merged in.',
               '',
-              '```bash',
-              'task backend:format',
-              '```',
-              '',
-              'Then commit and push the changes.',
+              'Run `task backend:format` to auto-fix, then commit and push the changes.',
             ].join('\n');
             const { data: comments } = await github.rest.issues.listComments({
               owner: context.repo.owner,
@@ -110,33 +115,61 @@ jobs:
               });
             }
 
-      - name: Fail if Java formatting issues found
+      - name: Fail if backend format check failed
         if: steps.spotless-check.outcome == 'failure'
         run: |
           echo "============================================"
-          echo "  Java Formatting Check Failed"
+          echo "  Backend Format Check Failed"
           echo "============================================"
           echo ""
-          echo "Your code has formatting issues."
-          echo "Run the following command to fix them:"
+          echo "There are formatting issues in your Java code"
+          echo "that will need to be fixed before they can be"
+          echo "merged in."
           echo ""
-          echo "  task backend:format"
-          echo ""
-          echo "Then commit and push the changes."
+          echo "Run 'task backend:format' to auto-fix, then"
+          echo "commit and push the changes."
           echo "============================================"
           exit 1
 
-      - name: Build with Gradle and spring security ${{ matrix.spring-security }}
+      - name: Remove backend format check comment on success
+        if: steps.spotless-check.outcome == 'success' && github.event_name == 'pull_request'
+        continue-on-error: true
+        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
+        with:
+          script: |
+            const marker = '<!-- java-formatting-check -->';
+            const { data: comments } = await github.rest.issues.listComments({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.issue.number,
+            });
+            const existing = comments.find(c => c.body.includes(marker));
+            if (existing) {
+              await github.rest.issues.deleteComment({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                comment_id: existing.id,
+              });
+            }
+
+      - name: Build with Gradle (flavor=${{ matrix.flavor }})
+        # STIRLING_FLAVOR is read by settings.gradle and expands into the
+        # right combination of DISABLE_ADDITIONAL_FEATURES + ENABLE_SAAS
+        # so we don't have to set them by hand. The saas flavor pulls in
+        # the app/saas subproject (unit tests only - no runtime tests).
         run: task backend:build:ci
         env:
           MAVEN_USER: ${{ secrets.MAVEN_USER }}
           MAVEN_PASSWORD: ${{ secrets.MAVEN_PASSWORD }}
           MAVEN_PUBLIC_URL: ${{ secrets.MAVEN_PUBLIC_URL }}
-          DISABLE_ADDITIONAL_FEATURES: ${{ matrix.spring-security }}
+          STIRLING_FLAVOR: ${{ matrix.flavor }}
 
       - name: Check Test Reports Exist
         if: always()
         run: |
+          # Common + core + proprietary always build (proprietary is
+          # excluded only at runtime, not from the gradle subproject
+          # graph). Saas builds add a fourth report dir.
           declare -a dirs=(
             "app/core/build/reports/tests/"
             "app/core/build/test-results/"
@@ -145,6 +178,9 @@ jobs:
             "app/proprietary/build/reports/tests/"
             "app/proprietary/build/test-results/"
           )
+          if [ "${{ matrix.flavor }}" = "saas" ]; then
+            dirs+=("app/saas/build/reports/tests/" "app/saas/build/test-results/")
+          fi
           for dir in "${dirs[@]}"; do
             if [ ! -d "$dir" ]; then
               echo "Missing $dir"
@@ -156,7 +192,7 @@ jobs:
         if: always()
         uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
         with:
-          name: test-reports-jdk-${{ matrix.jdk-version }}-spring-security-${{ matrix.spring-security }}
+          name: test-reports-jdk-${{ matrix.jdk-version }}-flavor-${{ matrix.flavor }}
           path: |
             app/**/build/reports/jacoco/test
             app/**/build/reports/tests/
@@ -166,7 +202,47 @@ jobs:
           retention-days: 3
           if-no-files-found: warn
 
-      - name: Add coverage to PR with spring security ${{ matrix.spring-security }} and JDK ${{ matrix.jdk-version }}
+      - name: Install defusedxml for coverage summary
+        # coverage-summary.py parses JaCoCo XML through defusedxml to
+        # silence security scanners that pattern-match on the stdlib
+        # xml.etree.ElementTree.parse call.
+        if: always() && matrix.flavor == 'saas'
+        run: python -m pip install --quiet defusedxml
+
+      - name: JaCoCo coverage step summary
+        # Only the saas leg posts the JUnit summary - it's a strict
+        # superset of the core + proprietary legs (same .exec files plus
+        # the saas subproject). Posting from all three would mean three
+        # near-identical tables crowding out the aggregate report.
+        if: always() && matrix.flavor == 'saas'
+        run: |
+          python scripts/coverage-summary.py \
+            --title "Backend JUnit coverage (JDK ${{ matrix.jdk-version }})" \
+            --jacoco "common=app/common/build/reports/jacoco/test/jacocoTestReport.xml" \
+            --jacoco "core=app/core/build/reports/jacoco/test/jacocoTestReport.xml" \
+            --jacoco "proprietary=app/proprietary/build/reports/jacoco/test/jacocoTestReport.xml" \
+            --jacoco "saas=app/saas/build/reports/jacoco/test/jacocoTestReport.xml" \
+            --github-step-summary
+
+      - name: Upload raw JUnit .exec for aggregate merge
+        # Same dedup rationale as the summary step: upload from the saas
+        # leg only (the most complete set, includes app/saas/.../test.exec)
+        # so the aggregate workflow merges the union rather than three
+        # overlapping subsets.
+        #
+        # Separate artifact from the HTML reports so the aggregate
+        # workflow can grab just the .exec files with a name pattern
+        # (`jacoco-exec-*`) instead of unpacking the whole test-reports
+        # tarball.
+        if: always() && matrix.flavor == 'saas'
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: jacoco-exec-junit-jdk-${{ matrix.jdk-version }}
+          path: app/*/build/jacoco/*.exec
+          retention-days: 7
+          if-no-files-found: warn
+
+      - name: Add coverage to PR (flavor=${{ matrix.flavor }}, JDK=${{ matrix.jdk-version }})
         # The action only supports the pull_request event (it posts a PR comment),
         # so skip it for merge_group runs and workflow_dispatch.
         if: github.event_name == 'pull_request'

.github/workflows/build-enterprise.yml

@@ -78,7 +78,7 @@ jobs:
           cache: "npm"
           cache-dependency-path: frontend/package-lock.json
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Install Playwright (chromium only)
         run: task e2e:install -- chromium
       - name: Build frontend (needed for playwright's vite preview webServer)

.github/workflows/build.yml

@@ -68,6 +68,18 @@ jobs:
     uses: ./.github/workflows/backend-build.yml
     secrets: inherit
 
+  db-migration-test:
+    # Boots the current bootJar against H2 fixtures captured from past
+    # releases (v2.0.0 / v2.5.0 / v2.10.0) and verifies admin login still
+    # works after Hibernate's ddl-auto=update migrates the schema. Gated on
+    # the `project` filter so doc-only PRs skip this ~5-minute job.
+    if: needs.files-changed.outputs.project == 'true'
+    needs: [files-changed]
+    permissions:
+      contents: read
+    uses: ./.github/workflows/db-migration-test.yml
+    secrets: inherit
+
   check-generateOpenApiDocs:
     if: needs.files-changed.outputs.openapi == 'true'
     needs: [files-changed]
@@ -173,6 +185,24 @@ jobs:
     uses: ./.github/workflows/dependency-review.yml
     secrets: inherit
 
+  # Coverage aggregate: merges the JUnit + e2e:live + cucumber .exec
+  # artifacts produced by the jobs above into one report, plus pulls
+  # in vitest + Playwright frontend coverage for the per-area matrix.
+  # `if: always()` so a producer failing partway still gets credit
+  # for whatever did record. Advisory only - intentionally NOT in
+  # all-checks-passed, so a flaky aggregate run never blocks merging.
+  coverage-aggregate:
+    if: always()
+    needs:
+      - build
+      - playwright-e2e-live
+      - docker-compose-tests
+      - frontend-validation
+    permissions:
+      contents: read
+    uses: ./.github/workflows/coverage-aggregate.yml
+    secrets: inherit
+
   # Single status check that branch protection should mark as required.
   # Succeeds when every upstream job is either `success` or `skipped` (path-
   # gated jobs that didn't apply this run). Any `failure` or `cancelled`
@@ -184,6 +214,7 @@ jobs:
     needs:
       - files-changed
       - build
+      - db-migration-test
       - check-generateOpenApiDocs
       - frontend-validation
       - playwright-e2e
@@ -208,6 +239,7 @@ jobs:
           RESULTS: |
             files-changed=${{ needs.files-changed.result }}
             build=${{ needs.build.result }}
+            db-migration-test=${{ needs.db-migration-test.result }}
             check-generateOpenApiDocs=${{ needs.check-generateOpenApiDocs.result }}
             frontend-validation=${{ needs.frontend-validation.result }}
             playwright-e2e=${{ needs.playwright-e2e.result }}

.github/workflows/check-licence.yml

@@ -40,11 +40,11 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
           cache-disabled: true
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Check licenses for compatibility
         run: task backend:licenses:check
         env:

.github/workflows/check-openapi.yml

@@ -45,11 +45,11 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
           cache-disabled: true
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Generate OpenAPI documentation
         run: task backend:swagger
         env:

.github/workflows/check_toml.yml

@@ -166,16 +166,16 @@ jobs:
 
             // Determine reference file
             let referenceFilePath;
-            if (changedFiles.includes("frontend/editor/public/locales/en-GB/translation.toml")) {
+            if (changedFiles.includes("frontend/editor/public/locales/en-US/translation.toml")) {
               console.log("Using PR branch reference file.");
               const { data: fileContent } = await github.rest.repos.getContent({
                 owner: prRepoOwner,
                 repo: prRepoName,
-                path: "frontend/editor/public/locales/en-GB/translation.toml",
+                path: "frontend/editor/public/locales/en-US/translation.toml",
                 ref: branch,
               });
 
-              referenceFilePath = "pr-branch-translation-en-GB.toml";
+              referenceFilePath = "pr-branch-translation-en-US.toml";
               const content = Buffer.from(fileContent.content, "base64").toString("utf-8");
               fs.writeFileSync(referenceFilePath, content);
             } else {
@@ -183,11 +183,11 @@ jobs:
               const { data: fileContent } = await github.rest.repos.getContent({
                 owner: repoOwner,
                 repo: repoName,
-                path: "frontend/editor/public/locales/en-GB/translation.toml",
+                path: "frontend/editor/public/locales/en-US/translation.toml",
                 ref: "main",
               });
 
-              referenceFilePath = "main-branch-translation-en-GB.toml";
+              referenceFilePath = "main-branch-translation-en-US.toml";
               const content = Buffer.from(fileContent.content, "base64").toString("utf-8");
               fs.writeFileSync(referenceFilePath, content);
             }
@@ -293,6 +293,6 @@ jobs:
         run: |
           echo "Cleaning up temporary files..."
           rm -rf pr-branch
-          rm -f pr-branch-translation-en-GB.toml main-branch-translation-en-GB.toml changed_files.txt result.txt
+          rm -f pr-branch-translation-en-US.toml main-branch-translation-en-US.toml changed_files.txt result.txt
           echo "Cleanup complete."
         continue-on-error: true # Ensure cleanup runs even if previous steps fail

.github/workflows/coverage-aggregate.yml

@@ -0,0 +1,230 @@
+name: Aggregate backend coverage
+
+# Reusable workflow called from build.yml after every backend coverage
+# producer (JUnit, e2e:live, cucumber) has run. Downloads each job's raw
+# .exec, merges them into one JaCoCo report, and posts a combined step
+# summary alongside the per-source ones.
+#
+# Kept separate from the per-source jobs so:
+#   - the per-source jobs stay fast and independent (no cross-job waits)
+#   - this job can `if: always()` and still produce something useful when
+#     one of the producers fails partway through
+#   - frontend producers can be added later without touching the
+#     producers themselves
+on:
+  workflow_call:
+
+permissions:
+  contents: read
+
+jobs:
+  pick:
+    uses: ./.github/workflows/_runner-pick.yml
+
+  aggregate:
+    needs: pick
+    runs-on: ${{ needs.pick.outputs.is_fork == 'true' && 'ubuntu-latest' || 'depot-ubuntu-24.04-4' }}
+    timeout-minutes: 15
+    steps:
+      - name: Harden Runner
+        uses: step-security/harden-runner@ab7a9404c0f3da075243ca237b5fac12c98deaa5 # v2.19.3
+        with:
+          egress-policy: audit
+      - name: Checkout repository
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Set up JDK 25
+        uses: actions/setup-java@be666c2fcd27ec809703dec50e508c2fdc7f6654 # v5.2.0
+        with:
+          java-version: "25"
+          distribution: "temurin"
+
+      - name: Cache Gradle dependency artifacts
+        uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
+        with:
+          path: |
+            ~/.gradle/wrapper
+            ~/.gradle/caches/modules-2/files-2.1
+            ~/.gradle/caches/modules-2/metadata-2.*
+          key: gradle-deps-${{ runner.os }}-jdk-25-${{ hashFiles('**/gradle/wrapper/gradle-wrapper.properties', '**/*.gradle', '**/*.gradle.kts', 'settings.gradle', 'settings.gradle.kts', 'gradle/libs.versions.toml') }}
+
+      - name: Setup Gradle
+        uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
+        with:
+          gradle-version: 9.3.1
+          cache-disabled: true
+
+      - name: Set up Python
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.12"
+
+      - name: Install defusedxml for coverage scripts
+        # Both coverage-summary.py and coverage-matrix.py parse JaCoCo
+        # XML through defusedxml - see the script headers for context.
+        run: python -m pip install --quiet defusedxml
+
+      # Pattern matches every artifact this PR's producers might upload:
+      #   jacoco-exec-junit-jdk-25            (uploaded only by the saas
+      #                                         leg of backend-build, which
+      #                                         is a strict superset of the
+      #                                         core + proprietary legs)
+      #   jacoco-exec-e2e-live
+      #   jacoco-exec-cucumber
+      # Each lands as a sibling dir under coverage-execs/, with the .exec
+      # files preserving their original relative paths.
+      - name: Download all .exec artifacts
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v6.0.0
+        with:
+          pattern: jacoco-exec-*
+          path: coverage-execs/
+          merge-multiple: false
+        continue-on-error: true
+
+      - name: Inventory .exec files
+        id: inventory
+        # Splits the downloaded artifacts into two buckets:
+        #   * e2e-only  = cucumber + Playwright live (user-flow coverage)
+        #   * all       = the above plus JUnit (everything we test)
+        #
+        # Bucketing is by artifact-name prefix: download-artifact preserves
+        # the artifact name as the top-level dir, so JUnit's `.exec`s live
+        # under coverage-execs/jacoco-exec-junit-*/... while the others
+        # are under coverage-execs/jacoco-exec-{e2e-live,cucumber}/...
+        #
+        # If nothing was uploaded (e.g. all producers crashed before
+        # writing) we exit gracefully so this advisory job never fails CI.
+        run: |
+          mapfile -t all_execs < <(find coverage-execs -name '*.exec' -type f | sort)
+          mapfile -t e2e_execs < <(find coverage-execs -name '*.exec' -type f -not -path '*/jacoco-exec-junit-*' | sort)
+          if [ "${#all_execs[@]}" -eq 0 ]; then
+            echo "::warning::No .exec artifacts found - skipping aggregate report"
+            echo "found_all=false" >> "$GITHUB_OUTPUT"
+            echo "found_e2e=false" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+          printf 'All %d .exec files:\n' "${#all_execs[@]}"
+          printf '  %s\n' "${all_execs[@]}"
+          IFS=','; all_joined="${all_execs[*]}"
+          echo "files_all=$all_joined" >> "$GITHUB_OUTPUT"
+          echo "found_all=true" >> "$GITHUB_OUTPUT"
+          if [ "${#e2e_execs[@]}" -eq 0 ]; then
+            echo "::notice::No e2e/cucumber .exec files - e2e-only report will be skipped"
+            echo "found_e2e=false" >> "$GITHUB_OUTPUT"
+          else
+            printf 'E2E-only %d .exec files:\n' "${#e2e_execs[@]}"
+            printf '  %s\n' "${e2e_execs[@]}"
+            unset IFS
+            IFS=','; e2e_joined="${e2e_execs[*]}"
+            echo "files_e2e=$e2e_joined" >> "$GITHUB_OUTPUT"
+            echo "found_e2e=true" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Compile classes for JaCoCo class lookup
+        # jacocoReportFromExec only needs the compiled .class files
+        # under each subproject's build/classes/java/main/. `classes`
+        # (compileJava + processResources) is enough; we skipped the
+        # heavier `assemble` to avoid building bootJar / fat jars that
+        # add 60+ seconds per run for no gain to the report.
+        if: steps.inventory.outputs.found_all == 'true'
+        run: ./gradlew classes -PnoSpotless
+
+      - name: Generate e2e-only JaCoCo report
+        # "Real user-flow" coverage: only counts code reached by an actual
+        # HTTP request from cucumber or live Playwright. Useful for
+        # questions like "how much of our backend does a user actually
+        # hit?". Skipped when neither producer uploaded a .exec.
+        if: steps.inventory.outputs.found_e2e == 'true'
+        run: |
+          ./gradlew jacocoReportFromExec \
+            -PexecFile="${{ steps.inventory.outputs.files_e2e }}" \
+            -PreportDir=build/reports/jacoco/aggregate-e2e \
+            -PnoSpotless
+
+      - name: Generate combined JaCoCo report (everything)
+        if: steps.inventory.outputs.found_all == 'true'
+        run: |
+          ./gradlew jacocoReportFromExec \
+            -PexecFile="${{ steps.inventory.outputs.files_all }}" \
+            -PreportDir=build/reports/jacoco/aggregate-all \
+            -PnoSpotless
+
+      - name: E2E-only step summary
+        # Rendered first so it gets prime real estate in the Summary
+        # tab - this is the number most readers actually want
+        # ("how much of the backend do real user flows cover?").
+        if: steps.inventory.outputs.found_e2e == 'true'
+        run: |
+          python scripts/coverage-summary.py \
+            --title "Real user-flow backend coverage (e2e:live + cucumber)" \
+            --jacoco "merged=build/reports/jacoco/aggregate-e2e/jacocoTestReport.xml" \
+            --github-step-summary
+
+      - name: ALL-sources step summary
+        # Separate call (not a multi-input one) because the helper's
+        # rightmost "Aggregate" column would sum the two reports - which
+        # is meaningless when one is a strict superset of the other.
+        if: steps.inventory.outputs.found_all == 'true'
+        run: |
+          python scripts/coverage-summary.py \
+            --title "Combined backend coverage (JUnit + e2e:live + cucumber)" \
+            --jacoco "merged=build/reports/jacoco/aggregate-all/jacocoTestReport.xml" \
+            --github-step-summary
+
+      - name: Upload combined aggregate report
+        if: steps.inventory.outputs.found_all == 'true'
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: jacoco-aggregate-all-${{ github.run_id }}
+          path: build/reports/jacoco/aggregate-all/
+          retention-days: 14
+
+      - name: Upload e2e-only aggregate report
+        if: steps.inventory.outputs.found_e2e == 'true'
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: jacoco-aggregate-e2e-${{ github.run_id }}
+          path: build/reports/jacoco/aggregate-e2e/
+          retention-days: 14
+
+      # --------------------------------------------------------------
+      # Per-area matrix: rolls backend + frontend coverage into one
+      # table indexed by core/proprietary/saas/desktop. Pulls the
+      # frontend artifacts now (after the JaCoCo step has done its
+      # work) so the per-source backend summaries still render first
+      # even if the matrix step fails.
+      # --------------------------------------------------------------
+      - name: Download vitest coverage artifact
+        # frontend-validation uploads as `frontend-coverage`. Tolerate
+        # absence so a backend-only PR still produces the matrix with
+        # just backend rows populated.
+        if: always()
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v6.0.0
+        with:
+          name: frontend-coverage
+          path: matrix-inputs/vitest/
+        continue-on-error: true
+
+      - name: Download Playwright frontend coverage artifact
+        # e2e-live uploads as `playwright-frontend-coverage-<run_id>`.
+        # Same tolerance as vitest - matrix script handles missing inputs.
+        if: always()
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v6.0.0
+        with:
+          name: playwright-frontend-coverage-${{ github.run_id }}
+          path: matrix-inputs/playwright/
+        continue-on-error: true
+
+      - name: Coverage matrix step summary
+        if: always()
+        # Matrix references the two aggregate JaCoCo XMLs (already
+        # generated above) plus whichever frontend artifacts landed.
+        # Every input is optional; missing ones render as "-".
+        run: |
+          python scripts/coverage-matrix.py \
+            ${{ steps.inventory.outputs.found_all == 'true' && '--jacoco-all build/reports/jacoco/aggregate-all/jacocoTestReport.xml' || '' }} \
+            ${{ steps.inventory.outputs.found_e2e == 'true' && '--jacoco-e2e build/reports/jacoco/aggregate-e2e/jacocoTestReport.xml' || '' }} \
+            --vitest matrix-inputs/vitest/coverage-summary.json \
+            --playwright-frontend matrix-inputs/playwright/coverage-pw-summary/coverage-summary.json \
+            --title "Coverage matrix (per-area, e2e vs all)" \
+            --github-step-summary

.github/workflows/db-migration-test.yml

@@ -0,0 +1,93 @@
+name: DB migration smoke test
+
+# Boots the current Stirling-PDF JAR against H2 fixtures captured from past
+# releases (v2.0.0 / v2.5.0 / v2.10.0) and verifies admin login still works.
+# Catches schema changes that would break existing user databases under
+# Hibernate's `ddl-auto=update` upgrade path.
+
+on:
+  workflow_call:
+
+permissions:
+  contents: read
+
+jobs:
+  pick:
+    uses: ./.github/workflows/_runner-pick.yml
+
+  migration-test:
+    needs: pick
+    runs-on: ${{ needs.pick.outputs.is_fork == 'true' && 'ubuntu-latest' || 'depot-ubuntu-24.04-8' }}
+    timeout-minutes: 30
+    env:
+      DEPOT_TOKEN: ${{ secrets.DEPOT_TOKEN }}
+    steps:
+      - name: Harden Runner
+        uses: step-security/harden-runner@ab7a9404c0f3da075243ca237b5fac12c98deaa5 # v2.19.3
+        with:
+          egress-policy: audit
+
+      - name: Checkout repository
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Set up JDK 25
+        uses: actions/setup-java@be666c2fcd27ec809703dec50e508c2fdc7f6654 # v5.2.0
+        with:
+          java-version: 25
+          distribution: temurin
+
+      - name: Cache Gradle dependency artifacts
+        uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
+        with:
+          path: |
+            ~/.gradle/wrapper
+            ~/.gradle/caches/modules-2/files-2.1
+            ~/.gradle/caches/modules-2/metadata-2.*
+          key: gradle-deps-${{ runner.os }}-jdk-25-${{ hashFiles('**/gradle/wrapper/gradle-wrapper.properties', '**/*.gradle', '**/*.gradle.kts', 'settings.gradle', 'settings.gradle.kts', 'gradle/libs.versions.toml') }}
+
+      - name: Setup Gradle
+        uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
+        with:
+          gradle-version: 9.5.1
+          cache-disabled: true
+
+      # No `-PnoSpotless` here yet because the upstream cache layer matches the
+      # backend build's; reuse keeps cold-cache cost identical.
+      - name: Build Stirling-PDF JAR
+        env:
+          MAVEN_USER: ${{ secrets.MAVEN_USER }}
+          MAVEN_PASSWORD: ${{ secrets.MAVEN_PASSWORD }}
+          MAVEN_PUBLIC_URL: ${{ secrets.MAVEN_PUBLIC_URL }}
+        run: ./gradlew :stirling-pdf:bootJar -PnoSpotless --no-daemon
+
+      - name: Locate built JAR
+        id: jar
+        run: |
+          jar=$(find app/core/build/libs -maxdepth 1 -name 'Stirling-PDF*.jar' -o -name 'stirling-pdf*.jar' 2>/dev/null \
+              | grep -vE '(-plain|-sources)\.jar$' | head -n 1)
+          if [[ -z "$jar" ]]; then
+            echo "::error::No JAR under app/core/build/libs"
+            ls -lah app/core/build/libs || true
+            exit 1
+          fi
+          # Absolute path - the migration script pushd's into a temp workdir
+          # before invoking java, which would dangle a relative path.
+          jar=$(realpath "$jar")
+          echo "path=$jar" >> "$GITHUB_OUTPUT"
+          echo "Built JAR: $jar"
+
+      - name: Run migration smoke test
+        env:
+          STIRLING_JAR: ${{ steps.jar.outputs.path }}
+        run: bash scripts/db-migration/run-migration-test.sh
+
+      - name: Upload app logs on failure
+        if: failure()
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: db-migration-app-logs
+          # Path matches the preserved workdir in run-migration-test.sh -
+          # only failing fixtures leave a directory behind.
+          path: /tmp/stirling-migration-failed-*/app.log
+          retention-days: 7
+          if-no-files-found: warn

.github/workflows/deploy-on-v2-commit.yml

@@ -107,7 +107,7 @@ jobs:
 
       - name: Build and push frontend image (Depot)
         if: env.USE_DEPOT == 'true' && steps.check-frontend.outputs.exists == 'false'
-        uses: depot/build-push-action@5f3b3c2e5a00f0093de47f657aeaefcedff27d18 # v1.16.0
+        uses: depot/build-push-action@98e78adca7817480b8185f474a400b451d74e287 # v1.16.0
         with:
           project: ${{ vars.DEPOT_PROJECT_ID }}
           context: .
@@ -136,7 +136,7 @@ jobs:
 
       - name: Build and push backend image (Depot)
         if: env.USE_DEPOT == 'true' && steps.check-backend.outputs.exists == 'false'
-        uses: depot/build-push-action@5f3b3c2e5a00f0093de47f657aeaefcedff27d18 # v1.16.0
+        uses: depot/build-push-action@98e78adca7817480b8185f474a400b451d74e287 # v1.16.0
         with:
           project: ${{ vars.DEPOT_PROJECT_ID }}
           context: .
@@ -188,7 +188,7 @@ jobs:
               environment:
                 DISABLE_ADDITIONAL_FEATURES: "true"
                 SECURITY_ENABLELOGIN: "false"
-                SYSTEM_DEFAULTLOCALE: en-GB
+                SYSTEM_DEFAULTLOCALE: en-US
                 UI_APPNAME: "Stirling-PDF V2"
                 UI_HOMEDESCRIPTION: "V2 Frontend/Backend Split"
                 UI_APPNAMENAVBAR: "V2 Deployment"

.github/workflows/docker-compose-tests.yml

@@ -61,7 +61,7 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
           cache-disabled: true
 
       - name: Set up Docker Buildx
@@ -87,6 +87,12 @@ jobs:
         run: |
           pip install --require-hashes --only-binary=:all: -r ./testing/cucumber/requirements.txt
 
+      - name: Extract JaCoCo agent for cucumber coverage
+        # Stages build/jacoco/jacocoagent.jar where the coverage override
+        # file bind-mounts it into the cucumber container. The agent jar
+        # never goes into the published image - this is host-only.
+        run: ./gradlew copyJacocoAgent -PnoSpotless
+
       - name: Run Docker Compose Tests
         run: |
           chmod +x ./testing/test_webpages.sh
@@ -98,6 +104,62 @@ jobs:
           MAVEN_PASSWORD: ${{ secrets.MAVEN_PASSWORD }}
           MAVEN_PUBLIC_URL: ${{ secrets.MAVEN_PUBLIC_URL }}
           DOCKER_BASE_CHANGED: ${{ inputs.docker-base-changed }}
+          # Tells test.sh to layer testing/compose/docker-compose-coverage.override.yml
+          # over the cucumber compose so the container starts with the
+          # JaCoCo agent attached via JAVA_CUSTOM_OPTS.
+          STIRLING_PDF_TEST_COVERAGE: "1"
+
+      - name: Generate cucumber JaCoCo report
+        # `if: always()` so a behave failure still produces partial
+        # coverage from whatever endpoints did run. The exec file only
+        # exists when the container shut down cleanly - guard so the step
+        # is silent on the (rare) crash path.
+        if: always()
+        id: cucumber-coverage
+        run: |
+          if [ -s testing/cucumber-coverage/cucumber.exec ]; then
+            ./gradlew jacocoReportFromExec \
+              -PexecFile=testing/cucumber-coverage/cucumber.exec \
+              -PreportDir=build/reports/jacoco/cucumber \
+              -PnoSpotless
+            echo "report=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "::warning::No cucumber .exec at testing/cucumber-coverage/cucumber.exec (container may have crashed before flushing)"
+            echo "report=false" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Install defusedxml for coverage summary
+        # coverage-summary.py parses JaCoCo XML through defusedxml -
+        # see the script header for context.
+        if: always() && steps.cucumber-coverage.outputs.report == 'true'
+        run: python -m pip install --quiet defusedxml
+
+      - name: Cucumber coverage step summary
+        if: always() && steps.cucumber-coverage.outputs.report == 'true'
+        run: |
+          python scripts/coverage-summary.py \
+            --title "Cucumber (docker) JaCoCo coverage" \
+            --jacoco "cucumber=build/reports/jacoco/cucumber/jacocoTestReport.xml" \
+            --github-step-summary
+
+      - name: Upload cucumber JaCoCo report
+        if: always() && steps.cucumber-coverage.outputs.report == 'true'
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: jacoco-cucumber-${{ github.run_id }}
+          path: build/reports/jacoco/cucumber/
+          retention-days: 7
+
+      - name: Upload raw cucumber .exec for aggregate merge
+        # Picked up by the coverage-aggregate workflow via the
+        # `jacoco-exec-*` artifact name pattern.
+        if: always() && steps.cucumber-coverage.outputs.report == 'true'
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: jacoco-exec-cucumber
+          path: testing/cucumber-coverage/cucumber.exec
+          retention-days: 7
+          if-no-files-found: warn
 
       - name: Upload Cucumber Report
         if: always()

.github/workflows/e2e-live.yml

@@ -42,16 +42,139 @@ jobs:
           cache: "npm"
           cache-dependency-path: frontend/package-lock.json
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Install Playwright (chromium only)
         run: task e2e:install -- chromium
       - name: Build frontend (production bundle for vite preview)
         env:
           VITE_BUILD_FOR_PREVIEW: "1"
         run: task frontend:build
-      - name: Run live E2E tests (chromium)
+      - name: Run live E2E tests (chromium) with coverage
         id: live-tests
+        env:
+          # Attaches the JaCoCo agent to the bootRun JVM (see
+          # .taskfiles/e2e.yml live:backend). The .exec gets flushed on
+          # graceful shutdown when the runner traps EXIT/INT/TERM, so the
+          # report step below sees a populated file.
+          COVERAGE: "1"
+          # Tells the Playwright fixture (test-base.ts) to capture per-test
+          # V8 JS coverage. Raw dumps land under
+          # .test-state/playwright/coverage-pw/ for the post-process step
+          # to aggregate. Chromium-only - other engines silently skip.
+          PW_COVERAGE: "1"
         run: task e2e:live
+      - name: Generate JaCoCo report from e2e:live .exec
+        if: always()
+        id: live-coverage
+        # `if: always()` so even a failed test run still produces a
+        # report from whatever flows did exercise the backend before
+        # the failure. The task itself tolerates a missing .exec
+        # (jacoco emits an empty report rather than crashing) but we
+        # guard with `test -s` to keep the job log clean.
+        run: |
+          if [ -s .test-state/playwright/jacoco.exec ]; then
+            ./gradlew jacocoReportFromExec \
+              -PexecFile=.test-state/playwright/jacoco.exec \
+              -PreportDir=build/reports/jacoco/e2e-live \
+              -PnoSpotless
+            echo "report=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "::warning::No e2e:live .exec found at .test-state/playwright/jacoco.exec; skipping report"
+            echo "report=false" >> "$GITHUB_OUTPUT"
+          fi
+      - name: Set up Python for coverage summary
+        if: always() && steps.live-coverage.outputs.report == 'true'
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.12"
+      - name: Install defusedxml for coverage summary
+        # coverage-summary.py uses defusedxml instead of stdlib xml.etree
+        # to dodge XXE / billion-laughs scanner findings.
+        if: always() && steps.live-coverage.outputs.report == 'true'
+        run: python -m pip install --quiet defusedxml
+      - name: e2e:live coverage step summary
+        if: always() && steps.live-coverage.outputs.report == 'true'
+        run: |
+          python scripts/coverage-summary.py \
+            --title "Playwright (live backend) JaCoCo coverage" \
+            --jacoco "e2e-live=build/reports/jacoco/e2e-live/jacocoTestReport.xml" \
+            --github-step-summary
+      - name: Upload e2e:live JaCoCo report
+        if: always() && steps.live-coverage.outputs.report == 'true'
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: jacoco-e2e-live-${{ github.run_id }}
+          path: build/reports/jacoco/e2e-live/
+          retention-days: 7
+
+      - name: Upload raw e2e:live .exec for aggregate merge
+        # Picked up by the coverage-aggregate workflow via the
+        # `jacoco-exec-*` artifact name pattern.
+        if: always() && steps.live-coverage.outputs.report == 'true'
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: jacoco-exec-e2e-live
+          path: .test-state/playwright/jacoco.exec
+          retention-days: 7
+          if-no-files-found: warn
+
+      - name: Set up Python for frontend coverage summary
+        # Separate from the backend-coverage python step because the
+        # frontend path doesn't depend on a JaCoCo report - it produces
+        # a summary even on backend failure, as long as some Playwright
+        # tests ran far enough to dump V8 coverage.
+        if: always()
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.12"
+
+      - name: Install defusedxml for frontend coverage summary
+        # Idempotent re-install: the backend-coverage step may have
+        # installed it already, but this leg can run on its own when the
+        # backend report step skips (e.g. .exec missing).
+        if: always()
+        run: python -m pip install --quiet defusedxml
+
+      - name: Aggregate Playwright frontend (V8) coverage
+        # Rolls per-test V8 dumps from the test-base fixture into one
+        # vitest-shaped coverage-summary.json. Tolerates a missing dump
+        # dir (firefox/webkit runs, or a failure before any test got
+        # far enough to dump).
+        if: always()
+        id: pw-frontend-coverage
+        run: |
+          if [ -d .test-state/playwright/coverage-pw ] && \
+             find .test-state/playwright/coverage-pw -name '*.json' -type f | grep -q .; then
+            python scripts/playwright-coverage-summary.py \
+              .test-state/playwright/coverage-pw \
+              --out .test-state/playwright/coverage-pw-summary/coverage-summary.json
+            echo "summary=true" >> "$GITHUB_OUTPUT"
+          else
+            echo "::notice::No Playwright frontend coverage dumps found (chromium-only feature)"
+            echo "summary=false" >> "$GITHUB_OUTPUT"
+          fi
+
+      - name: Playwright frontend coverage step summary
+        if: always() && steps.pw-frontend-coverage.outputs.summary == 'true'
+        run: |
+          python scripts/coverage-summary.py \
+            --title "Playwright (live) frontend coverage" \
+            --vitest .test-state/playwright/coverage-pw-summary/coverage-summary.json \
+            --github-step-summary
+
+      - name: Upload Playwright frontend coverage
+        # Bundle both the aggregated summary and the raw V8 dumps so
+        # someone debugging "why is this function showing as covered"
+        # can trace it back to the source dump.
+        if: always() && steps.pw-frontend-coverage.outputs.summary == 'true'
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: playwright-frontend-coverage-${{ github.run_id }}
+          path: |
+            .test-state/playwright/coverage-pw-summary/
+            .test-state/playwright/coverage-pw/
+          retention-days: 7
+
       - name: Print backend log on failure
         if: failure() && steps.live-tests.conclusion == 'failure'
         run: |
@@ -65,10 +188,30 @@ jobs:
           name: backend-log-live-${{ github.run_id }}
           path: .test-state/playwright/backend.log
           retention-days: 7
-      - name: Upload Playwright report
+      - name: List Playwright output locations (debug)
+        if: always()
+        run: |
+          echo "::group::Playwright output dirs"
+          # Playwright anchors its default outputDir + HTML report to the
+          # nearest package.json, which is frontend/ (frontend/editor has
+          # none), so artifacts land under frontend/, not frontend/editor/.
+          ls -la frontend/playwright-report 2>/dev/null \
+            || echo "no playwright-report at frontend/"
+          ls -la frontend/test-results 2>/dev/null \
+            || echo "no test-results at frontend/"
+          find . -name node_modules -prune -o -name 'trace.zip' -print 2>/dev/null || true
+          echo "::endgroup::"
+      - name: Upload Playwright report + traces
         if: always()
         uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
         with:
           name: playwright-report-live-${{ github.run_id }}
-          path: frontend/editor/playwright-report/
+          # test-results/ holds the per-test trace.zip (with browser console
+          # logs) + screenshots/video; playwright-report/ is the HTML report.
+          # Both live under frontend/ (Playwright anchors them to the nearest
+          # package.json, which is frontend/; frontend/editor has none).
+          path: |
+            frontend/playwright-report/
+            frontend/test-results/
           retention-days: 7
+          if-no-files-found: warn

.github/workflows/e2e-stubbed.yml

@@ -36,7 +36,7 @@ jobs:
           cache: "npm"
           cache-dependency-path: frontend/package-lock.json
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Install Playwright (chromium only)
         run: task e2e:install -- chromium
       - name: Build frontend (production bundle for vite preview)

.github/workflows/frontend-backend-licenses-update.yml

@@ -97,7 +97,7 @@ jobs:
         run: npm ci --ignore-scripts --audit=false --fund=false
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Generate frontend license report (internal PR)
         if: github.event_name == 'pull_request' && github.event.pull_request.head.repo.fork == false
         env:
@@ -110,8 +110,8 @@ jobs:
           NPM_CONFIG_IGNORE_SCRIPTS: "true"
         working-directory: frontend
         run: |
-          mkdir -p src/assets
-          npx --yes license-report --only=prod --output=json > src/assets/3rdPartyLicenses.json
+          mkdir -p editor/src/assets
+          npx --yes license-report --only=prod --output=json > editor/src/assets/3rdPartyLicenses.json
 
       - name: Postprocess with project script (BASE version)
         if: github.event_name == 'pull_request' && github.event.pull_request.head.repo.fork == true
@@ -349,10 +349,10 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Check licenses and generate report
         id: license-check
         run: task backend:licenses:generate || echo "LICENSE_CHECK_FAILED=true" >> $GITHUB_ENV

.github/workflows/frontend-validation.yml

@@ -31,7 +31,7 @@ jobs:
           cache: "npm"
           cache-dependency-path: frontend/package-lock.json
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Quality-check frontend
         id: frontend-check
         run: task frontend:check:all
@@ -109,6 +109,41 @@ jobs:
                 comment_id: existing.id,
               });
             }
+      - name: Vitest coverage
+        # Separate from `frontend:check:all` so the quality-gate run stays
+        # uninstrumented (faster signal) and coverage stays an informational
+        # follow-up. Continue-on-error keeps the workflow green even when
+        # a handful of test files refuse to import (e.g. missing icon
+        # specifiers) - the summary still gets posted with whatever
+        # vitest managed to instrument.
+        id: frontend-coverage
+        continue-on-error: true
+        run: task frontend:test:coverage
+      - name: Set up Python for coverage summary
+        if: always()
+        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
+        with:
+          python-version: "3.12"
+      - name: Install defusedxml for coverage summary
+        # See coverage-summary.py header - it parses XML through defusedxml
+        # to dodge the stdlib parser's exposure to XXE / billion-laughs.
+        if: always()
+        run: python -m pip install --quiet defusedxml
+      - name: Vitest coverage step summary
+        if: always()
+        run: |
+          python scripts/coverage-summary.py \
+            --title "Frontend Vitest coverage" \
+            --vitest frontend/editor/coverage/coverage-summary.json \
+            --github-step-summary
+      - name: Upload vitest coverage report
+        if: always()
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: frontend-coverage
+          path: frontend/editor/coverage/
+          retention-days: 7
+          if-no-files-found: warn
       - name: Upload frontend build artifacts
         uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
         with:

.github/workflows/multiOSReleases.yml

@@ -73,10 +73,10 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Get version number
         id: versionNumber
         run: |
@@ -148,7 +148,7 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Setup Node.js
         if: matrix.variant.build_frontend == true
@@ -159,7 +159,7 @@ jobs:
           cache-dependency-path: frontend/package-lock.json
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
 
       - name: Build JAR
         run: ./gradlew build ${{ matrix.variant.build_frontend && '-PbuildWithFrontend=true' || '' }} -x spotlessApply -x spotlessCheck -x test -x sonarqube
@@ -252,10 +252,10 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
 
       # Build the universal JRE before desktop:prepare so the jlink:runtime
       # task short-circuits on its `test -d runtime/jre` status check.
@@ -442,10 +442,6 @@ jobs:
           echo "Generated tauri.windows.conf.json (alias masked):"
           sed "s/${KEYPAIR_ALIAS}/***/g" ./frontend/editor/src-tauri/tauri.windows.conf.json
 
-      - name: Sign JPDFium dylibs inside bootJar (macOS only)
-        if: matrix.platform == 'macos-15' && env.APPLE_CERTIFICATE != ''
-        run: bash frontend/scripts/sign-jpdfium-dylibs-in-bootjar.sh
-
       - name: Import release GPG signing key (Linux)
         if: matrix.platform == 'ubuntu-22.04' && env.RELEASE_GPG_PRIVATE_KEY != '' && (github.event_name == 'release' || (github.event_name == 'workflow_dispatch' && github.event.inputs.sign != 'false') || github.ref == 'refs/heads/V2-master')
         run: |
@@ -495,6 +491,7 @@ jobs:
           projectPath: ./frontend/editor
           tauriScript: npx tauri
           args: ${{ matrix.args }}
+          updaterJsonKeepUniversal: true
 
       - name: Clear release GPG key from runner keyring (Linux)
         if: always() && matrix.platform == 'ubuntu-22.04' && env.RELEASE_GPG_PRIVATE_KEY != '' && (github.event_name == 'release' || (github.event_name == 'workflow_dispatch' && github.event.inputs.sign != 'false') || github.ref == 'refs/heads/V2-master')
@@ -586,21 +583,37 @@ jobs:
         if: always() && steps.digicert-setup.conclusion != 'failure'
         shell: bash
         run: |
-          mkdir -p ./dist
+          # Absolute dist path so the cd below can't break the copy targets.
+          DIST="$GITHUB_WORKSPACE/dist"
+          mkdir -p "$DIST"
           cd ./frontend/editor/src-tauri/target
 
-          # Find and rename artifacts based on platform
+          echo "=== tauri bundle artifacts ==="
+          find . -path "*/bundle/*" \( -name "*.msi" -o -name "*.deb" \
+            -o -name "*.rpm" -o -name "*.AppImage" -o -name "*.dmg" \
+            -o -name "*.app.tar.gz" -o -name "*.sig" \) 2>/dev/null | sort || true
+          echo "=============================="
+
+          # createUpdaterArtifacts:true signs the native installers in place;
+          # each <bundle> ships with a sibling <bundle>.sig consumed by latest.json.
           if [ "${{ matrix.platform }}" = "windows-latest" ]; then
-            # Only ship the MSI installer on Windows. The loose exe and WiX toolset exes
-            # are not the user-facing installer - the MSI contains the signed inner exe.
-            find . -name "*.msi" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.msi" \;
+            find . -name "*.msi" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.msi" \;
+            find . -name "*.msi.sig" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.msi.sig" \;
           elif [ "${{ matrix.platform }}" = "macos-15" ]; then
-            find . -name "*.dmg" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.dmg" \;
-            find . -name "*.app" -exec cp -r {} "../../../dist/Stirling-PDF-${{ matrix.name }}.app" \;
+            # DMG = manual install; .app.tar.gz (+ .sig) = updater payload.
+            # Raw .app is intentionally not shipped (hundreds of MB of uncompressed input).
+            find . -name "*.dmg" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.dmg" \;
+            find . -name "*.app.tar.gz" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.app.tar.gz" \;
+            find . -name "*.app.tar.gz.sig" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.app.tar.gz.sig" \;
           else
-            find . -name "*.deb" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.deb" \;
-            find . -name "*.rpm" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.rpm" \;
-            find . -name "*.AppImage" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.AppImage" \;
+            # The raw .AppImage IS its updater payload (signed -> .AppImage.sig),
+            # not a .tar.gz wrapper - that's only produced under v1Compatible.
+            find . -name "*.deb" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.deb" \;
+            find . -name "*.deb.sig" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.deb.sig" \;
+            find . -name "*.rpm" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.rpm" \;
+            find . -name "*.rpm.sig" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.rpm.sig" \;
+            find . -name "*.AppImage" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.AppImage" \;
+            find . -name "*.AppImage.sig" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.AppImage.sig" \;
           fi
 
       - name: Upload build artifacts
@@ -611,8 +624,7 @@ jobs:
           path: ./dist/*
           retention-days: 1
 
-  create-release:
-    if: (github.event_name == 'workflow_dispatch' && github.event.inputs.test_mode != 'true') || github.event_name == 'release' || github.ref == 'refs/heads/V2-master'
+  collect-and-release:
     needs: [pick, determine-matrix, build, build-jars]
     runs-on: ${{ needs.pick.outputs.is_fork == 'true' && 'ubuntu-latest' || 'depot-ubuntu-24.04-4' }}
     permissions:
@@ -623,6 +635,16 @@ jobs:
         with:
           egress-policy: audit
 
+      # Sparse-check out the verifier + pubkey before the artifact downloads
+      # so the checkout cannot clobber ./artifacts.
+      - name: Checkout updater verifier
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          sparse-checkout: |
+            .github/scripts/verify-updater-signatures.py
+            frontend/editor/src-tauri/tauri.conf.json
+          sparse-checkout-cone-mode: false
+
       - name: Download all Tauri artifacts
         uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
         with:
@@ -650,17 +672,146 @@ jobs:
       - name: Display structure of downloaded files
         run: ls -R ./artifacts
 
+      # tauri-action only emits latest.json when it also publishes the release
+      # (tagName/releaseId set). We publish separately via action-gh-release,
+      # so build latest.json here from the per-platform .sig files.
+      - name: Generate updater latest.json
+        env:
+          VERSION: ${{ needs.determine-matrix.outputs.version }}
+          TAG: v${{ needs.determine-matrix.outputs.version }}
+          REPO: ${{ github.repository }}
+        run: |
+          python3 - << 'PYEOF'
+          import json, os, sys
+          from pathlib import Path
+          from datetime import datetime, timezone
+
+          VERSION = os.environ['VERSION']
+          TAG = os.environ['TAG']
+          REPO = os.environ['REPO']
+
+          ART = Path('./artifacts/tauri')
+
+          # Tauri updater looks up {os}-{arch}-{installer} (e.g. linux-x86_64-deb)
+          # before bare {os}-{arch}, so per-format Linux keys let deb/rpm/appimage
+          # each self-update from their matching file. macOS universal serves both
+          # arches from the one .app.tar.gz.
+          PLATFORM_MAP = [
+              {
+                  'bundles': ['Stirling-PDF-linux-x86_64.deb'],
+                  'targets': ['linux-x86_64-deb'],
+              },
+              {
+                  'bundles': ['Stirling-PDF-linux-x86_64.rpm'],
+                  'targets': ['linux-x86_64-rpm'],
+              },
+              {
+                  'bundles': ['Stirling-PDF-linux-x86_64.AppImage'],
+                  'targets': ['linux-x86_64-appimage'],
+              },
+              {
+                  'bundles': ['Stirling-PDF-windows-x86_64.msi'],
+                  'targets': ['windows-x86_64-msi', 'windows-x86_64'],
+              },
+              {
+                  'bundles': ['Stirling-PDF-macos-universal.app.tar.gz'],
+                  'targets': ['darwin-x86_64', 'darwin-aarch64'],
+              },
+          ]
+
+          # rglob() because download-artifact varies layout: one artifact -> flat,
+          # many -> nested under <artifact-name>/.
+          def find_signed(name):
+              for bundle_path in sorted(ART.rglob(name)):
+                  sig_path = bundle_path.with_name(bundle_path.name + '.sig')
+                  if sig_path.exists():
+                      return bundle_path, sig_path
+              return None
+
+          platforms = {}
+          skipped = []
+          for entry in PLATFORM_MAP:
+              picked = None
+              for name in entry['bundles']:
+                  picked = find_signed(name)
+                  if picked:
+                      break
+              if not picked:
+                  skipped.append(
+                      f"{entry['targets']} (no signed bundle among "
+                      f"{entry['bundles']} - TAURI_SIGNING_PRIVATE_KEY unset "
+                      f"or createUpdaterArtifacts disabled?)"
+                  )
+                  continue
+              bundle_path, sig_path = picked
+              signature = sig_path.read_text(encoding='utf-8').strip()
+              url = f"https://github.com/{REPO}/releases/download/{TAG}/{bundle_path.name}"
+              for target in entry['targets']:
+                  platforms[target] = {'signature': signature, 'url': url}
+              print(f"Added {entry['targets']} from {bundle_path.name}")
+
+          if skipped:
+              print("Skipped platforms:")
+              for s in skipped:
+                  print(f"  - {s}")
+
+          if not platforms:
+              print(
+                  "WARN: no signed updater bundles found - "
+                  "skipping latest.json generation"
+              )
+              sys.exit(0)
+
+          manifest = {
+              'version': VERSION,
+              'notes': f"See https://github.com/{REPO}/releases/tag/{TAG}",
+              'pub_date': datetime.now(timezone.utc).strftime('%Y-%m-%dT%H:%M:%SZ'),
+              'platforms': platforms,
+          }
+
+          out = Path('./artifacts/latest.json')
+          out.write_text(json.dumps(manifest, indent=2) + '\n', encoding='utf-8')
+          print(f"Generated {out} with platforms: {sorted(platforms.keys())}")
+          PYEOF
+
+      - name: Upload merged artifacts for review
+        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f # v7.0.0
+        with:
+          name: release-artifacts
+          path: ./artifacts/
+          retention-days: 7
+
+      # Gate publish on valid updater sigs. Runs after the review upload (so
+      # artifacts survive for debugging) and before action-gh-release.
+      - name: Verify updater signatures
+        run: |
+          python3 -m pip install --quiet 'cryptography==44.0.0'
+          python3 .github/scripts/verify-updater-signatures.py \
+            ./artifacts/tauri frontend/editor/src-tauri/tauri.conf.json
+
+      # workflow_dispatch path requires platform=='all' so a single-platform
+      # dispatch can't overwrite an existing release's full latest.json with a
+      # partial one (action-gh-release defaults overwrite_files:true).
+      # release / V2-master always build the full matrix so no extra guard needed.
+      # fail_on_unmatched_files makes a missing latest.json or installer fail loudly
+      # instead of silently shipping a broken auto-update.
       - name: Upload binaries to Release
+        if: (github.event_name == 'workflow_dispatch' && github.event.inputs.test_mode != 'true' && github.event.inputs.platform == 'all') || github.event_name == 'release' || github.ref == 'refs/heads/V2-master'
         uses: softprops/action-gh-release@b4309332981a82ec1c5618f44dd2e27cc8bfbfda # v3.0.0
         with:
           tag_name: v${{ needs.determine-matrix.outputs.version }}
           generate_release_notes: true
+          fail_on_unmatched_files: true
+          # Installers + updater payloads + manifest. .sig contents are embedded
+          # in latest.json so the .sig files themselves are not uploaded.
           files: |
             ./artifacts/**/*.jar
             ./artifacts/**/*.msi
             ./artifacts/**/*.dmg
+            ./artifacts/**/*.app.tar.gz
             ./artifacts/**/*.deb
             ./artifacts/**/*.rpm
             ./artifacts/**/*.AppImage
+            ./artifacts/latest.json
           draft: false
           prerelease: false

.github/workflows/nightly.yml

@@ -37,7 +37,7 @@ jobs:
           cache-dependency-path: frontend/package-lock.json
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Install all Playwright browsers
         run: task e2e:install
 

.github/workflows/push-docker-base.yml

@@ -75,7 +75,7 @@ jobs:
 
       - name: Generate tags for base image
         id: meta
-        uses: docker/metadata-action@030e881283bb7a6894de51c315a6bfe6a94e05cf # v6.0.0
+        uses: docker/metadata-action@80c7e94dd9b9319bd5eb7a0e0fe9291e23a2a2e9 # v6.1.0
         with:
           images: |
             ${{ secrets.DOCKER_HUB_ORG_USERNAME }}/stirling-pdf-base

.github/workflows/push-docker.yml

@@ -78,14 +78,14 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Set up Docker Buildx
         id: buildx
         uses: docker/setup-buildx-action@4d04d5d9486b7bd6fa91e7baf45bbb4f8b9deedd # v4.0.0
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Get version number
         id: versionNumber
         run: echo "versionNumber=$(./gradlew printVersion --quiet | tail -1)" >> $GITHUB_OUTPUT
@@ -129,7 +129,7 @@ jobs:
       - name: Generate tags for latest
         id: meta
         if: env.RUN_MAIN_APP == 'true'
-        uses: docker/metadata-action@030e881283bb7a6894de51c315a6bfe6a94e05cf # v6.0.0
+        uses: docker/metadata-action@80c7e94dd9b9319bd5eb7a0e0fe9291e23a2a2e9 # v6.1.0
         with:
           images: |
             ${{ secrets.DOCKER_HUB_USERNAME }}/s-pdf
@@ -178,7 +178,7 @@ jobs:
 
       - name: Generate tags for latest-fat
         id: meta-fat
-        uses: docker/metadata-action@030e881283bb7a6894de51c315a6bfe6a94e05cf # v6.0.0
+        uses: docker/metadata-action@80c7e94dd9b9319bd5eb7a0e0fe9291e23a2a2e9 # v6.1.0
         if: env.RUN_MAIN_APP == 'true' && github.ref != 'refs/heads/main' && github.ref != 'refs/heads/testMain'
         with:
           images: |
@@ -222,7 +222,7 @@ jobs:
 
       - name: Generate tags for ultra-lite
         id: meta-lite
-        uses: docker/metadata-action@030e881283bb7a6894de51c315a6bfe6a94e05cf # v6.0.0
+        uses: docker/metadata-action@80c7e94dd9b9319bd5eb7a0e0fe9291e23a2a2e9 # v6.1.0
         if: env.RUN_MAIN_APP == 'true' && github.ref != 'refs/heads/main' && github.ref != 'refs/heads/testMain'
         with:
           images: |

.github/workflows/stale.yml

@@ -22,7 +22,7 @@ jobs:
           egress-policy: audit
 
       - name: 30 days stale issues
-        uses: actions/stale@b5d41d4e1d5dceea10e7104786b73624c18a190f # v10.2.0
+        uses: actions/stale@eb5cf3af3ac0a1aa4c9c45633dd1ae542a27a899 # v10.3.0
         with:
           repo-token: ${{ secrets.GITHUB_TOKEN }}
           days-before-stale: 30

.github/workflows/swagger.yml

@@ -48,7 +48,7 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Generate Swagger documentation
         run: ./gradlew :stirling-pdf:generateOpenApiDocs
@@ -63,7 +63,7 @@ jobs:
           SWAGGERHUB_USER: "Frooodle"
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Get version number
         id: versionNumber
         run: echo "versionNumber=$(./gradlew printVersion --quiet | tail -1)" >> $GITHUB_OUTPUT

.github/workflows/sync_files_v2.yml

@@ -62,7 +62,7 @@ jobs:
 
       - name: Sync translation TOML files
         run: |
-          python .github/scripts/check_language_toml.py --reference-file "frontend/editor/public/locales/en-GB/translation.toml" --branch main
+          python .github/scripts/check_language_toml.py --reference-file "frontend/editor/public/locales/en-US/translation.toml" --branch main
 
       - name: pre-commit run
         run: |
@@ -100,7 +100,7 @@ jobs:
             This Pull Request was automatically generated to synchronize updates to translation files and documentation. Below are the details of the changes made:
 
             #### **1. Synchronization of Translation Files**
-            - Updated translation files (`frontend/editor/public/locales/*/translation.toml`) to reflect changes in the reference file `en-GB/translation.toml`.
+            - Updated translation files (`frontend/editor/public/locales/*/translation.toml`) to reflect changes in the reference file `en-US/translation.toml`.
             - Ensured consistency and synchronization across all supported language files.
             - Highlighted any missing or incomplete translations.
             - **Format**: TOML

.github/workflows/tauri-build.yml

@@ -136,10 +136,10 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Setup Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
 
       - name: Build universal macOS JRE
         if: matrix.platform == 'macos-15'
@@ -157,6 +157,9 @@ jobs:
           JPDFIUM_PLATFORMS: ${{ matrix.jpdfium_platforms }}
         run: task desktop:prepare
 
+      - name: Run Tauri/Cargo tests
+        run: task desktop:test
+
       # DigiCert KeyLocker Setup (Cloud HSM)
       - name: Setup DigiCert KeyLocker
         id: digicert-setup
@@ -268,10 +271,6 @@ jobs:
           echo "APPLE_SIGNING_IDENTITY=$CERT_ID" >> $GITHUB_ENV
           echo "Certificate imported successfully."
 
-      - name: Sign JPDFium dylibs inside bootJar (macOS only)
-        if: matrix.platform == 'macos-15' && env.APPLE_CERTIFICATE != ''
-        run: bash frontend/scripts/sign-jpdfium-dylibs-in-bootjar.sh
-
       - name: Check DMG creation dependencies (macOS only)
         if: matrix.platform == 'macos-15'
         run: |
@@ -417,20 +416,22 @@ jobs:
       - name: Rename artifacts
         shell: bash
         run: |
-          mkdir -p ./dist
+          # Absolute dist path so the cd below can't break the copy targets.
+          DIST="$GITHUB_WORKSPACE/dist"
+          mkdir -p "$DIST"
           cd ./frontend/editor/src-tauri/target
 
           # Find and rename artifacts based on platform
           if [ "${{ matrix.platform }}" = "windows-latest" ]; then
             # Only ship the MSI installer. The loose exe and WiX toolset exes
             # are not the user-facing installer - the MSI contains the signed inner exe.
-            find . -name "*.msi" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.msi" \;
+            find . -name "*.msi" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.msi" \;
           elif [ "${{ matrix.platform }}" = "macos-15" ]; then
-            find . -name "*.dmg" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.dmg" \;
+            find . -name "*.dmg" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.dmg" \;
           else
-            find . -name "*.deb" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.deb" \;
-            find . -name "*.rpm" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.rpm" \;
-            find . -name "*.AppImage" -exec cp {} "../../../dist/Stirling-PDF-${{ matrix.name }}.AppImage" \;
+            find . -name "*.deb" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.deb" \;
+            find . -name "*.rpm" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.rpm" \;
+            find . -name "*.AppImage" -exec cp {} "$DIST/Stirling-PDF-${{ matrix.name }}.AppImage" \;
           fi
 
       # Verify the MSI AND the inner exe extracted from it are signed.

.github/workflows/test-build-docker.yml

@@ -106,11 +106,11 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
           cache-disabled: true
 
       - name: Install Task
-        uses: go-task/setup-task@3be4020d41929789a01026e0e427a4321ce0ad44 # v2.0.0
+        uses: go-task/setup-task@01a4adf9db2d14c1de7a560f09170b6e0df736aa # v2.1.0
       - name: Build application
         run: task backend:build
         env:
@@ -157,7 +157,7 @@ jobs:
 
       - name: Build ${{ matrix.docker-rev }} (Depot)
         if: env.USE_DEPOT == 'true'
-        uses: depot/build-push-action@5f3b3c2e5a00f0093de47f657aeaefcedff27d18 # v1.16.0
+        uses: depot/build-push-action@98e78adca7817480b8185f474a400b451d74e287 # v1.16.0
         with:
           project: ${{ vars.DEPOT_PROJECT_ID }}
           context: .
@@ -230,7 +230,7 @@ jobs:
 
       - name: Build docker/unoserver/Dockerfile (Depot)
         if: env.USE_DEPOT == 'true'
-        uses: depot/build-push-action@5f3b3c2e5a00f0093de47f657aeaefcedff27d18 # v1.16.0
+        uses: depot/build-push-action@98e78adca7817480b8185f474a400b451d74e287 # v1.16.0
         with:
           project: ${{ vars.DEPOT_PROJECT_ID }}
           context: .

.github/workflows/testdriver.yml

@@ -51,7 +51,7 @@ jobs:
       - name: Setup Gradle
         uses: gradle/actions/setup-gradle@50e97c2cd7a37755bbfafc9c5b7cafaece252f6e # v6.1.0
         with:
-          gradle-version: 9.3.1
+          gradle-version: 9.5.1
 
       - name: Build with Gradle
         run: ./gradlew build
@@ -83,7 +83,7 @@ jobs:
 
       - name: Build and push test image (Depot)
         if: env.USE_DEPOT == 'true'
-        uses: depot/build-push-action@5f3b3c2e5a00f0093de47f657aeaefcedff27d18 # v1.16.0
+        uses: depot/build-push-action@98e78adca7817480b8185f474a400b451d74e287 # v1.16.0
         with:
           project: ${{ vars.DEPOT_PROJECT_ID }}
           context: .
@@ -129,7 +129,7 @@ jobs:
               environment:
                 DISABLE_ADDITIONAL_FEATURES: "true"
                 SECURITY_ENABLELOGIN: "false"
-                SYSTEM_DEFAULTLOCALE: en-GB
+                SYSTEM_DEFAULTLOCALE: en-US
                 UI_APPNAME: "Stirling-PDF Test"
                 UI_HOMEDESCRIPTION: "Test Deployment"
                 UI_APPNAMENAVBAR: "Test"

.gitignore

@@ -23,6 +23,10 @@ customFiles/
 configs/
 watchedFolders/
 clientWebUI/
+# Scratch dir used by local fixture-regeneration runs (see
+# app/proprietary/src/test/resources/db-migration-fixtures/README.md).
+# Holds downloaded JARs and disposable workdirs. Never committed.
+.alpha-local/
 !cucumber/
 !cucumber/exampleFiles/
 !cucumber/exampleFiles/example_html.zip
@@ -53,6 +57,8 @@ app/core/src/main/resources/static/robots.txt
 app/core/src/main/resources/static/pdfium/
 app/core/src/main/resources/static/pdfjs/
 app/core/src/main/resources/static/vendor/
+app/core/src/main/resources/static/**/*.gz
+app/core/src/main/resources/static/**/*.br
 # Note: Keep backend-managed files like fonts/, css/, js/, pdfjs/, etc.
 
 # Gradle
@@ -174,7 +180,6 @@ venv.bak/
 
 # Env files (secrets / local overrides). Subproject .gitignore files whitelist any committed defaults.
 .env*
-!.env.saas.example
 
 # VS Code
 /.vscode/**/*
@@ -209,7 +214,7 @@ out/
 *.asc
 
 # Allow test fixture certificates (synthetic, no real credentials)
-!frontend/src/core/tests/test-fixtures/certs/**
+!frontend/editor/src/core/tests/test-fixtures/certs/**
 
 # SSH Keys
 *.pub
@@ -251,7 +256,7 @@ node_modules/
 *compact*.json
 test_batch.json
 *.backup.*.json
-frontend/public/locales/*/translation.backup*.json
+frontend/editor/public/locales/*/translation.backup*.json
 
 # Development/build artifacts
 .gradle-cache/
@@ -274,3 +279,6 @@ docs/type3/signatures/
 # Playwright MCP screenshots / traces
 .playwright-mcp/
 *.playwright-mcp.png
+
+# Local screenshot artifacts from *-screenshots.spec.ts
+frontend/editor/screenshots/

.gitleaksignore

@@ -0,0 +1,14 @@
+# PostHog project-level key — phc_ prefix keys are public/client-side by design
+# (PostHog client-side tracking embeds them in the browser bundle). Committed
+# intentionally in #6150 so engine/.env has a working default, with real
+# credentials overridden via engine/.env.local.
+engine/.env:generic-api-key:41
+
+# MCP test fixtures / harness - no real secrets:
+# - test-only API key constant in an integration test
+# - JDBC URL + throwaway Keycloak creds in the local test compose
+# - placeholder / shell-variable Bearer headers in curl-based validation scripts
+app/proprietary/src/test/java/stirling/software/proprietary/mcp/security/McpApiKeyIntegrationTest.java:generic-api-key:40
+testing/compose/docker-compose-keycloak-mcp.yml:generic-api-key:25
+testing/compose/validate-mcp-apikey.sh:curl-auth-header:73
+testing/compose/validate-mcp-test.sh:curl-auth-header:92

.taskfiles/backend.yml

@@ -18,16 +18,26 @@ version: '3'
 tasks:
   dev:
     desc: "Start backend dev server"
+    cmds:
+      - task: dev:proprietary
+        vars:
+          PORT: '{{.PORT}}'
+          AIENGINE_URL: '{{.AIENGINE_URL}}'
+          AIENGINE_TIMEOUTSECONDS: '{{.AIENGINE_TIMEOUTSECONDS}}'
+
+  dev:proprietary:
+    desc: "Start backend dev server in proprietary mode"
     ignore_error: true
     vars:
       PORT: '{{.PORT | default "8080"}}'
       AIENGINE_URL: '{{.AIENGINE_URL | default ""}}'
+      AIENGINE_TIMEOUTSECONDS: '{{.AIENGINE_TIMEOUTSECONDS | default "120"}}'
     env:
       SERVER_PORT: '{{.PORT}}'
     cmds:
-      - cmd: '{{if .AIENGINE_URL}}AIENGINE_URL={{.AIENGINE_URL}} AIENGINE_ENABLED=true {{end}}cmd /c ".\gradlew.bat :stirling-pdf:bootRun"'
+      - cmd: '{{if .AIENGINE_URL}}AIENGINE_URL={{.AIENGINE_URL}} AIENGINE_ENABLED=true AIENGINE_TIMEOUTSECONDS={{.AIENGINE_TIMEOUTSECONDS}} {{end}}cmd /c ".\gradlew.bat :stirling-pdf:bootRun"'
         platforms: [windows]
-      - cmd: '{{if .AIENGINE_URL}}AIENGINE_URL={{.AIENGINE_URL}} AIENGINE_ENABLED=true {{end}}./gradlew :stirling-pdf:bootRun'
+      - cmd: '{{if .AIENGINE_URL}}AIENGINE_URL={{.AIENGINE_URL}} AIENGINE_ENABLED=true AIENGINE_TIMEOUTSECONDS={{.AIENGINE_TIMEOUTSECONDS}} {{end}}./gradlew :stirling-pdf:bootRun'
         platforms: [linux, darwin]
 
   dev:bundled:
@@ -40,18 +50,23 @@ tasks:
         platforms: [linux, darwin]
 
   dev:saas:
-    desc: "Start backend in SaaS flavor against Supabase (loads .env.saas.local)"
+    desc: "Start backend in SaaS flavor against Supabase"
     # `dotenv:` reads from the root Taskfile's directory (".") because this
-    # subtaskfile is included with `dir: .`. Drop the file at the repo root.
-    dotenv: ['.env.saas.local']
+    # subtaskfile is included with `dir: .`.
+    dotenv: ['app/.env.saas.local', 'app/.env.saas']
     ignore_error: true
     vars:
       PORT: '{{.PORT | default "8080"}}'
       # Override to "" to run the pure `saas` profile against your own SAAS_DB_*.
       PROFILES: '{{.PROFILES | default "dev"}}'
+      AIENGINE_URL: '{{.AIENGINE_URL | default ""}}'
+      AIENGINE_TIMEOUTSECONDS: '{{.AIENGINE_TIMEOUTSECONDS | default "120"}}'
     env:
       SERVER_PORT: '{{.PORT}}'
       STIRLING_FLAVOR: saas
+      AIENGINE_URL: '{{.AIENGINE_URL}}'
+      AIENGINE_ENABLED: '{{if .AIENGINE_URL}}true{{else}}false{{end}}'
+      AIENGINE_TIMEOUTSECONDS: '{{.AIENGINE_TIMEOUTSECONDS}}'
     cmds:
       - cmd: cmd /c ".\gradlew.bat :stirling-pdf:bootRun {{if .PROFILES}}--args=\"--spring.profiles.include={{.PROFILES}}\"{{end}}"
         platforms: [windows]

.taskfiles/desktop.yml

@@ -1,7 +1,9 @@
 version: '3'
 
 vars:
-  JLINK_MODULES: "java.base,java.compiler,java.desktop,java.instrument,java.logging,java.management,java.naming,java.net.http,java.prefs,java.rmi,java.scripting,java.security.jgss,java.security.sasl,java.sql,java.transaction.xa,java.xml,java.xml.crypto,jdk.crypto.ec,jdk.crypto.cryptoki,jdk.unsupported"
+  # jdk.dynalink is required by VeraPDF (PDF/A validation); without it the bundled JRE throws
+  # NoClassDefFoundError: jdk/dynalink/Namespace at runtime in get-info-on-pdf and verify-pdf
+  JLINK_MODULES: "java.base,java.compiler,java.desktop,java.instrument,java.logging,java.management,java.naming,java.net.http,java.prefs,java.rmi,java.scripting,java.security.jgss,java.security.sasl,java.sql,java.transaction.xa,java.xml,java.xml.crypto,jdk.crypto.ec,jdk.crypto.cryptoki,jdk.unsupported,jdk.dynalink"
 
   # Override via JPDFIUM_PLATFORMS env (csv of platform keys, or 'all').
   JPDFIUM_PLATFORMS:
@@ -62,21 +64,28 @@ tasks:
     deps: [prepare]
     dir: editor
     cmds:
-      - npx tauri build --bundles app
+      - npx tauri build --bundles app --config '{"bundle":{"createUpdaterArtifacts":false}}'
 
   build:dev:windows:
     desc: "Build Tauri desktop NSIS installer (Windows)"
     deps: [prepare]
     dir: editor
     cmds:
-      - npx tauri build --bundles nsis
+      - npx tauri build --bundles nsis --config '{"bundle":{"createUpdaterArtifacts":false}}'
 
   build:dev:linux:
     desc: "Build Tauri desktop AppImage (Linux)"
     deps: [prepare]
     dir: editor
     cmds:
-      - npx tauri build --bundles appimage
+      - npx tauri build --bundles appimage --config '{"bundle":{"createUpdaterArtifacts":false}}'
+
+  test:
+    desc: "Run Tauri/Cargo tests"
+    deps: [prepare]
+    dir: editor/src-tauri
+    cmds:
+      - cargo test
 
   clean:
     desc: "Clean Tauri/Cargo build artifacts"
@@ -118,16 +127,38 @@ tasks:
     cmds:
       - rm -rf runtime/jre
       - mkdir -p runtime
-      - >-
-        jlink
-        --add-modules {{.JLINK_MODULES}}
-        --strip-debug
-        --compress=zip-6
-        --no-header-files
-        --no-man-pages
-        --output runtime/jre
+      - |
+        JLINK_COMPRESS="$(jlink --help 2>&1 | grep -q 'zip-\[0-9\]' && echo zip-6 || echo 2)"
+        jlink \
+          --add-modules {{.JLINK_MODULES}} \
+          --strip-debug \
+          --compress="$JLINK_COMPRESS" \
+          --no-header-files \
+          --no-man-pages \
+          --output runtime/jre
+      # jlink emits its files mode 444 (read-only). Tauri's build-script
+      # resource copier preserves source permissions when staging
+      # `runtime/jre/**/*` into `target/<profile>/runtime/jre/...`, so the
+      # staged copies are read-only too. On any subsequent incremental
+      # build the copier tries to overwrite them and fails with a bare
+      # `Permission denied (os error 13)` (Rust's io::Error Display drops
+      # the path, so the failure is opaque). Make the source writable here
+      # so the staged destinations are writable and can be overwritten.
+      #
+      # Trade-off: this task runs for both `task desktop:dev` and
+      # `task desktop:build`, so production bundles also ship mode-644
+      # JRE files instead of 444. Functionally harmless on POSIX (the
+      # `other` bit is `r--` either way, and on macOS code signing is the
+      # real integrity check) and on Windows the DOS read-only attribute
+      # isn't load-bearing for the bundled JDK. If we ever need strict
+      # 444 in production, split the chmod into a dev-only step and have
+      # `desktop:build` run `jlink:clean` first to force a fresh build.
+      - cmd: chmod -R u+w runtime/jre
+        platforms: [linux, darwin]
+      - cmd: powershell -NoProfile -Command "Get-ChildItem -Recurse runtime/jre | ForEach-Object { $_.IsReadOnly = $false }"
+        platforms: [windows]
     status:
-      - test -d editor/src-tauri/runtime/jre
+      - test -f runtime/jre/release
 
   jlink:clean:
     desc: "Remove JLink runtime and bundled JARs"

.taskfiles/docker.yml

@@ -20,6 +20,11 @@ tasks:
     cmds:
       - docker build -t stirling-pdf-ultra-lite -f {{.EMBEDDED_DIR}}/Dockerfile.ultra-lite .
 
+  build:backend:
+    desc: "Build backend-only Docker image (no embedded frontend)"
+    cmds:
+      - docker build -t stirling-pdf-backend -f docker/backend/Dockerfile .
+
   build:frontend:
     desc: "Build frontend-only Docker image"
     cmds:

.taskfiles/e2e.yml

@@ -34,6 +34,9 @@ tasks:
     ignore_error: true
     vars:
       BASE_DIR: '{{.ROOT_DIR}}/.test-state/playwright'
+    # COVERAGE=1 in the calling environment attaches the JaCoCo agent to
+    # the bootRun JVM and writes to BASE_DIR/jacoco.exec on shutdown.
+    # Off by default to keep local dev runs uninstrumented; CI flips it.
     env:
       STIRLING_BASE_PATH: '{{.BASE_DIR}}'
       # Suppress the analytics opt-in modal that fires on first admin login.
@@ -58,12 +61,19 @@ tasks:
           set -e
           rm -rf "{{.BASE_DIR}}"
           mkdir -p "{{.BASE_DIR}}"
+          GRADLE_ARGS=":stirling-pdf:bootRun"
+          if [ -n "${COVERAGE:-}" ]; then
+            # copyJacocoAgent is wired as a dependency of bootRun when
+            # -PjacocoAgent=true, so we do not need to invoke it separately.
+            GRADLE_ARGS="$GRADLE_ARGS -PjacocoAgent=true -PjacocoExec={{.BASE_DIR}}/jacoco.exec"
+            echo "JaCoCo coverage enabled, writing to {{.BASE_DIR}}/jacoco.exec"
+          fi
           # Background gradle and record its PID so the runner can clean up
           # the exact process tree (wrapper + forked Spring Boot JVM) without
           # resorting to fuzzy `pkill -f` patterns. `wait` keeps this script
           # alive for the lifetime of gradle so Task'"'"'s parallel deps stay
           # synchronised.
-          bash gradlew :stirling-pdf:bootRun > "{{.BASE_DIR}}/backend.log" 2>&1 &
+          bash gradlew $GRADLE_ARGS > "{{.BASE_DIR}}/backend.log" 2>&1 &
           GRADLE_PID=$!
           echo $GRADLE_PID > "{{.BASE_DIR}}/backend.pid"
           wait $GRADLE_PID
@@ -202,3 +212,55 @@ tasks:
     desc: "Stop the SAML keycloak test environment"
     cmds:
       - docker compose -f testing/compose/docker-compose-keycloak-saml.yml down -v
+
+  mcp:up:
+    desc: "Start the MCP keycloak test environment (Stirling as OAuth resource server)"
+    summary: |
+      Brings up Keycloak (OAuth authorization server) + Stirling configured as an
+      MCP resource server, then you can exercise /mcp with real Keycloak tokens.
+      Set LICENSE_KEY=<KEY> to skip the interactive license prompt:
+        task e2e:mcp:up LICENSE_KEY=abc123
+      Pass extra flags via -- :
+        task e2e:mcp:up -- --validate --nobuild
+    ignore_error: true
+    cmds:
+      - bash testing/compose/start-mcp-test.sh {{if .LICENSE_KEY}}--license-key "{{.LICENSE_KEY}}"{{end}} {{.CLI_ARGS}}
+
+  mcp:manual:
+    desc: "Start the MCP keycloak test env in manual mode (prints URLs + a live token for your client)"
+    summary: |
+      Brings the stack up and prints copy-paste URLs/commands plus a freshly minted
+      access token so you can drive your own MCP client (Inspector, curl, ...).
+        task e2e:mcp:manual LICENSE_KEY=<your-license-key>
+      Add --nobuild if the images are already built:
+        task e2e:mcp:manual LICENSE_KEY=<your-license-key> -- --nobuild
+    ignore_error: true
+    cmds:
+      - bash testing/compose/start-mcp-test.sh --manual {{if .LICENSE_KEY}}--license-key "{{.LICENSE_KEY}}"{{end}} {{.CLI_ARGS}}
+
+  mcp:apikey:
+    desc: "Start the MCP test env in API-KEY manual mode (no OAuth/IdP): mints a key + prints client settings"
+    summary: |
+      Brings Stirling up in apikey auth mode and prints copy-paste client settings with a freshly
+      minted X-API-KEY - ideal for clients whose OAuth layer can't reach localhost.
+        task e2e:mcp:apikey LICENSE_KEY=<your-license-key>
+      Add --nobuild if images are already built:
+        task e2e:mcp:apikey LICENSE_KEY=<your-license-key> -- --nobuild
+    ignore_error: true
+    cmds:
+      - bash testing/compose/start-mcp-test.sh --apikey {{if .LICENSE_KEY}}--license-key "{{.LICENSE_KEY}}"{{end}} {{.CLI_ARGS}}
+
+  mcp:validate:
+    desc: "Validate the running MCP keycloak test environment end-to-end (oauth mode + real MCP SDK client)"
+    cmds:
+      - bash testing/compose/validate-mcp-test.sh
+
+  mcp:validate-apikey:
+    desc: "Validate the MCP server in API-KEY auth mode (mints a key + real MCP SDK client), then restore oauth"
+    cmds:
+      - bash testing/compose/validate-mcp-apikey.sh
+
+  mcp:down:
+    desc: "Stop the MCP keycloak test environment"
+    cmds:
+      - docker compose -f testing/compose/docker-compose-keycloak-mcp.yml down -v

.taskfiles/engine.yml

@@ -33,7 +33,7 @@ tasks:
     env:
       PYTHONUNBUFFERED: "1"
     cmds:
-      - uv run uvicorn stirling.api.app:app --host 0.0.0.0 --port {{.PORT}}
+      - uv run uvicorn stirling.api.app:app --host 0.0.0.0 --port {{.PORT}} --workers "${STIRLING_ENGINE_WORKERS:-4}"
 
   dev:
     desc: "Start engine dev server with hot reload"

.taskfiles/frontend.yml

@@ -112,6 +112,12 @@ tasks:
       - task: dev:_run
         vars: { MODE: prototypes, PORT: '{{.PORT}}', BACKEND_URL: '{{.BACKEND_URL}}', OPEN: '{{.OPEN}}' }
 
+  dev:portal:
+    desc: "Start developer portal dev server"
+    deps: [install]
+    cmds:
+      - npx vite portal --port {{.PORT | default "5173"}}{{if .OPEN}} --open{{end}}
+
   # ============================================================
   # Build
   # ============================================================
@@ -156,6 +162,24 @@ tasks:
     cmds:
       - npx vite build editor --mode prototypes
 
+  build:portal:
+    desc: "Build developer portal"
+    deps: [install]
+    cmds:
+      - npx vite build portal
+
+  storybook:
+    desc: "Start Storybook dev server"
+    deps: [install]
+    cmds:
+      - npx storybook dev -p 6006 {{.CLI_ARGS}}
+
+  storybook:build:
+    desc: "Build static Storybook"
+    deps: [install]
+    cmds:
+      - npx storybook build {{.CLI_ARGS}}
+
   # ============================================================
   # Code quality
   # ============================================================
@@ -163,9 +187,23 @@ tasks:
   lint:
     desc: "Run linting"
     deps: [install]
+    cmds:
+      - task: lint:eslint
+      - task: lint:dpdm
+
+  lint:eslint:
+    desc: "Run ESLint linting"
+    deps: [install]
     cmds:
       - npx eslint --max-warnings=0
-      - npx dpdm editor/src --circular --no-warning --no-tree --exit-code circular:1
+
+  lint:dpdm:
+    desc: "Run circular import linting"
+    deps: [install]
+    cmds:
+      # Globs so dpdm walks the whole tree. dpdm expands the braces itself, so this is
+      # shell-agnostic. Covers editor, portal, and the shared design system.
+      - npx dpdm "editor/src/**/*.{ts,tsx}" "portal/src/**/*.{ts,tsx}" "shared/**/*.{ts,tsx}" --circular --no-warning --no-tree --exit-code circular:1
 
   lint:fix:
     desc: "Auto-fix lint issues"
@@ -236,6 +274,18 @@ tasks:
     cmds:
       - npx tsc --noEmit --project editor/src/prototypes/tsconfig.json
 
+  typecheck:portal:
+    desc: "Typecheck developer portal build variant"
+    deps: [install]
+    cmds:
+      - npx tsc --noEmit --project portal/tsconfig.json
+
+  typecheck:shared:
+    desc: "Typecheck the shared design system"
+    deps: [install]
+    cmds:
+      - npx tsc --noEmit --project shared/tsconfig.json
+
   typecheck:all:
     desc: "Typecheck all build variants"
     cmds:
@@ -245,6 +295,8 @@ tasks:
       - task: typecheck:desktop
       - task: typecheck:scripts
       - task: typecheck:prototypes
+      - task: typecheck:portal
+      - task: typecheck:shared
 
   # ============================================================
   # Quality Gate
@@ -265,7 +317,9 @@ tasks:
       - task: lint
       - task: format:check
       - task: build
+      - task: build:portal
       - task: test
+      - task: storybook:build
 
   # ============================================================
   # Test
@@ -273,21 +327,36 @@ tasks:
 
   test:
     desc: "Run tests"
-    deps: [install]
+    deps: [prepare]
     cmds:
       - npx vitest run --root editor
 
   test:watch:
     desc: "Run tests in watch mode"
-    deps: [install]
+    deps: [prepare]
     cmds:
       - npx vitest --watch --root editor
 
   test:coverage:
-    desc: "Run tests with coverage"
-    deps: [install]
+    desc: "Run tests with coverage (one-shot; CI-friendly)."
+    deps: [prepare]
     cmds:
-      - npx vitest --coverage --root editor
+      # `vitest run` makes this CI-safe (the bare `vitest` form enters watch
+      # mode). Explicit reporter list because v8 + json-summary is what the
+      # coverage-summary.py helper consumes; html/text are kept for humans.
+      #
+      # reportsDirectory is pinned to ./coverage relative to vitest's root
+      # (--root editor), so output lands at frontend/editor/coverage/. The
+      # CI upload step reads from that path. An earlier attempt with
+      # `./editor/coverage` double-nested into frontend/editor/editor/coverage;
+      # pinning future-proofs against vitest changing the default.
+      - >
+        npx vitest run --root editor --coverage
+        --coverage.provider=v8
+        --coverage.reporter=text-summary
+        --coverage.reporter=json-summary
+        --coverage.reporter=html
+        --coverage.reportsDirectory=./coverage
 
   # ============================================================
   # Code Generation

ADDING_TOOLS.md

@@ -200,9 +200,9 @@ const [ToolName] = (props: BaseToolProps) => {
 ```
 
 ## 5. Add Translations
-Update translation files. **Important: Only update `en-GB` files** - other languages are handled separately.
+Update translation files. **Important: Only update `en-US` files** - other languages are handled separately.
 
-**File to update:** `frontend/editor/public/locales/en-GB/translation.toml`
+**File to update:** `frontend/editor/public/locales/en-US/translation.toml`
 
 **Required Translation Keys**:
 ```toml
@@ -251,7 +251,7 @@ Update translation files. **Important: Only update `en-GB` files** - other langu
 ```
 
 **Translation Notes:**
-- **Only update `en-GB/translation.toml`** - other locale files are managed separately
+- **Only update `en-US/translation.toml`** - other locale files are managed separately
 - Use descriptive keys that match your component's `t()` calls
 - Include tooltip translations if you created tooltip hooks
 - Add `options.*` keys if your tool has settings with descriptions

AGENTS.md

@@ -426,7 +426,7 @@ The frontend is organized with a clear separation of concerns:
 
 ## Translation Rules
 
-- **CRITICAL**: Always update translations in `en-GB` only, never `en-US`
+- **CRITICAL**: Always update translations in `en-US` only - all other languages (including `en-GB`) are handled separately
 - Translation files are located in `frontend/editor/public/locales/`
 
 ## Important Notes

DeveloperGuide.md

@@ -52,6 +52,17 @@ This guide focuses on developing for Stirling 2.0, including both the React fron
 - Rust and Cargo (required for Tauri desktop app development)
 - Tauri CLI (install with `cargo install tauri-cli`)
 
+### Optional System Dependencies
+
+These are not required to run the app but enable specific features. The app detects them at startup and disables the relevant features if they are missing.
+
+| Dependency | Feature | Install |
+|---|---|---|
+| LibreOffice | File-to-PDF conversions | `brew install libreoffice` / `apt install libreoffice` |
+| Tesseract | OCR | `brew install tesseract` / `apt install tesseract-ocr` |
+| WeasyPrint | AI document creation | `brew install weasyprint` / `apt install weasyprint` |
+| qpdf | PDF optimisation | `brew install qpdf` / `apt install qpdf` |
+
 ### Setup Steps
 
 1. Clone the repository:
@@ -576,7 +587,7 @@ When adding a new feature or modifying existing ones in Stirling-PDF, you'll nee
 Find the existing `messages.properties` files in the `stirling-pdf/src/main/resources` directory. You'll see files like:
 
 - `messages.properties` (default, usually English)
-- `messages_en_GB.properties`
+- `messages_en_US.properties`
 - `messages_fr_FR.properties`
 - `messages_de_DE.properties`
 - etc.

LICENSE

@@ -10,14 +10,16 @@ if that directory exists, is licensed under the license defined in "app/propriet
 if that directory exists, is licensed under the license defined in "app/saas/LICENSE".
 * All content that resides under the "engine/" directory of this repository,
 if that directory exists, is licensed under the license defined in "engine/LICENSE".
-* All content that resides under the "frontend/src/proprietary/" directory of this repository,
-if that directory exists, is licensed under the license defined in "frontend/src/proprietary/LICENSE".
-* All content that resides under the "frontend/src/desktop/" directory of this repository,
-if that directory exists, is licensed under the license defined in "frontend/src/desktop/LICENSE".
-* All content that resides under the "frontend/src/saas/" directory of this repository,
-if that directory exists, is licensed under the license defined in "frontend/src/saas/LICENSE".
-* All content that resides under the "frontend/src/prototypes/" directory of this repository,
-if that directory exists, is licensed under the license defined in "frontend/src/prototypes/LICENSE".
+* All content that resides under the "frontend/editor/src/proprietary/" directory of this repository,
+if that directory exists, is licensed under the license defined in "frontend/editor/src/proprietary/LICENSE".
+* All content that resides under the "frontend/editor/src/desktop/" directory of this repository,
+if that directory exists, is licensed under the license defined in "frontend/editor/src/desktop/LICENSE".
+* All content that resides under the "frontend/editor/src/saas/" directory of this repository,
+if that directory exists, is licensed under the license defined in "frontend/editor/src/saas/LICENSE".
+* All content that resides under the "frontend/editor/src/prototypes/" directory of this repository,
+if that directory exists, is licensed under the license defined in "frontend/editor/src/prototypes/LICENSE".
+* All content that resides under the "frontend/portal/" directory of this repository,
+if that directory exists, is licensed under the license defined in "frontend/portal/LICENSE".
 * Content outside of the above mentioned directories or restrictions above is
 available under the MIT License as defined below.
 

Taskfile.yml

@@ -58,9 +58,22 @@ tasks:
           BACKEND_URL: 'http://localhost:{{.BACKEND_PORT}}'
           OPEN: "true"
 
+  dev:saas:
+    desc: "Start SaaS backend + frontend concurrently on free ports"
+    cmds:
+      - task: dev:_all
+        vars: { FRONTEND: saas, BACKEND: saas }
+
   dev:all:
     desc: "Start backend + frontend + engine concurrently on free ports"
+    cmds:
+      - task: dev:_all
+
+  dev:_all:
+    internal: true
     vars:
+      FRONTEND: '{{.FRONTEND | default "proprietary"}}'
+      BACKEND: '{{.BACKEND | default "proprietary"}}'
       PORTS:
         sh: '{{if eq OS "windows"}}{{.FIND_FREE_PORT_PS}} 8080 5173 5001{{else}}{{.FIND_FREE_PORT_SH}} 8080 5173 5001{{end}}'
       BACKEND_PORT: '{{index (splitList "\n" .PORTS) 0}}'
@@ -70,11 +83,11 @@ tasks:
       - task: engine:dev
         vars:
           PORT: '{{.ENGINE_PORT}}'
-      - task: backend:dev
+      - task: 'backend:dev:{{.BACKEND}}'
         vars:
           PORT: '{{.BACKEND_PORT}}'
           AIENGINE_URL: 'http://localhost:{{.ENGINE_PORT}}'
-      - task: frontend:dev:prototypes
+      - task: 'frontend:dev:{{.FRONTEND}}'
         vars:
           PORT: '{{.FRONTEND_PORT}}'
           BACKEND_URL: 'http://localhost:{{.BACKEND_PORT}}'

app/.env.saas

@@ -1,20 +1,20 @@
 ###############################################################################
-# Stirling-PDF SaaS local environment template.
+# Stirling-PDF SaaS environment defaults.
 #
-# Copy this file to `.env.saas.local` (gitignored) and fill in real values.
-# Loaded by `task backend:dev:saas` via Taskfile's `dotenv:` directive, then
-# read by Spring Boot's `${...}` placeholders in application-saas.properties
-# and application-dev.properties.
+# This file is committed and provides non-secret defaults loaded by
+# `task backend:dev:saas`. Put real values for secrets (passwords, project
+# refs, edge function secrets) in `.env.saas.local` - any variable set there
+# takes precedence over what's defined here.
 #
-# DO NOT commit `.env.saas.local`. Only `.env.saas.example` is checked in.
+# DO NOT commit `.env.saas.local`. Only `.env.saas` is checked in.
 ###############################################################################
 
 # ---------- Supabase project ----------
 # Project reference (the subdomain part of <ref>.supabase.co). Required.
-# Example dev project:
+# Set in .env.saas.local.
 SAAS_DB_PROJECT_REF=
 
-# Edge function secret used by billing/license rollup calls.
+# Edge function secret used by billing/license rollup calls. Set in .env.saas.local.
 SUPABASE_EDGE_FUNCTION_SECRET=
 
 # ---------- Database (saas profile) ----------
@@ -28,7 +28,7 @@ SAAS_DB_PASSWORD=
 # ---------- Database (dev profile overrides) ----------
 # Used when `--spring.profiles.include=dev` is active. The dev profile
 # defaults the URL/username to the shared dev Supabase project, but the
-# password must still be provided here.
+# password must still be provided in .env.saas.local.
 SAAS_DEV_DB_URL=
 SAAS_DEV_DB_USERNAME=postgres
 SAAS_DEV_DB_PASSWORD=

app/.gitignore

@@ -0,0 +1,3 @@
+# Whitelist committed env defaults. `.env.saas.local` (and any other .env*)
+# stays ignored via the root .gitignore.
+!.env.saas

app/allowed-licenses.json

@@ -44,6 +44,10 @@
       "moduleName": ".*",
       "moduleLicense": "The MIT License"
     },
+    {
+      "moduleName": ".*",
+      "moduleLicense": "MIT-0"
+    },
     {
       "moduleName": "com.github.jai-imageio:jai-imageio-core",
       "moduleLicense": "LICENSE.txt"

app/common/build.gradle

@@ -60,7 +60,7 @@ dependencies {
         exclude group: 'com.google.code.gson', module: 'gson'
     }
 
-    api 'com.stirling:jpdfium:1.0.1'
+    api 'com.stirling:jpdfium:1.0.2'
 
     // -PjpdfiumPlatforms=all|<csv of linux-x64,linux-arm64,darwin-x64,darwin-arm64,windows-x64>
     def jpdfiumPlatformsProp = (project.findProperty('jpdfiumPlatforms') ?: 'all').toString().trim()
@@ -75,9 +75,12 @@ dependencies {
     }
     logger.lifecycle("JPDFium native platforms: ${jpdfiumPlatforms.join(', ')}")
     jpdfiumPlatforms.each { platform ->
-        runtimeOnly "com.stirling:jpdfium-natives-${platform}:1.0.1"
+        runtimeOnly "com.stirling:jpdfium-natives-${platform}:1.0.2"
     }
 
+    // Bucket4j (local in-process token bucket for RateLimitStore default impl)
+    implementation 'com.bucket4j:bucket4j_jdk17-core:8.19.0'
+
     // ArchUnit: enforces module dependency direction (see ArchitectureTest)
     testImplementation 'com.tngtech.archunit:archunit-junit5:1.4.2'
 }

app/common/src/main/java/stirling/software/SPDF/pdf/parser/CompositeTableParser.java

@@ -1,73 +0,0 @@
-package stirling.software.SPDF.pdf.parser;
-
-import static stirling.software.SPDF.pdf.parser.PdfModels.*;
-
-import java.io.IOException;
-import java.util.List;
-
-import org.apache.pdfbox.pdmodel.PDDocument;
-import org.springframework.context.annotation.Primary;
-import org.springframework.stereotype.Service;
-
-import lombok.RequiredArgsConstructor;
-import lombok.extern.slf4j.Slf4j;
-
-/**
- * Chains table parsers in priority order: Tabula lattice → Tabula stream → {@link
- * LineAlignmentTableParser}. The first parser returning a result above {@link
- * #TABULA_CONFIDENCE_THRESHOLD} wins; results from different parsers are never mixed on one page.
- */
-@Service
-@Primary
-@RequiredArgsConstructor
-@Slf4j
-public class CompositeTableParser implements TableParser {
-
-    /** Min Tabula confidence to accept results; below this LineAlignment is tried instead. */
-    static final float TABULA_CONFIDENCE_THRESHOLD = 0.5f;
-
-    private final TabulaTableParser tabulaParser;
-    private final LineAlignmentTableParser lineAlignmentParser;
-
-    @Override
-    public List<TableFragment> parse(PDDocument document, RawPage rawPage) throws IOException {
-        // Step 1: Tabula lattice mode (ruled/bordered tables).
-        List<TableFragment> latticeResults = filterConfident(tabulaParser.parse(document, rawPage));
-        if (!latticeResults.isEmpty()) {
-            log.debug(
-                    "Page {}: using Tabula lattice ({} table(s))",
-                    rawPage.pageNumber(),
-                    latticeResults.size());
-            return latticeResults;
-        }
-
-        // Step 2: Tabula stream mode (borderless/whitespace-delimited tables).
-        // parseStream is not on the TableParser interface — this intentionally couples to the
-        // concrete TabulaTableParser since stream mode is a Tabula-specific concept.
-        List<TableFragment> streamResults =
-                filterConfident(tabulaParser.parseStream(document, rawPage));
-        if (!streamResults.isEmpty()) {
-            log.debug(
-                    "Page {}: using Tabula stream ({} table(s))",
-                    rawPage.pageNumber(),
-                    streamResults.size());
-            return streamResults;
-        }
-
-        // Step 3: Geometry-based line-alignment fallback.
-        List<TableFragment> lineResults = lineAlignmentParser.parse(document, rawPage);
-        if (!lineResults.isEmpty()) {
-            log.debug(
-                    "Page {}: using LineAlignment ({} table(s))",
-                    rawPage.pageNumber(),
-                    lineResults.size());
-            return lineResults;
-        }
-
-        return List.of();
-    }
-
-    private List<TableFragment> filterConfident(List<TableFragment> tables) {
-        return tables.stream().filter(t -> t.confidence() >= TABULA_CONFIDENCE_THRESHOLD).toList();
-    }
-}

app/common/src/main/java/stirling/software/SPDF/pdf/parser/LineAlignmentTableParser.java

@@ -1,528 +0,0 @@
-package stirling.software.SPDF.pdf.parser;
-
-import static stirling.software.SPDF.pdf.parser.PdfModels.*;
-
-import java.io.IOException;
-import java.util.ArrayList;
-import java.util.Arrays;
-import java.util.Collections;
-import java.util.Comparator;
-import java.util.HashMap;
-import java.util.List;
-import java.util.Map;
-import java.util.Optional;
-import java.util.TreeMap;
-import java.util.regex.Pattern;
-
-import org.apache.pdfbox.pdmodel.PDDocument;
-import org.springframework.stereotype.Service;
-
-import lombok.extern.slf4j.Slf4j;
-
-/**
- * Fallback {@link TableParser} for borderless financial tables using text geometry.
- *
- * <p>Identifies "anchor lines" (≥2 numeric tokens), builds a column grid from their right-edge
- * positions, groups vertically proximate anchor lines into table candidates, then scores each group
- * on column consistency and anchor density (confidence ceiling 0.85).
- */
-@Service
-@Slf4j
-public class LineAlignmentTableParser implements TableParser {
-
-    /** Width in points of each column position bucket. */
-    static final float COLUMN_BUCKET_PT = 5f;
-
-    /** Tolerance in buckets when matching a token's right-edge to a confirmed column position. */
-    private static final int COLUMN_MATCH_BUCKETS = 2;
-
-    /** Maximum gap (as a multiple of modal line spacing) before splitting a group. */
-    private static final float MAX_GAP_FACTOR = 2.5f;
-
-    /** Minimum anchor rows (numeric-heavy) to form a valid table. */
-    static final int MIN_TABLE_ROWS = 3;
-
-    /** Minimum confirmed column positions to form a valid table. */
-    static final int MIN_COLUMNS = 2;
-
-    /**
-     * Min fraction of anchor lines a column must appear on to be confirmed (permissive for N/A
-     * rows).
-     */
-    private static final double COLUMN_MIN_FREQUENCY = 0.40;
-
-    /**
-     * Matches financial numeric tokens: integers, decimals, parenthetical negatives, currency,
-     * percent, nil dashes.
-     */
-    private static final Pattern NUMERIC =
-            Pattern.compile("^[\\(\\-\\$£€¥]?\\d[\\d,\\.]*[\\)%]?$|^[-–—]$");
-
-    /**
-     * Lines within this y-distance are merged into one row (restores rows split by LineBuilder's
-     * column-gap logic).
-     */
-    static final float ROW_MERGE_TOLERANCE_PT = 2f;
-
-    // ── public API ───────────────────────────────────────────────────────────────────────────────
-
-    @Override
-    public List<TableFragment> parse(PDDocument document, RawPage rawPage) throws IOException {
-        List<RawLine> lines = rawPage.lines();
-        if (lines.size() < MIN_TABLE_ROWS) return List.of();
-
-        float modalSpacing = computeModalSpacing(lines);
-        List<TokenizedLine> tokenized =
-                mergeCoincidentLines(lines.stream().map(this::tokenize).toList());
-
-        List<TokenizedLine> anchors = tokenized.stream().filter(TokenizedLine::isAnchor).toList();
-
-        if (anchors.size() < MIN_TABLE_ROWS) return List.of();
-
-        List<Float> columnGrid = buildColumnGrid(anchors);
-        if (columnGrid.size() < MIN_COLUMNS) {
-            log.debug(
-                    "Page {}: LineAlignment — fewer than {} confirmed columns, skipping",
-                    rawPage.pageNumber(),
-                    MIN_COLUMNS);
-            return List.of();
-        }
-
-        List<List<TokenizedLine>> groups = groupRows(tokenized, columnGrid, modalSpacing);
-
-        List<TableFragment> results = new ArrayList<>();
-        for (int i = 0; i < groups.size(); i++) {
-            buildFragment(groups.get(i), columnGrid, rawPage.pageNumber(), i)
-                    .ifPresent(results::add);
-        }
-
-        log.debug(
-                "Page {}: LineAlignment detected {} table(s) ({} anchor lines, {} columns)",
-                rawPage.pageNumber(),
-                results.size(),
-                anchors.size(),
-                columnGrid.size());
-        return results;
-    }
-
-    // ── coincident-line merging ──────────────────────────────────────────────────────────────────
-
-    /**
-     * Merges tokenised lines sharing the same y-position into one row, rejoining label/value halves
-     * split by LineBuilder.
-     */
-    List<TokenizedLine> mergeCoincidentLines(List<TokenizedLine> tokenized) {
-        if (tokenized.size() < 2) return tokenized;
-
-        List<TokenizedLine> result = new ArrayList<>();
-        int i = 0;
-
-        while (i < tokenized.size()) {
-            float baseY = tokenized.get(i).line().bounds().y();
-            int j = i + 1;
-            while (j < tokenized.size()
-                    && Math.abs(tokenized.get(j).line().bounds().y() - baseY)
-                            <= ROW_MERGE_TOLERANCE_PT) {
-                j++;
-            }
-
-            if (j == i + 1) {
-                result.add(tokenized.get(i));
-            } else {
-                result.add(mergeGroup(tokenized.subList(i, j)));
-            }
-            i = j;
-        }
-
-        return result;
-    }
-
-    private TokenizedLine mergeGroup(List<TokenizedLine> group) {
-        List<TextFragment> mergedFragments =
-                group.stream()
-                        .flatMap(tl -> tl.line().fragments().stream())
-                        .sorted(Comparator.comparingDouble(f -> f.bounds().x()))
-                        .toList();
-
-        Bounds mergedBounds =
-                group.stream()
-                        .map(tl -> tl.line().bounds())
-                        .reduce(Bounds::merge)
-                        .orElse(group.get(0).line().bounds());
-
-        RawLine mergedLine =
-                new RawLine(
-                        group.get(0).line().lineId(),
-                        mergedFragments,
-                        mergedBounds,
-                        group.get(0).line().pageNumber());
-
-        return tokenize(mergedLine);
-    }
-
-    // ── tokenisation ─────────────────────────────────────────────────────────────────────────────
-
-    /**
-     * Splits fragments into word-level tokens; x-positions are estimated linearly within each
-     * fragment.
-     */
-    TokenizedLine tokenize(RawLine line) {
-        List<LineToken> tokens = new ArrayList<>();
-        for (TextFragment frag : line.fragments()) {
-            tokens.addAll(tokensFromFragment(frag));
-        }
-        List<LineToken> numeric = tokens.stream().filter(LineToken::numeric).toList();
-        return new TokenizedLine(line, tokens, numeric);
-    }
-
-    private List<LineToken> tokensFromFragment(TextFragment frag) {
-        String raw = frag.text();
-        if (raw == null || raw.isBlank()) return List.of();
-
-        float fragX = frag.bounds().x();
-        float fragWidth = frag.bounds().width();
-        int rawLen = raw.length();
-
-        List<LineToken> result = new ArrayList<>();
-        int offset = 0;
-        for (String part : raw.split("\\s+")) {
-            if (part.isEmpty()) {
-                offset++;
-                continue;
-            }
-            int idx = raw.indexOf(part, offset);
-            if (idx < 0) idx = offset;
-
-            float tokenX = rawLen > 0 ? fragX + ((float) idx / rawLen) * fragWidth : fragX;
-            float tokenRight =
-                    rawLen > 0
-                            ? fragX + ((float) (idx + part.length()) / rawLen) * fragWidth
-                            : fragX + fragWidth;
-
-            result.add(new LineToken(part, tokenX, tokenRight, NUMERIC.matcher(part).matches()));
-            offset = idx + part.length();
-        }
-        return result;
-    }
-
-    // ── column grid ──────────────────────────────────────────────────────────────────────────────
-
-    /**
-     * Returns confirmed column right-edge positions — those appearing on ≥ {@value
-     * #COLUMN_MIN_FREQUENCY} × N anchor lines.
-     */
-    private List<Float> buildColumnGrid(List<TokenizedLine> anchors) {
-        // bucket → set of line indices that contributed a numeric token to that bucket
-        Map<Integer, List<Integer>> bucketLines = new HashMap<>();
-        for (int i = 0; i < anchors.size(); i++) {
-            for (LineToken t : anchors.get(i).numeric()) {
-                int bucket = bucket(t.right());
-                bucketLines.computeIfAbsent(bucket, k -> new ArrayList<>()).add(i);
-            }
-        }
-
-        int minHits =
-                Math.max(MIN_TABLE_ROWS, (int) Math.ceil(anchors.size() * COLUMN_MIN_FREQUENCY));
-
-        // Confirmed buckets → average right-edge for that bucket
-        TreeMap<Integer, Float> confirmed = new TreeMap<>();
-        for (Map.Entry<Integer, List<Integer>> entry : bucketLines.entrySet()) {
-            // Count distinct lines
-            long distinctLines = entry.getValue().stream().distinct().count();
-            if (distinctLines >= minHits) {
-                double avg =
-                        entry.getValue().stream()
-                                .distinct() // weight each line equally regardless of token count
-                                .mapToDouble(
-                                        lineIdx ->
-                                                avgRightEdgeForBucket(
-                                                        anchors, lineIdx, entry.getKey()))
-                                .average()
-                                .orElse(entry.getKey() * (double) COLUMN_BUCKET_PT);
-                confirmed.put(entry.getKey(), (float) avg);
-            }
-        }
-
-        return new ArrayList<>(confirmed.values()); // already sorted by bucket (left to right)
-    }
-
-    /**
-     * Returns the average right-edge position of tokens in {@code line} whose bucket matches {@code
-     * targetBucket}, falling back to the bucket's nominal centre when no tokens match.
-     */
-    private double avgRightEdgeForBucket(
-            List<TokenizedLine> anchors, int lineIdx, int targetBucket) {
-        return anchors.get(lineIdx).numeric().stream()
-                .filter(t -> bucket(t.right()) == targetBucket)
-                .mapToDouble(LineToken::right)
-                .average()
-                .orElse(targetBucket * (double) COLUMN_BUCKET_PT);
-    }
-
-    // ── grouping ─────────────────────────────────────────────────────────────────────────────────
-
-    /**
-     * Groups anchor lines into table candidates, including adjacent label rows; a gap &gt;
-     * MAX_GAP_FACTOR × modal spacing splits groups.
-     */
-    private List<List<TokenizedLine>> groupRows(
-            List<TokenizedLine> all, List<Float> columnGrid, float modalSpacing) {
-        float maxGap = modalSpacing > 0 ? modalSpacing * MAX_GAP_FACTOR : 30f;
-
-        List<List<TokenizedLine>> groups = new ArrayList<>();
-        List<TokenizedLine> current = new ArrayList<>();
-
-        for (int i = 0; i < all.size(); i++) {
-            TokenizedLine tl = all.get(i);
-            boolean fits = tl.isAnchor() && matchesGrid(tl, columnGrid);
-
-            if (current.isEmpty()) {
-                if (fits) current.add(tl);
-                continue;
-            }
-
-            float gap =
-                    tl.line().bounds().y()
-                            - current.get(current.size() - 1).line().bounds().bottom();
-
-            if (gap > maxGap) {
-                groups.add(current);
-                current = new ArrayList<>();
-                if (fits) current.add(tl);
-                continue;
-            }
-
-            if (fits) {
-                current.add(tl);
-            } else if (!tl.line().text().isBlank()) {
-                // Include non-anchor lines (labels) only if they have text and are within
-                // proximity.
-                current.add(tl);
-            }
-        }
-
-        if (!current.isEmpty()) groups.add(current);
-
-        return groups.stream().filter(g -> hasEnoughAnchorRows(g, columnGrid)).toList();
-    }
-
-    private boolean hasEnoughAnchorRows(List<TokenizedLine> group, List<Float> columnGrid) {
-        return group.stream().filter(r -> r.isAnchor() && matchesGrid(r, columnGrid)).count()
-                >= MIN_TABLE_ROWS;
-    }
-
-    /** A line "matches" the grid when ≥ 60 % of its numeric tokens land in confirmed columns. */
-    private boolean matchesGrid(TokenizedLine tl, List<Float> columnGrid) {
-        if (tl.numeric().isEmpty()) return false;
-        long matches =
-                tl.numeric().stream()
-                        .filter(t -> nearestColumnIndex(t.right(), columnGrid) >= 0)
-                        .count();
-        return (double) matches / tl.numeric().size() >= 0.60;
-    }
-
-    private boolean hasInconsistentColumnMatch(TokenizedLine tl, List<Float> columnGrid) {
-        if (tl.numeric().isEmpty()) return false;
-        long hits =
-                tl.numeric().stream()
-                        .filter(t -> nearestColumnIndex(t.right(), columnGrid) >= 0)
-                        .count();
-        return (double) hits / tl.numeric().size() < 0.60;
-    }
-
-    // ── fragment assembly ────────────────────────────────────────────────────────────────────────
-
-    private Optional<TableFragment> buildFragment(
-            List<TokenizedLine> group, List<Float> columnGrid, int pageNumber, int tableIndex) {
-
-        long anchorCount =
-                group.stream().filter(r -> r.isAnchor() && matchesGrid(r, columnGrid)).count();
-        if (anchorCount < MIN_TABLE_ROWS) return Optional.empty();
-
-        List<String> warnings = new ArrayList<>();
-        List<List<String>> rawRows = new ArrayList<>();
-        List<TableRow> rows = new ArrayList<>();
-
-        for (int rowIdx = 0; rowIdx < group.size(); rowIdx++) {
-            TokenizedLine tl = group.get(rowIdx);
-            List<String> rawRow = buildRawRow(tl, columnGrid);
-            rawRows.add(Collections.unmodifiableList(rawRow));
-            rows.add(buildTableRow(rowIdx, tl, rawRow, columnGrid));
-        }
-
-        // Column count = 1 label column + confirmed numeric columns
-        int colCount = columnGrid.size() + 1;
-        Bounds bounds = computeGroupBounds(group);
-        float confidence = computeConfidence(group, columnGrid, warnings);
-
-        return Optional.of(
-                new TableFragment(
-                        "tbl-la-p" + pageNumber + "-" + tableIndex,
-                        pageNumber,
-                        bounds,
-                        List.of(),
-                        Collections.unmodifiableList(rows),
-                        Collections.unmodifiableList(rawRows),
-                        colCount,
-                        confidence,
-                        Collections.unmodifiableList(warnings),
-                        null));
-    }
-
-    /**
-     * Builds a raw row as a list of strings: index 0 = label text, indices 1..N = column values.
-     */
-    private List<String> buildRawRow(TokenizedLine tl, List<Float> columnGrid) {
-        String[] cells = new String[columnGrid.size() + 1];
-        Arrays.fill(cells, "");
-
-        // Separate label tokens (those not landing in any confirmed column) from column tokens.
-        List<String> labelParts = new ArrayList<>();
-        for (LineToken token : tl.all()) {
-            int col = nearestColumnIndex(token.right(), columnGrid);
-            if (col >= 0 && token.numeric()) {
-                int cellIdx = col + 1;
-                cells[cellIdx] =
-                        cells[cellIdx].isEmpty()
-                                ? token.text()
-                                : cells[cellIdx] + " " + token.text();
-            } else {
-                labelParts.add(token.text());
-            }
-        }
-        cells[0] = String.join(" ", labelParts).trim();
-        return Arrays.asList(cells);
-    }
-
-    private TableRow buildTableRow(
-            int rowIdx, TokenizedLine tl, List<String> rawRow, List<Float> columnGrid) {
-        List<TableCell> cells = new ArrayList<>(rawRow.size());
-
-        // Label cell: use the line's full bounds as an approximation.
-        cells.add(TableCell.of(0, rawRow.get(0), tl.line().bounds()));
-
-        for (int col = 0; col < columnGrid.size(); col++) {
-            String text = col + 1 < rawRow.size() ? rawRow.get(col + 1) : "";
-            float right = columnGrid.get(col);
-            float left = col > 0 ? columnGrid.get(col - 1) : right - 50f;
-            Bounds cellBounds =
-                    new Bounds(
-                            left,
-                            tl.line().bounds().y(),
-                            right - left,
-                            tl.line().bounds().height());
-            cells.add(TableCell.of(col + 1, text, cellBounds));
-        }
-        return new TableRow(rowIdx, Collections.unmodifiableList(cells));
-    }
-
-    // ── confidence scoring ───────────────────────────────────────────────────────────────────────
-
-    /**
-     * Heuristic score in [0.0, 0.85] (ceiling keeps results below Tabula lattice which starts at
-     * 1.0). Base 0.70; +0.05/col beyond 2 (max +0.10); +0.05 at ≥5 anchors, +0.05 at ≥8; −0.15 if
-     * &gt;30 % of anchors have inconsistent columns; −0.10 if non-anchors outnumber anchors.
-     */
-    private float computeConfidence(
-            List<TokenizedLine> group, List<Float> columnGrid, List<String> warnings) {
-        float score = 0.70f;
-
-        long anchorCount =
-                group.stream().filter(r -> r.isAnchor() && matchesGrid(r, columnGrid)).count();
-        long totalRows = group.size();
-
-        // More columns
-        int extraCols = Math.min(columnGrid.size() - MIN_COLUMNS, 2);
-        score += extraCols * 0.05f;
-
-        // More anchor rows
-        if (anchorCount >= 5) score += 0.05f;
-        if (anchorCount >= 8) score += 0.05f;
-
-        // Inconsistent column matching
-        long inconsistent =
-                group.stream()
-                        .filter(TokenizedLine::isAnchor)
-                        .filter(tl -> hasInconsistentColumnMatch(tl, columnGrid))
-                        .count();
-        if (inconsistent > anchorCount * 0.30) {
-            score -= 0.15f;
-            warnings.add(
-                    "Column match inconsistent on "
-                            + inconsistent
-                            + "/"
-                            + anchorCount
-                            + " anchor rows");
-        }
-
-        // Label-heavy
-        long nonAnchor = totalRows - anchorCount;
-        if (nonAnchor > anchorCount) {
-            score -= 0.10f;
-            warnings.add(
-                    "Non-anchor rows ("
-                            + nonAnchor
-                            + ") outnumber anchor rows ("
-                            + anchorCount
-                            + ")");
-        }
-
-        return Math.max(0f, Math.min(0.85f, score));
-    }
-
-    // ── utility ──────────────────────────────────────────────────────────────────────────────────
-
-    /**
-     * Returns the grid index nearest to {@code rightEdge}, or -1 if none is within {@value
-     * #COLUMN_MATCH_BUCKETS} buckets.
-     */
-    private int nearestColumnIndex(float rightEdge, List<Float> grid) {
-        int nearest = -1;
-        float minDist = COLUMN_MATCH_BUCKETS * COLUMN_BUCKET_PT + 1f;
-        for (int i = 0; i < grid.size(); i++) {
-            float dist = Math.abs(rightEdge - grid.get(i));
-            if (dist < minDist) {
-                minDist = dist;
-                nearest = i;
-            }
-        }
-        return nearest;
-    }
-
-    private Bounds computeGroupBounds(List<TokenizedLine> group) {
-        return group.stream()
-                .map(tl -> tl.line().bounds())
-                .reduce(Bounds::merge)
-                .orElse(new Bounds(0, 0, 0, 0));
-    }
-
-    /** Modal gap between consecutive line edges, used to calibrate the group-split threshold. */
-    private float computeModalSpacing(List<RawLine> lines) {
-        if (lines.size() < 2) return 0f;
-        Map<Float, Long> freq = new HashMap<>();
-        for (int i = 1; i < lines.size(); i++) {
-            float gap = lines.get(i).bounds().y() - lines.get(i - 1).bounds().bottom();
-            if (gap > 0) freq.merge(Math.round(gap / 2f) * 2f, 1L, Long::sum);
-        }
-        return freq.entrySet().stream()
-                .max(Map.Entry.comparingByValue())
-                .map(Map.Entry::getKey)
-                .orElse(0f);
-    }
-
-    private static int bucket(float x) {
-        return Math.round(x / COLUMN_BUCKET_PT);
-    }
-
-    // ── private data types ───────────────────────────────────────────────────────────────────────
-
-    /** A word-level token with an approximate right-edge x-position. */
-    record LineToken(String text, float x, float right, boolean numeric) {}
-
-    /** A {@link RawLine} with tokens pre-computed; an "anchor" has ≥ 2 numeric tokens. */
-    record TokenizedLine(RawLine line, List<LineToken> all, List<LineToken> numeric) {
-        boolean isAnchor() {
-            return numeric.size() >= 2;
-        }
-    }
-}

app/common/src/main/java/stirling/software/SPDF/pdf/parser/LineBuilder.java

@@ -1,139 +0,0 @@
-package stirling.software.SPDF.pdf.parser;
-
-import static stirling.software.SPDF.pdf.parser.PdfModels.*;
-
-import java.util.ArrayList;
-import java.util.Comparator;
-import java.util.List;
-
-import org.springframework.stereotype.Service;
-
-import lombok.extern.slf4j.Slf4j;
-
-/**
- * Groups {@link TextFragment} objects into visual {@link RawLine}s using baseline proximity.
- *
- * <p>Fragments are on the same line when their baselines are within a font-size-derived tolerance.
- * A new line starts whenever the horizontal gap exceeds an adaptive column-gap threshold ({@code
- * max(effectiveWidth * COLUMN_GAP_RATIO, COLUMN_GAP_MIN_PT)}), splitting two-column text.
- */
-@Service
-@Slf4j
-public class LineBuilder {
-
-    /** Baseline tolerance as a fraction of font size; 0.5 keeps mixed-size text on one line. */
-    private static final float BASELINE_TOLERANCE_FACTOR = 0.5f;
-
-    /** Absolute minimum tolerance so tiny font sizes don't collapse multi-line content. */
-    private static final float MIN_BASELINE_TOLERANCE = 2f;
-
-    /**
-     * Column-gap threshold as a fraction of page width; 0.10 clears tab stops but stays below
-     * two-column gutters.
-     */
-    static final float COLUMN_GAP_RATIO = 0.10f;
-
-    /** Floor for the column-gap threshold so narrow pages don't over-split lines. */
-    static final float COLUMN_GAP_MIN_PT = 40f;
-
-    public List<RawLine> build(List<TextFragment> fragments, int pageNumber) {
-        if (fragments.isEmpty()) return List.of();
-
-        float effectiveWidth = inferEffectiveWidth(fragments);
-        float columnGapThreshold = Math.max(effectiveWidth * COLUMN_GAP_RATIO, COLUMN_GAP_MIN_PT);
-        log.debug(
-                "LineBuilder page {}: effectiveWidth={:.1f}pt, columnGapThreshold={:.1f}pt",
-                pageNumber,
-                effectiveWidth,
-                columnGapThreshold);
-
-        // Sort top-to-bottom first, then left-to-right within the same baseline band.
-        List<TextFragment> sorted =
-                fragments.stream()
-                        .sorted(
-                                Comparator.comparingDouble(TextFragment::baseline)
-                                        .thenComparingDouble(f -> f.bounds().x()))
-                        .toList();
-
-        List<List<TextFragment>> groups = groupByBaseline(sorted, columnGapThreshold);
-
-        List<RawLine> lines = new ArrayList<>(groups.size());
-        for (int i = 0; i < groups.size(); i++) {
-            List<TextFragment> group =
-                    groups.get(i).stream()
-                            .sorted(Comparator.comparingDouble(f -> f.bounds().x()))
-                            .toList();
-
-            Bounds lineBounds =
-                    group.stream()
-                            .map(TextFragment::bounds)
-                            .reduce(Bounds::merge)
-                            .orElse(new Bounds(0, 0, 0, 0));
-
-            lines.add(new RawLine("ln-p" + pageNumber + "-" + i, group, lineBounds, pageNumber));
-        }
-        return lines;
-    }
-
-    private List<List<TextFragment>> groupByBaseline(
-            List<TextFragment> sorted, float columnGapThreshold) {
-        List<List<TextFragment>> groups = new ArrayList<>();
-        List<TextFragment> current = new ArrayList<>();
-        float currentBaseline = Float.NaN;
-
-        for (TextFragment fragment : sorted) {
-            if (current.isEmpty()) {
-                current.add(fragment);
-                currentBaseline = fragment.baseline();
-                continue;
-            }
-
-            float maxFontSize =
-                    Math.max(
-                            fragment.fontSize(),
-                            (float)
-                                    current.stream()
-                                            .mapToDouble(TextFragment::fontSize)
-                                            .max()
-                                            .orElse(0));
-            float tolerance =
-                    Math.max(maxFontSize * BASELINE_TOLERANCE_FACTOR, MIN_BASELINE_TOLERANCE);
-
-            boolean sameBaseline = Math.abs(fragment.baseline() - currentBaseline) <= tolerance;
-            boolean columnGap = sameBaseline && hasColumnGap(fragment, current, columnGapThreshold);
-
-            if (sameBaseline && !columnGap) {
-                current.add(fragment);
-                // Anchor to the weighted mean baseline so long lines stay stable.
-                currentBaseline =
-                        (currentBaseline * (current.size() - 1) + fragment.baseline())
-                                / current.size();
-            } else {
-                groups.add(current);
-                current = new ArrayList<>();
-                current.add(fragment);
-                currentBaseline = fragment.baseline();
-            }
-        }
-
-        if (!current.isEmpty()) groups.add(current);
-        return groups;
-    }
-
-    /**
-     * True when the gap from the rightmost fragment in {@code group} to {@code next} exceeds {@code
-     * threshold}.
-     */
-    private static boolean hasColumnGap(
-            TextFragment next, List<TextFragment> group, float threshold) {
-        float lastRight = group.get(group.size() - 1).bounds().right();
-        return next.bounds().x() - lastRight > threshold;
-    }
-
-    /** Infers effective page width from the rightmost fragment right-edge plus a 10 % margin. */
-    private static float inferEffectiveWidth(List<TextFragment> fragments) {
-        double maxRight =
-                fragments.stream().mapToDouble(f -> f.bounds().right()).max().orElse(500.0);
-        return (float) maxRight * 1.10f;
-    }
-}

app/common/src/main/java/stirling/software/SPDF/pdf/parser/PageColumnLayout.java

@@ -0,0 +1,142 @@
+package stirling.software.SPDF.pdf.parser;
+
+import java.util.ArrayList;
+import java.util.List;
+
+/**
+ * Detects whether a page is one- or two-column from per-line bounding boxes, and classifies an
+ * X-span into the column it belongs to. Detection is a midpoint vote at {@code pageWidth / 2}.
+ *
+ * <p>Capped at two columns by design — sufficient for the redaction target set (single-column
+ * documents and IEEE-style two-column papers). 3+ column layouts (newspapers, magazines) and
+ * off-centre gutters (asymmetric two-column) would need a histogram or clustering approach to
+ * detect the actual gutter X. (future work)
+ *
+ * <p>Coordinates are PDFTextStripper screen space (top-left origin, Y increases downward).
+ */
+public final class PageColumnLayout {
+
+    /**
+     * Slack when checking "crosses a gutter" so single-pixel overshoots don't mark a line as
+     * spanning.
+     */
+    public static final float SPAN_SLACK_PT = 2f;
+
+    /**
+     * Slack on each side of the page midpoint inside which a line is considered "spanning"
+     * (covering both columns) rather than belonging to one side.
+     */
+    private static final float MIDPOINT_SLACK_PT = 30f;
+
+    /**
+     * Minimum line width (points) for a line to count toward the two-column tally. Avoids false
+     * positives where right-aligned dates, page numbers, or short "Link" fragments next to a
+     * heading look like a second column when they're really just inline metadata.
+     */
+    private static final float MIN_COLUMN_LINE_WIDTH_PT = 100f;
+
+    /**
+     * Minimum number of clearly leftish AND clearly rightish lines (each of width &ge; {@link
+     * #MIN_COLUMN_LINE_WIDTH_PT}) required to call the page two-column. Anything below this falls
+     * back to single-column.
+     */
+    private static final int MIN_SIDE_LINES = 3;
+
+    private final List<float[]> columns;
+    private final List<float[]> gutters;
+
+    private PageColumnLayout(List<float[]> columns, List<float[]> gutters) {
+        this.columns = columns;
+        this.gutters = gutters;
+    }
+
+    /**
+     * Determines column layout from per-line bounding boxes ({@code [x1, _, x2, _]}). Counts lines
+     * whose X-midpoint sits clearly left of, or clearly right of, the page midpoint (with {@link
+     * #MIDPOINT_SLACK_PT} slack each side). If both sides have at least {@link #MIN_SIDE_LINES}
+     * lines, the page is treated as two-column with the gutter at the page midpoint. Otherwise it's
+     * single-column.
+     *
+     * <p>Cross-column lines must already be split: callers should feed boxes from a line extractor
+     * that splits same-Y glyphs at large X gaps (see {@code AllTextLineExtractor}). Without that
+     * split, IEEE-style aligned-baseline 2-column PDFs produce one wide merged box per row and the
+     * side tallies all end up classified as "spanning", falling to single-column.
+     */
+    public static PageColumnLayout fromLineBoxes(List<float[]> lineBoxes, float pageWidth) {
+        if (lineBoxes == null || lineBoxes.isEmpty()) {
+            return new PageColumnLayout(List.of(new float[] {0f, pageWidth}), List.of());
+        }
+        float pageMid = pageWidth / 2f;
+        int left = 0, right = 0;
+        for (float[] lb : lineBoxes) {
+            if (lb == null || lb.length < 3) continue;
+            float width = lb[2] - lb[0];
+            // Skip narrow lines — dates, page numbers, "Link" labels next to a heading should
+            // not, on their own, make a single-column doc look two-column.
+            if (width < MIN_COLUMN_LINE_WIDTH_PT) continue;
+            float mid = (lb[0] + lb[2]) * 0.5f;
+            if (mid < pageMid - MIDPOINT_SLACK_PT) left++;
+            else if (mid > pageMid + MIDPOINT_SLACK_PT) right++;
+        }
+        if (left < MIN_SIDE_LINES || right < MIN_SIDE_LINES) {
+            return new PageColumnLayout(List.of(new float[] {0f, pageWidth}), List.of());
+        }
+        float gutterL = pageMid - MIDPOINT_SLACK_PT;
+        float gutterR = pageMid + MIDPOINT_SLACK_PT;
+        return new PageColumnLayout(
+                List.of(new float[] {0f, gutterL}, new float[] {gutterR, pageWidth}),
+                List.of(new float[] {gutterL, gutterR}));
+    }
+
+    /** All columns, left-to-right, as {@code [leftX, rightX]} pairs. Never empty. */
+    public List<float[]> columns() {
+        return columns;
+    }
+
+    /** Gutters between columns, left-to-right, as {@code [leftX, rightX]} pairs. */
+    public List<float[]> gutters() {
+        return gutters;
+    }
+
+    public int columnCount() {
+        return columns.size();
+    }
+
+    /**
+     * Returns the column index containing the X-midpoint of {@code [x1, x2]}, falling back to the
+     * closest column if the midpoint sits inside a gutter.
+     */
+    public int columnOf(float x1, float x2) {
+        float mid = (x1 + x2) * 0.5f;
+        int best = 0;
+        float bestDist = Float.MAX_VALUE;
+        for (int i = 0; i < columns.size(); i++) {
+            float[] c = columns.get(i);
+            if (mid >= c[0] && mid <= c[1]) return i;
+            float dist = mid < c[0] ? c[0] - mid : mid - c[1];
+            if (dist < bestDist) {
+                bestDist = dist;
+                best = i;
+            }
+        }
+        return best;
+    }
+
+    /**
+     * Returns every column index whose X-range overlaps {@code [x1, x2]} with at least {@link
+     * #SPAN_SLACK_PT} of intrusion. A normal in-column line returns one index; a line crossing a
+     * gutter returns two or more.
+     */
+    public int[] columnsCrossing(float x1, float x2) {
+        List<Integer> hits = new ArrayList<>();
+        for (int i = 0; i < columns.size(); i++) {
+            float[] c = columns.get(i);
+            float overlap = Math.min(x2, c[1]) - Math.max(x1, c[0]);
+            if (overlap > SPAN_SLACK_PT) hits.add(i);
+        }
+        if (hits.isEmpty()) hits.add(columnOf(x1, x2));
+        int[] out = new int[hits.size()];
+        for (int i = 0; i < hits.size(); i++) out[i] = hits.get(i);
+        return out;
+    }
+}

app/common/src/main/java/stirling/software/SPDF/pdf/parser/PageImageLocator.java

@@ -0,0 +1,133 @@
+package stirling.software.SPDF.pdf.parser;
+
+import java.awt.geom.Point2D;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+
+import org.apache.pdfbox.contentstream.PDFGraphicsStreamEngine;
+import org.apache.pdfbox.cos.COSName;
+import org.apache.pdfbox.pdmodel.PDPage;
+import org.apache.pdfbox.pdmodel.graphics.image.PDImage;
+import org.apache.pdfbox.util.Matrix;
+
+/**
+ * PDFGraphicsStreamEngine that intercepts {@code drawImage} calls and records each image's bounding
+ * box in PDF user-space (origin bottom-left, Y up) by transforming the unit square through the
+ * current transformation matrix (CTM).
+ *
+ * <p>Usage:
+ *
+ * <pre>{@code
+ * PageImageLocator locator = new PageImageLocator(page, pageIndex);
+ * locator.processPage(page);
+ * List<ImageBox> boxes = locator.getImageBoxes();
+ * }</pre>
+ *
+ * <p>Each {@link ImageBox} carries the 0-based page index and the axis-aligned bounding box {@code
+ * (x1, y1, x2, y2)} in PDF user-space coordinates.
+ */
+public final class PageImageLocator extends PDFGraphicsStreamEngine {
+
+    /**
+     * Bounding box of a raster or vector image found on a PDF page.
+     *
+     * @param pageIndex 0-based page index
+     * @param x1 left edge in PDF user-space (origin bottom-left)
+     * @param y1 bottom edge in PDF user-space
+     * @param x2 right edge
+     * @param y2 top edge
+     */
+    public record ImageBox(int pageIndex, float x1, float y1, float x2, float y2) {}
+
+    private final int pageIndex;
+    private final List<ImageBox> imageBoxes = new ArrayList<>();
+    private final Point2D.Float currentPoint = new Point2D.Float();
+
+    /**
+     * @param page the PDPage to process
+     * @param pageIndex 0-based index of this page in the document (stored on each returned {@link
+     *     ImageBox})
+     */
+    public PageImageLocator(PDPage page, int pageIndex) {
+        super(page);
+        this.pageIndex = pageIndex;
+    }
+
+    /** Returns all image bounding boxes collected during {@link #processPage}. */
+    public List<ImageBox> getImageBoxes() {
+        return imageBoxes;
+    }
+
+    @Override
+    public void drawImage(PDImage pdImage) throws IOException {
+        Matrix ctm = getGraphicsState().getCurrentTransformationMatrix();
+        // An image occupies the unit square (0,0)→(1,1) in image space.
+        // Transform all four corners through the CTM to get the page-space bounding box.
+        float a = ctm.getScaleX();
+        float b = ctm.getShearY();
+        float c = ctm.getShearX();
+        float d = ctm.getScaleY();
+        float e = ctm.getTranslateX();
+        float f = ctm.getTranslateY();
+        float[] xs = {e, a + e, c + e, a + c + e};
+        float[] ys = {f, b + f, d + f, b + d + f};
+        float x1 = Float.MAX_VALUE, y1 = Float.MAX_VALUE;
+        float x2 = -Float.MAX_VALUE, y2 = -Float.MAX_VALUE;
+        for (float x : xs) {
+            x1 = Math.min(x1, x);
+            x2 = Math.max(x2, x);
+        }
+        for (float y : ys) {
+            y1 = Math.min(y1, y);
+            y2 = Math.max(y2, y);
+        }
+        imageBoxes.add(new ImageBox(pageIndex, x1, y1, x2, y2));
+    }
+
+    // ---------- required abstract methods (no-op for path operations) ----------
+
+    @Override
+    public void appendRectangle(Point2D p0, Point2D p1, Point2D p2, Point2D p3) {}
+
+    @Override
+    public void clip(int windingRule) {}
+
+    @Override
+    public void moveTo(float x, float y) {
+        currentPoint.setLocation(x, y);
+    }
+
+    @Override
+    public void lineTo(float x, float y) {
+        currentPoint.setLocation(x, y);
+    }
+
+    @Override
+    public void curveTo(float x1, float y1, float x2, float y2, float x3, float y3) {
+        currentPoint.setLocation(x3, y3);
+    }
+
+    @Override
+    public Point2D getCurrentPoint() {
+        return currentPoint;
+    }
+
+    @Override
+    public void closePath() {}
+
+    @Override
+    public void endPath() {}
+
+    @Override
+    public void strokePath() {}
+
+    @Override
+    public void fillPath(int windingRule) {}
+
+    @Override
+    public void fillAndStrokePath(int windingRule) {}
+
+    @Override
+    public void shadingFill(COSName shadingName) {}
+}

app/common/src/main/java/stirling/software/SPDF/pdf/parser/PdfIngester.java

@@ -1,79 +0,0 @@
-package stirling.software.SPDF.pdf.parser;
-
-import static stirling.software.SPDF.pdf.parser.PdfModels.*;
-
-import java.io.IOException;
-import java.util.ArrayList;
-import java.util.List;
-
-import org.apache.pdfbox.pdmodel.PDDocument;
-import org.apache.pdfbox.pdmodel.PDPage;
-import org.apache.pdfbox.pdmodel.common.PDRectangle;
-import org.springframework.stereotype.Service;
-
-import lombok.RequiredArgsConstructor;
-import lombok.extern.slf4j.Slf4j;
-
-/**
- * Runs the per-page ingestion pipeline: {@link WordExtractingStripper} → {@link LineBuilder} →
- * {@link TableParser}, producing a {@link PdfModels.ParsedPage} per page. The caller owns the
- * {@link PDDocument} lifecycle.
- */
-@Service
-@RequiredArgsConstructor
-@Slf4j
-public class PdfIngester {
-
-    private final LineBuilder lineBuilder;
-    private final TableParser tableParser;
-
-    public List<ParsedPage> parse(PDDocument document) throws IOException {
-        return parse(document, document.getNumberOfPages());
-    }
-
-    public List<ParsedPage> parse(PDDocument document, int maxPages) throws IOException {
-        int pageCount = Math.min(document.getNumberOfPages(), maxPages);
-        List<ParsedPage> pages = new ArrayList<>(pageCount);
-        long fragmentsMs = 0;
-        long tablesMs = 0;
-        long t0 = System.currentTimeMillis();
-
-        for (int p = 1; p <= pageCount; p++) {
-            long ft = System.currentTimeMillis();
-            List<TextFragment> fragments = extractFragments(document, p);
-            fragmentsMs += System.currentTimeMillis() - ft;
-
-            PDPage page = document.getPage(p - 1);
-            PDRectangle mediaBox = page.getMediaBox();
-            List<RawLine> lines = lineBuilder.build(fragments, p);
-            RawPage rawPage = new RawPage(p, mediaBox.getWidth(), mediaBox.getHeight(), lines);
-
-            long tt = System.currentTimeMillis();
-            List<TableFragment> tables = tableParser.parse(document, rawPage);
-            tablesMs += System.currentTimeMillis() - tt;
-
-            log.debug(
-                    "Page {}: {} fragments → {} lines, {} table(s)",
-                    p,
-                    fragments.size(),
-                    lines.size(),
-                    tables.size());
-            pages.add(new ParsedPage(p, mediaBox.getWidth(), mediaBox.getHeight(), tables, lines));
-        }
-
-        log.info(
-                "[timing] parse pages={} total={}ms fragments={}ms tables={}ms",
-                pageCount,
-                System.currentTimeMillis() - t0,
-                fragmentsMs,
-                tablesMs);
-        return pages;
-    }
-
-    private List<TextFragment> extractFragments(PDDocument document, int pageNumber)
-            throws IOException {
-        WordExtractingStripper stripper = new WordExtractingStripper(pageNumber);
-        stripper.getText(document);
-        return stripper.getFragments();
-    }
-}

app/common/src/main/java/stirling/software/SPDF/pdf/parser/WordExtractingStripper.java

@@ -1,113 +0,0 @@
-package stirling.software.SPDF.pdf.parser;
-
-import static stirling.software.SPDF.pdf.parser.PdfModels.*;
-
-import java.io.IOException;
-import java.util.ArrayList;
-import java.util.Collections;
-import java.util.List;
-
-import org.apache.pdfbox.pdmodel.PDPage;
-import org.apache.pdfbox.pdmodel.font.PDFont;
-import org.apache.pdfbox.text.PDFTextStripper;
-import org.apache.pdfbox.text.TextPosition;
-
-/**
- * Extends {@link PDFTextStripper} to capture per-fragment geometry and font metadata.
- *
- * <p>Overrides {@link #writeString} to split each content-stream string into word-level {@link
- * TextFragment}s with bounding boxes, baseline, font name, and bold flag. Coordinates are in
- * PDFTextStripper space: (0,0) top-left, Y increases downward, {@code getY()} is the baseline.
- */
-class WordExtractingStripper extends PDFTextStripper {
-
-    private final int targetPage;
-    private final List<TextFragment> fragments = new ArrayList<>();
-    private int fragmentIndex = 0;
-
-    WordExtractingStripper(int pageNumber) throws IOException {
-        this.targetPage = pageNumber;
-        setStartPage(pageNumber);
-        setEndPage(pageNumber);
-        setSortByPosition(true);
-    }
-
-    @Override
-    protected void startPage(PDPage page) throws IOException {
-        super.startPage(page);
-        fragments.clear();
-        fragmentIndex = 0;
-    }
-
-    @Override
-    protected void writeString(String text, List<TextPosition> textPositions) throws IOException {
-        if (text == null || text.isBlank()) return;
-
-        // Fast path: no whitespace → emit one fragment (most financial PDFs have each
-        // number as its own string operation, so this is the common case).
-        if (text.indexOf(' ') < 0) {
-            emitFragment(text, textPositions);
-            return;
-        }
-
-        // Per-word splitting requires 1:1 text-char to TextPosition correspondence.
-        // Fall back to one fragment when sizes differ (ligatures, encoding edge cases).
-        if (textPositions.size() != text.length()) {
-            emitFragment(text, textPositions);
-            return;
-        }
-
-        // Emit one TextFragment per whitespace-delimited word with accurate per-word bounds.
-        int start = 0;
-        for (int i = 0; i <= text.length(); i++) {
-            if (i == text.length() || text.charAt(i) == ' ') {
-                if (start < i) {
-                    emitFragment(text.substring(start, i), textPositions.subList(start, i));
-                }
-                start = i + 1;
-            }
-        }
-    }
-
-    private void emitFragment(String text, List<TextPosition> positions) {
-        if (positions.isEmpty()) return;
-
-        float minX = Float.MAX_VALUE;
-        float minY = Float.MAX_VALUE;
-        float maxRight = -Float.MAX_VALUE;
-        float maxBaseline = -Float.MAX_VALUE;
-        TextPosition first = null;
-
-        for (TextPosition tp : positions) {
-            if (tp == null) continue;
-            if (first == null) first = tp;
-
-            float x = tp.getX();
-            // getY() is the baseline; top of character = getY() - getHeight().
-            float top = tp.getY() - tp.getHeight();
-            float right = x + tp.getWidth();
-            float baseline = tp.getY();
-
-            minX = Math.min(minX, x);
-            minY = Math.min(minY, top);
-            maxRight = Math.max(maxRight, right);
-            maxBaseline = Math.max(maxBaseline, baseline);
-        }
-
-        if (first == null) return;
-
-        PDFont font = first.getFont();
-        String fontName = font != null ? font.getName() : "";
-        boolean bold = fontName != null && fontName.toLowerCase().contains("bold");
-        // getHeight() gives the rendered glyph height, which is the most reliable visual size.
-        float fontSize = first.getHeight();
-
-        Bounds bounds = new Bounds(minX, minY, maxRight - minX, maxBaseline - minY);
-        String id = "tf-p" + targetPage + "-" + fragmentIndex++;
-        fragments.add(new TextFragment(id, text, bounds, maxBaseline, fontSize, fontName, bold));
-    }
-
-    List<TextFragment> getFragments() {
-        return Collections.unmodifiableList(fragments);
-    }
-}

app/common/src/main/java/stirling/software/common/annotations/AutoJobPostMapping.java

@@ -77,6 +77,10 @@ public @interface AutoJobPostMapping {
     /**
      * Relative resource weight (1-100). See {@link
      * stirling.software.common.enumeration.ResourceWeight} for the standard tiers.
+     *
+     * <p>The default is a sentinel ({@link Integer#MIN_VALUE}); {@code
+     * AutoJobPostMappingWeightTest} fails the build if any endpoint leaves it unset. Runtime
+     * readers clamp the value into {@code [1, 100]}.
      */
-    int resourceWeight() default 1;
+    int resourceWeight() default Integer.MIN_VALUE;
 }

app/common/src/main/java/stirling/software/common/cluster/ClusterBackplane.java

@@ -0,0 +1,24 @@
+package stirling.software.common.cluster;
+
+/** Health and identity facade for the active cluster backplane. */
+public interface ClusterBackplane {
+
+    /** Returns {@code true} when the backplane is reachable; used for health endpoints. */
+    boolean isHealthy();
+
+    /** Returns {@code "inprocess"} or {@code "valkey"}. */
+    String backplaneType();
+
+    /** Returns this JVM's stable node id (matches {@code Cluster.resolvedNodeId()}). */
+    String localNodeId();
+
+    /**
+     * Whether this JVM should run the local {@link
+     * stirling.software.common.service.TaskManager#cleanupOldJobs()} loop. Distributed backplanes
+     * own job expiry via their own TTL, so they should override this to return {@code false}.
+     * Defaults to {@code true} so in-process behavior is preserved without an explicit override.
+     */
+    default boolean shouldRunLocalCleanup() {
+        return true;
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/ClusterConfig.java

@@ -0,0 +1,63 @@
+package stirling.software.common.cluster;
+
+import org.springframework.context.annotation.Configuration;
+
+import jakarta.annotation.PostConstruct;
+
+import lombok.RequiredArgsConstructor;
+import lombok.extern.slf4j.Slf4j;
+
+import stirling.software.common.model.ApplicationProperties;
+import stirling.software.common.model.ApplicationProperties.Cluster;
+
+/**
+ * Validates that cluster mode is internally consistent.
+ *
+ * <p>Cluster settings are bound on the central {@link ApplicationProperties} under {@code
+ * cluster.*}; this class reads {@link ApplicationProperties#getCluster()} and runs guards in {@link
+ * PostConstruct}. When {@code cluster.enabled=false} (the default) all checks are skipped so a
+ * single-instance install needs no new config.
+ */
+@Slf4j
+@Configuration
+@RequiredArgsConstructor
+public class ClusterConfig {
+
+    private final ApplicationProperties applicationProperties;
+
+    @PostConstruct
+    void validate() {
+        Cluster cluster = applicationProperties.getCluster();
+        if (!cluster.isEnabled()) {
+            return;
+        }
+        String backplane = cluster.getBackplane();
+        if ("valkey".equalsIgnoreCase(backplane)) {
+            String url = cluster.getValkey() == null ? null : cluster.getValkey().getUrl();
+            if (url == null || url.isBlank()) {
+                throw new IllegalStateException(
+                        "cluster.enabled=true with backplane=valkey requires"
+                                + " cluster.valkey.url to be set (e.g."
+                                + " redis://valkey:6379).");
+            }
+        } else if ("inprocess".equalsIgnoreCase(backplane)) {
+            // enabled+inprocess only coordinates the local JVM; cross-node lookups will 410.
+            log.warn(
+                    "cluster.enabled=true with backplane=inprocess - only the local"
+                            + " JVM is coordinated. Cross-node lookups and the file proxy will fail."
+                            + " Use backplane=valkey for real multi-node deployments.");
+        } else {
+            // Fail fast on typos like "valky" so Spring doesn't later report a cryptic
+            // "no ClusterBackplane bean" - the operator-facing error names the bad value.
+            throw new IllegalStateException(
+                    "cluster.enabled=true with unknown backplane '"
+                            + backplane
+                            + "'. Valid values: inprocess | valkey.");
+        }
+        log.info(
+                "Cluster mode enabled (backplane={}, nodeRole={}, nodeId={}).",
+                backplane,
+                cluster.resolvedRole(),
+                cluster.resolvedNodeId());
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/ClusterNode.java

@@ -0,0 +1,12 @@
+package stirling.software.common.cluster;
+
+import java.time.Instant;
+
+/**
+ * Snapshot of a peer node as recorded in the {@link InstanceRegistry}.
+ *
+ * @param internalAddress {@code host:port} the node listens on for {@code /internal/cluster/**}
+ * @param role one of {@code WEB}, {@code WORKER}, {@code BOTH}
+ */
+public record ClusterNode(
+        String nodeId, String internalAddress, Instant lastHeartbeat, String role) {}

app/common/src/main/java/stirling/software/common/cluster/DistributedLock.java

@@ -0,0 +1,21 @@
+package stirling.software.common.cluster;
+
+import java.time.Duration;
+import java.util.Optional;
+
+/** Cluster-wide mutual exclusion primitive; non-reentrant by contract. */
+public interface DistributedLock {
+
+    Optional<LockHandle> tryAcquire(String lockKey, Duration leaseTime);
+
+    interface LockHandle extends AutoCloseable {
+        void release();
+
+        boolean renew(Duration leaseTime);
+
+        @Override
+        default void close() {
+            release();
+        }
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/FileStore.java

@@ -0,0 +1,69 @@
+package stirling.software.common.cluster;
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.file.Files;
+import java.nio.file.Path;
+
+/** Low-level storage seam for result/job files. */
+public interface FileStore {
+
+    /** Stored file record. */
+    record Stored(String fileId, long size) {}
+
+    /**
+     * Store the given stream and return a generated file id and total bytes written. {@code owner}
+     * may be null to indicate the file has no associated user (anonymous / desktop / async job with
+     * no propagated security context); a non-null value is persisted alongside the data so {@link
+     * #getOwner(String)} can return it later for authorization checks.
+     */
+    Stored store(InputStream in, String originalName, String owner) throws IOException;
+
+    /** Store with no owner. Equivalent to {@link #store(InputStream, String, String)} with null. */
+    default Stored store(InputStream in, String originalName) throws IOException {
+        return store(in, originalName, null);
+    }
+
+    /**
+     * Store the file at {@code source} and return a generated file id and total bytes written.
+     *
+     * <p>Default implementation opens {@code source} as a stream and delegates to {@link
+     * #store(InputStream, String, String)}. Local-disk implementations should override to use a
+     * direct file-to-file copy ({@code Files.copy(source, dest)} can use {@code sendfile(2)} on
+     * Linux), which avoids the two-memory-copy hit of streaming a disk-backed upload through the
+     * JVM heap.
+     */
+    default Stored store(Path source, String originalName, String owner) throws IOException {
+        try (InputStream in = Files.newInputStream(source)) {
+            return store(in, originalName, owner);
+        }
+    }
+
+    /** Store with no owner. Equivalent to {@link #store(Path, String, String)} with null. */
+    default Stored store(Path source, String originalName) throws IOException {
+        return store(source, originalName, null);
+    }
+
+    /** Open the stored file for streaming reads. Caller closes. */
+    InputStream retrieve(String fileId) throws IOException;
+
+    /** Load the stored file into a byte array. */
+    byte[] retrieveBytes(String fileId) throws IOException;
+
+    /** Size of the stored file in bytes. */
+    long size(String fileId) throws IOException;
+
+    /** Delete the stored file. Returns true if a file was removed. */
+    boolean delete(String fileId);
+
+    /** Whether the file id exists in the store. */
+    boolean exists(String fileId);
+
+    /**
+     * Returns the owner identifier recorded at store time, or {@code null} if the file does not
+     * exist or was stored without an owner. Implementations must not throw when the file is missing
+     * or when the owner record is absent; they should return null so callers can treat "no owner"
+     * as a non-authoritative case.
+     */
+    String getOwner(String fileId) throws IOException;
+}

app/common/src/main/java/stirling/software/common/cluster/InstanceRegistry.java

@@ -0,0 +1,18 @@
+package stirling.software.common.cluster;
+
+import java.time.Duration;
+import java.util.Collection;
+import java.util.Optional;
+
+/** Maps {@code nodeId} to its internal cluster address, with TTL'd heartbeats. */
+public interface InstanceRegistry {
+
+    /** Register or refresh this node. Idempotent so a wiped backplane self-heals on next tick. */
+    void register(ClusterNode node, Duration heartbeatTtl);
+
+    Optional<ClusterNode> lookup(String nodeId);
+
+    Collection<ClusterNode> activeNodes();
+
+    void deregister(String nodeId);
+}

app/common/src/main/java/stirling/software/common/cluster/JobStore.java

@@ -0,0 +1,24 @@
+package stirling.software.common.cluster;
+
+import java.time.Duration;
+import java.util.Collection;
+import java.util.Optional;
+
+/** Cluster-visible storage for job status and result metadata, with TTL'd entries. */
+public interface JobStore {
+
+    /** Persist or overwrite a job entry. {@code ttl} sets the lifetime of the entry. */
+    void put(JobStoreEntry entry, Duration ttl);
+
+    Optional<JobStoreEntry> get(String jobId);
+
+    void delete(String jobId);
+
+    boolean exists(String jobId);
+
+    /** Reverse lookup: which job owns this result file id? */
+    Optional<String> findJobIdByFileId(String fileId);
+
+    /** Snapshot of every active entry. Used by admin/stats endpoints; may be O(n). */
+    Collection<JobStoreEntry> all();
+}

app/common/src/main/java/stirling/software/common/cluster/JobStoreEntry.java

@@ -0,0 +1,30 @@
+package stirling.software.common.cluster;
+
+import java.time.Instant;
+import java.util.List;
+import java.util.Map;
+
+/**
+ * Cluster-visible projection of a job's status and result metadata, as persisted in {@link
+ * JobStore}.
+ *
+ * @param owningNodeId the node id that originally executed the job
+ */
+public record JobStoreEntry(
+        String jobId,
+        JobState state,
+        String owningNodeId,
+        Instant createdAt,
+        Instant completedAt,
+        String error,
+        List<String> fileIds,
+        Map<String, String> resultMeta) {
+
+    /** Lifecycle states for a job as observed by the cluster. */
+    public enum JobState {
+        PENDING,
+        RUNNING,
+        COMPLETE,
+        FAILED
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/KeyValueCache.java

@@ -0,0 +1,16 @@
+package stirling.software.common.cluster;
+
+import java.time.Duration;
+import java.util.Optional;
+
+/** Short-TTL namespaced key/value cache backed by the cluster backplane. */
+public interface KeyValueCache {
+
+    void put(String namespace, String key, String value, Duration ttl);
+
+    Optional<String> get(String namespace, String key);
+
+    void evict(String namespace, String key);
+
+    void evictNamespace(String namespace);
+}

app/common/src/main/java/stirling/software/common/cluster/RateLimitStore.java

@@ -0,0 +1,24 @@
+package stirling.software.common.cluster;
+
+import java.time.Duration;
+
+/**
+ * Token-bucket rate limiting backed by the cluster backplane.
+ *
+ * <p>In-process implementations enforce a per-JVM limit; distributed implementations enforce a
+ * single global limit across every node. Both use a Bucket4j greedy-refill token bucket so the
+ * semantics match across single-node and cluster deployments.
+ */
+public interface RateLimitStore {
+
+    /**
+     * Attempt to consume one token from the bucket identified by {@code bucketKey}.
+     *
+     * @param bucketKey opaque key identifying the bucket (e.g. {@code api:user:123})
+     * @param capacity bucket capacity
+     * @param refillPeriod time window over which {@code capacity} tokens refill
+     */
+    RateLimitDecision tryConsume(String bucketKey, long capacity, Duration refillPeriod);
+
+    record RateLimitDecision(boolean allowed, long remainingTokens, long nanosToWaitForRefill) {}
+}

app/common/src/main/java/stirling/software/common/cluster/StickyMissRecorder.java

@@ -0,0 +1,7 @@
+package stirling.software.common.cluster;
+
+/** Records one increment per sticky-session miss (a 410 Gone for a job owned by another node). */
+@FunctionalInterface
+public interface StickyMissRecorder {
+    void recordStickyMiss();
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/InProcessClusterBackplane.java

@@ -0,0 +1,31 @@
+package stirling.software.common.cluster.inprocess;
+
+import lombok.extern.slf4j.Slf4j;
+
+import stirling.software.common.cluster.ClusterBackplane;
+import stirling.software.common.model.ApplicationProperties;
+
+@Slf4j
+public class InProcessClusterBackplane implements ClusterBackplane {
+
+    private final ApplicationProperties applicationProperties;
+
+    public InProcessClusterBackplane(ApplicationProperties applicationProperties) {
+        this.applicationProperties = applicationProperties;
+    }
+
+    @Override
+    public boolean isHealthy() {
+        return true;
+    }
+
+    @Override
+    public String backplaneType() {
+        return "inprocess";
+    }
+
+    @Override
+    public String localNodeId() {
+        return applicationProperties.getCluster().resolvedNodeId();
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/InProcessClusterConfiguration.java

@@ -0,0 +1,65 @@
+package stirling.software.common.cluster.inprocess;
+
+import org.springframework.boot.autoconfigure.condition.ConditionalOnExpression;
+import org.springframework.boot.autoconfigure.condition.ConditionalOnMissingBean;
+import org.springframework.context.annotation.Bean;
+import org.springframework.context.annotation.Configuration;
+
+import lombok.extern.slf4j.Slf4j;
+
+import stirling.software.common.cluster.ClusterBackplane;
+import stirling.software.common.cluster.DistributedLock;
+import stirling.software.common.cluster.InstanceRegistry;
+import stirling.software.common.cluster.JobStore;
+import stirling.software.common.cluster.KeyValueCache;
+import stirling.software.common.cluster.RateLimitStore;
+import stirling.software.common.model.ApplicationProperties;
+
+/**
+ * Default cluster backplane wiring: every interface gets an {@code InProcess*} bean. Active when
+ * cluster mode is off or {@code cluster.backplane=inprocess}.
+ */
+@Slf4j
+@Configuration
+@ConditionalOnExpression(
+        "!${cluster.enabled:false} ||"
+                + " '${cluster.backplane:inprocess}'.equalsIgnoreCase('inprocess')")
+public class InProcessClusterConfiguration {
+
+    @Bean
+    @ConditionalOnMissingBean
+    public ClusterBackplane clusterBackplane(ApplicationProperties applicationProperties) {
+        log.info("Cluster backplane: in-process (single node)");
+        return new InProcessClusterBackplane(applicationProperties);
+    }
+
+    @Bean
+    @ConditionalOnMissingBean
+    public JobStore jobStore() {
+        return new InProcessJobStore();
+    }
+
+    @Bean
+    @ConditionalOnMissingBean
+    public RateLimitStore rateLimitStore() {
+        return new InProcessRateLimitStore();
+    }
+
+    @Bean
+    @ConditionalOnMissingBean
+    public DistributedLock distributedLock() {
+        return new InProcessDistributedLock();
+    }
+
+    @Bean
+    @ConditionalOnMissingBean
+    public KeyValueCache keyValueCache() {
+        return new InProcessKeyValueCache();
+    }
+
+    @Bean
+    @ConditionalOnMissingBean
+    public InstanceRegistry instanceRegistry() {
+        return new InProcessInstanceRegistry();
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/InProcessDistributedLock.java

@@ -0,0 +1,128 @@
+package stirling.software.common.cluster.inprocess;
+
+import java.time.Duration;
+import java.util.Optional;
+import java.util.concurrent.ConcurrentHashMap;
+import java.util.concurrent.atomic.AtomicLong;
+
+import stirling.software.common.cluster.DistributedLock;
+
+/**
+ * In-process {@link DistributedLock}, non-reentrant per the interface contract, with lease-expiry
+ * semantics that mirror a SET-NX-EX style distributed backend.
+ *
+ * <p>Each lock state carries a per-acquire {@code ownerToken} and an {@code expiryNanos}; another
+ * caller can take over once the lease has elapsed even if the original holder never called {@link
+ * LockHandle#release()}. This matters mostly for parity with the Valkey-backed implementation
+ * (Redis {@code SETEX} auto-expires the key); within a single JVM a crashed holder takes its lock
+ * state with it, but tests and code that rely on the {@code leaseTime} parameter still need it to
+ * be honored.
+ *
+ * <p>Expiry is lazy: an expired lock state lingers in the map until the next acquire attempt for
+ * the same key replaces it. Per-key cleanup also happens on explicit {@link LockHandle#release()},
+ * so a balanced acquire/release workload keeps the map size bounded.
+ */
+public class InProcessDistributedLock implements DistributedLock {
+
+    private final ConcurrentHashMap<String, LockState> locks = new ConcurrentHashMap<>();
+    private final AtomicLong tokenSeq = new AtomicLong();
+
+    /**
+     * Lease state for a single acquired lock. {@code ownerToken} prevents a former holder from
+     * releasing or renewing a lock now owned by someone else after lease expiry; {@code
+     * expiryNanos} is read/written only inside {@link ConcurrentHashMap#compute} so the bin lock
+     * provides the necessary happens-before guarantee.
+     */
+    private static final class LockState {
+        final long ownerToken;
+        long expiryNanos;
+
+        LockState(long ownerToken, long expiryNanos) {
+            this.ownerToken = ownerToken;
+            this.expiryNanos = expiryNanos;
+        }
+    }
+
+    @Override
+    public Optional<LockHandle> tryAcquire(String lockKey, Duration leaseTime) {
+        long token = tokenSeq.incrementAndGet();
+        long nowNanos = System.nanoTime();
+        long expiryNanos = nowNanos + leaseTime.toNanos();
+        boolean[] acquired = {false};
+        locks.compute(
+                lockKey,
+                (k, existing) -> {
+                    if (existing == null || existing.expiryNanos - nowNanos <= 0L) {
+                        // No lock, or the previous lease has expired - we take it. Subtraction
+                        // form avoids the long-overflow trap that would bite a naive
+                        // expiryNanos <= nowNanos comparison around System.nanoTime() rollover.
+                        acquired[0] = true;
+                        return new LockState(token, expiryNanos);
+                    }
+                    return existing;
+                });
+        if (!acquired[0]) {
+            return Optional.empty();
+        }
+        return Optional.of(new InProcessHandle(lockKey, token));
+    }
+
+    private void releaseInternal(String lockKey, long token) {
+        locks.compute(
+                lockKey,
+                (k, existing) -> {
+                    if (existing == null || existing.ownerToken != token) {
+                        // Already removed, expired-and-replaced, or never ours.
+                        return existing;
+                    }
+                    return null;
+                });
+    }
+
+    private boolean renewInternal(String lockKey, long token, Duration leaseTime) {
+        long nowNanos = System.nanoTime();
+        boolean[] renewed = {false};
+        locks.compute(
+                lockKey,
+                (k, existing) -> {
+                    if (existing == null
+                            || existing.ownerToken != token
+                            || existing.expiryNanos - nowNanos <= 0L) {
+                        // Lock is gone or expired; renewal is a no-op so the caller can detect it.
+                        return existing;
+                    }
+                    existing.expiryNanos = nowNanos + leaseTime.toNanos();
+                    renewed[0] = true;
+                    return existing;
+                });
+        return renewed[0];
+    }
+
+    private final class InProcessHandle implements LockHandle {
+        private final String lockKey;
+        private final long token;
+        private boolean released;
+
+        InProcessHandle(String lockKey, long token) {
+            this.lockKey = lockKey;
+            this.token = token;
+        }
+
+        @Override
+        public synchronized void release() {
+            if (released) {
+                return;
+            }
+            released = true;
+            releaseInternal(lockKey, token);
+        }
+
+        @Override
+        public synchronized boolean renew(Duration leaseTime) {
+            if (released) {
+                return false;
+            }
+            return renewInternal(lockKey, token, leaseTime);
+        }
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/InProcessInstanceRegistry.java

@@ -0,0 +1,42 @@
+package stirling.software.common.cluster.inprocess;
+
+import java.time.Duration;
+import java.util.Collection;
+import java.util.Collections;
+import java.util.Optional;
+import java.util.concurrent.atomic.AtomicReference;
+
+import stirling.software.common.cluster.ClusterNode;
+import stirling.software.common.cluster.InstanceRegistry;
+
+public class InProcessInstanceRegistry implements InstanceRegistry {
+
+    private final AtomicReference<ClusterNode> self = new AtomicReference<>();
+
+    @Override
+    public void register(ClusterNode node, Duration heartbeatTtl) {
+        self.set(node);
+    }
+
+    @Override
+    public Optional<ClusterNode> lookup(String nodeId) {
+        ClusterNode current = self.get();
+        return current != null && current.nodeId().equals(nodeId)
+                ? Optional.of(current)
+                : Optional.empty();
+    }
+
+    @Override
+    public Collection<ClusterNode> activeNodes() {
+        ClusterNode current = self.get();
+        return current == null ? Collections.emptyList() : Collections.singletonList(current);
+    }
+
+    @Override
+    public void deregister(String nodeId) {
+        ClusterNode current = self.get();
+        if (current != null && current.nodeId().equals(nodeId)) {
+            self.set(null);
+        }
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/InProcessJobStore.java

@@ -0,0 +1,96 @@
+package stirling.software.common.cluster.inprocess;
+
+import java.time.Duration;
+import java.time.Instant;
+import java.util.ArrayList;
+import java.util.Collection;
+import java.util.List;
+import java.util.Map;
+import java.util.Optional;
+import java.util.concurrent.ConcurrentHashMap;
+
+import lombok.extern.slf4j.Slf4j;
+
+import stirling.software.common.cluster.JobStore;
+import stirling.software.common.cluster.JobStoreEntry;
+
+@Slf4j
+public class InProcessJobStore implements JobStore {
+
+    private final ConcurrentHashMap<String, Holder> entries = new ConcurrentHashMap<>();
+
+    @Override
+    public void put(JobStoreEntry entry, Duration ttl) {
+        Instant expiry = ttl == null ? Instant.MAX : Instant.now().plus(ttl);
+        entries.put(entry.jobId(), new Holder(entry, expiry));
+    }
+
+    @Override
+    public Optional<JobStoreEntry> get(String jobId) {
+        Holder h = entries.get(jobId);
+        if (h == null) {
+            return Optional.empty();
+        }
+        if (h.isExpired()) {
+            entries.remove(jobId, h);
+            return Optional.empty();
+        }
+        return Optional.of(h.entry);
+    }
+
+    @Override
+    public void delete(String jobId) {
+        entries.remove(jobId);
+    }
+
+    @Override
+    public boolean exists(String jobId) {
+        return get(jobId).isPresent();
+    }
+
+    @Override
+    public Optional<String> findJobIdByFileId(String fileId) {
+        for (Map.Entry<String, Holder> e : entries.entrySet()) {
+            Holder h = e.getValue();
+            if (h.isExpired()) {
+                continue;
+            }
+            List<String> fileIds = h.entry.fileIds();
+            if (fileIds != null && fileIds.contains(fileId)) {
+                return Optional.of(e.getKey());
+            }
+        }
+        return Optional.empty();
+    }
+
+    @Override
+    public Collection<JobStoreEntry> all() {
+        List<JobStoreEntry> result = new ArrayList<>(entries.size());
+        for (Holder h : entries.values()) {
+            if (!h.isExpired()) {
+                result.add(h.entry);
+            }
+        }
+        return result;
+    }
+
+    /** Drop entries whose TTL has elapsed. Called by the {@code TaskManager} cleanup scheduler. */
+    public int purgeExpired() {
+        int removed = 0;
+        Instant now = Instant.now();
+        for (Map.Entry<String, Holder> e : entries.entrySet()) {
+            if (!e.getValue().expiry.equals(Instant.MAX) && e.getValue().expiry.isBefore(now)) {
+                if (entries.remove(e.getKey(), e.getValue())) {
+                    removed++;
+                }
+            }
+        }
+        return removed;
+    }
+
+    private record Holder(JobStoreEntry entry, Instant expiry) {
+        boolean isExpired() {
+            return !expiry.equals(Instant.MAX) && expiry.isBefore(Instant.now());
+        }
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/InProcessKeyValueCache.java

@@ -0,0 +1,55 @@
+package stirling.software.common.cluster.inprocess;
+
+import java.time.Duration;
+import java.time.Instant;
+import java.util.Map;
+import java.util.Optional;
+import java.util.concurrent.ConcurrentHashMap;
+
+import stirling.software.common.cluster.KeyValueCache;
+
+public class InProcessKeyValueCache implements KeyValueCache {
+
+    private final ConcurrentHashMap<String, ConcurrentHashMap<String, Expiring>> namespaces =
+            new ConcurrentHashMap<>();
+
+    @Override
+    public void put(String namespace, String key, String value, Duration ttl) {
+        Instant expiry = ttl == null ? Instant.MAX : Instant.now().plus(ttl);
+        namespaces
+                .computeIfAbsent(namespace, n -> new ConcurrentHashMap<>())
+                .put(key, new Expiring(value, expiry));
+    }
+
+    @Override
+    public Optional<String> get(String namespace, String key) {
+        Map<String, Expiring> ns = namespaces.get(namespace);
+        if (ns == null) {
+            return Optional.empty();
+        }
+        Expiring e = ns.get(key);
+        if (e == null) {
+            return Optional.empty();
+        }
+        if (e.expiry.isBefore(Instant.now())) {
+            ns.remove(key, e);
+            return Optional.empty();
+        }
+        return Optional.of(e.value);
+    }
+
+    @Override
+    public void evict(String namespace, String key) {
+        Map<String, Expiring> ns = namespaces.get(namespace);
+        if (ns != null) {
+            ns.remove(key);
+        }
+    }
+
+    @Override
+    public void evictNamespace(String namespace) {
+        namespaces.remove(namespace);
+    }
+
+    private record Expiring(String value, Instant expiry) {}
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/InProcessRateLimitStore.java

@@ -0,0 +1,49 @@
+package stirling.software.common.cluster.inprocess;
+
+import java.time.Duration;
+import java.util.Collections;
+import java.util.LinkedHashMap;
+import java.util.Map;
+
+import io.github.bucket4j.Bandwidth;
+import io.github.bucket4j.Bucket;
+import io.github.bucket4j.ConsumptionProbe;
+import io.github.bucket4j.local.LocalBucketBuilder;
+
+import stirling.software.common.cluster.RateLimitStore;
+
+/** Bucket4j-backed token bucket implementation of {@link RateLimitStore}. */
+public class InProcessRateLimitStore implements RateLimitStore {
+
+    /** Cap to bound memory; oldest accessed buckets are evicted. */
+    private static final int MAX_BUCKETS = 10_000;
+
+    private final Map<String, Bucket> buckets =
+            Collections.synchronizedMap(
+                    new LinkedHashMap<String, Bucket>(256, 0.75f, true) {
+                        @Override
+                        protected boolean removeEldestEntry(Map.Entry<String, Bucket> eldest) {
+                            return size() > MAX_BUCKETS;
+                        }
+                    });
+
+    @Override
+    public RateLimitDecision tryConsume(String bucketKey, long capacity, Duration refillPeriod) {
+        String compositeKey = bucketKey + "|" + capacity + "|" + refillPeriod.toNanos();
+        Bucket bucket =
+                buckets.computeIfAbsent(compositeKey, k -> buildBucket(capacity, refillPeriod));
+        ConsumptionProbe probe = bucket.tryConsumeAndReturnRemaining(1);
+        return new RateLimitDecision(
+                probe.isConsumed(),
+                probe.getRemainingTokens(),
+                probe.isConsumed() ? 0L : probe.getNanosToWaitForRefill());
+    }
+
+    private static Bucket buildBucket(long capacity, Duration refillPeriod) {
+        Bandwidth limit =
+                Bandwidth.builder().capacity(capacity).refillGreedy(capacity, refillPeriod).build();
+        LocalBucketBuilder builder = Bucket.builder();
+        builder.addLimit(limit);
+        return builder.build();
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/LocalDiskFileStore.java

@@ -0,0 +1,205 @@
+package stirling.software.common.cluster.inprocess;
+
+import java.io.BufferedInputStream;
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.charset.StandardCharsets;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.UUID;
+import java.util.concurrent.locks.ReentrantLock;
+import java.util.regex.Pattern;
+
+import lombok.extern.slf4j.Slf4j;
+
+import stirling.software.common.cluster.FileStore;
+
+/** Local-disk {@link FileStore} storing files under a base directory keyed by a UUID file id. */
+@Slf4j
+public class LocalDiskFileStore implements FileStore {
+
+    private static final String OWNER_SUFFIX = ".owner";
+
+    // File ids are generated as random UUIDs; reject anything else so a tainted id can never reach
+    // Files.* APIs (defence in depth on top of the resolve() prefix check, and silences CodeQL's
+    // path-injection finding on the resolveOwner sidecar lookup).
+    private static final Pattern UUID_PATTERN =
+            Pattern.compile(
+                    "^[0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}$");
+
+    private final String baseDirPath;
+    // Fixed-size lock stripes so concurrent store/delete on the same (or colliding) fileId
+    // serialise the data-file + owner-sidecar pair as one critical section. Striped (not
+    // per-id) so the map never has to be cleaned up; collisions across unrelated ids are
+    // harmless contention.
+    private static final int LOCK_STRIPES = 64;
+    private final ReentrantLock[] stripes = new ReentrantLock[LOCK_STRIPES];
+
+    public LocalDiskFileStore(String baseDirPath) {
+        this.baseDirPath = baseDirPath;
+        for (int i = 0; i < LOCK_STRIPES; i++) {
+            stripes[i] = new ReentrantLock();
+        }
+    }
+
+    @Override
+    public Stored store(InputStream in, String originalName, String owner) throws IOException {
+        String fileId = UUID.randomUUID().toString();
+        Path filePath = resolve(fileId);
+        Files.createDirectories(filePath.getParent());
+        ReentrantLock lock = acquire(fileId);
+        boolean success = false;
+        try {
+            long size = Files.copy(in, filePath);
+            writeOwner(fileId, owner);
+            success = true;
+            return new Stored(fileId, size);
+        } finally {
+            if (!success) {
+                cleanupAfterFailedStore(fileId, filePath);
+            }
+            release(fileId, lock);
+        }
+    }
+
+    /**
+     * File-to-file copy. {@link Files#copy(Path, Path, java.nio.file.CopyOption...)} can use {@code
+     * sendfile(2)} on Linux for a zero-copy kernel transfer when source and destination share a
+     * filesystem, avoiding the streaming overhead of pulling the bytes through the JVM heap. Reads
+     * the source size before copying so the post-copy stat is unnecessary.
+     */
+    @Override
+    public Stored store(Path source, String originalName, String owner) throws IOException {
+        String fileId = UUID.randomUUID().toString();
+        Path filePath = resolve(fileId);
+        Files.createDirectories(filePath.getParent());
+        long size = Files.size(source);
+        ReentrantLock lock = acquire(fileId);
+        boolean success = false;
+        try {
+            Files.copy(source, filePath);
+            writeOwner(fileId, owner);
+            success = true;
+            return new Stored(fileId, size);
+        } finally {
+            if (!success) {
+                cleanupAfterFailedStore(fileId, filePath);
+            }
+            release(fileId, lock);
+        }
+    }
+
+    private void writeOwner(String fileId, String owner) throws IOException {
+        if (owner == null || owner.isBlank()) {
+            return;
+        }
+        Path ownerPath = resolveOwner(fileId);
+        Files.write(ownerPath, owner.getBytes(StandardCharsets.UTF_8));
+    }
+
+    private void cleanupAfterFailedStore(String fileId, Path filePath) {
+        try {
+            Files.deleteIfExists(filePath);
+        } catch (IOException cleanupEx) {
+            log.warn("Failed to clean up partial file {} after store failure", filePath, cleanupEx);
+        }
+        try {
+            Files.deleteIfExists(resolveOwner(fileId));
+        } catch (IOException cleanupEx) {
+            log.warn("Failed to clean up owner sidecar for {} after store failure", fileId);
+        }
+    }
+
+    @Override
+    public InputStream retrieve(String fileId) throws IOException {
+        return new BufferedInputStream(Files.newInputStream(resolve(fileId)));
+    }
+
+    @Override
+    public byte[] retrieveBytes(String fileId) throws IOException {
+        Path filePath = resolve(fileId);
+        if (!Files.exists(filePath)) {
+            throw new IOException("File not found with ID: " + fileId);
+        }
+        return Files.readAllBytes(filePath);
+    }
+
+    @Override
+    public long size(String fileId) throws IOException {
+        Path filePath = resolve(fileId);
+        if (!Files.exists(filePath)) {
+            throw new IOException("File not found with ID: " + fileId);
+        }
+        return Files.size(filePath);
+    }
+
+    @Override
+    public boolean delete(String fileId) {
+        ReentrantLock lock = acquire(fileId);
+        try {
+            // Data first, owner second: a concurrent retrieve that observes the transient
+            // (data-gone, owner-still-present) window simply fails with IOException; the inverse
+            // order would briefly look like an unowned file and could grant cross-user access.
+            boolean removed;
+            try {
+                removed = Files.deleteIfExists(resolve(fileId));
+            } catch (IOException e) {
+                log.error("Error deleting file with ID: {}", fileId, e);
+                return false;
+            }
+            try {
+                Files.deleteIfExists(resolveOwner(fileId));
+            } catch (IOException e) {
+                log.warn("Error deleting owner sidecar for file ID: {}", fileId, e);
+            }
+            return removed;
+        } finally {
+            release(fileId, lock);
+        }
+    }
+
+    @Override
+    public boolean exists(String fileId) {
+        return Files.exists(resolve(fileId));
+    }
+
+    @Override
+    public String getOwner(String fileId) throws IOException {
+        Path ownerPath = resolveOwner(fileId);
+        if (!Files.exists(ownerPath)) {
+            return null;
+        }
+        byte[] bytes = Files.readAllBytes(ownerPath);
+        if (bytes.length == 0) {
+            return null;
+        }
+        return new String(bytes, StandardCharsets.UTF_8);
+    }
+
+    public Path resolve(String fileId) {
+        if (fileId == null || !UUID_PATTERN.matcher(fileId).matches()) {
+            throw new IllegalArgumentException("Invalid file ID");
+        }
+        Path basePath = Path.of(baseDirPath).normalize().toAbsolutePath();
+        Path resolvedPath = basePath.resolve(fileId).normalize();
+        if (!resolvedPath.startsWith(basePath)) {
+            throw new IllegalArgumentException("File ID resolves to an invalid path");
+        }
+        return resolvedPath;
+    }
+
+    private Path resolveOwner(String fileId) {
+        Path data = resolve(fileId);
+        return data.resolveSibling(data.getFileName().toString() + OWNER_SUFFIX);
+    }
+
+    private ReentrantLock acquire(String fileId) {
+        ReentrantLock lock = stripes[(fileId.hashCode() & Integer.MAX_VALUE) % LOCK_STRIPES];
+        lock.lock();
+        return lock;
+    }
+
+    private void release(String fileId, ReentrantLock lock) {
+        lock.unlock();
+    }
+}

app/common/src/main/java/stirling/software/common/cluster/inprocess/LocalDiskFileStoreConfiguration.java

@@ -0,0 +1,29 @@
+package stirling.software.common.cluster.inprocess;
+
+import org.springframework.beans.factory.annotation.Value;
+import org.springframework.boot.autoconfigure.condition.ConditionalOnMissingBean;
+import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty;
+import org.springframework.context.annotation.Bean;
+import org.springframework.context.annotation.Configuration;
+
+import stirling.software.common.cluster.FileStore;
+
+/**
+ * Always-on wiring for the per-node local-disk {@link FileStore}. Active when {@code
+ * cluster.artifactStore=local} (the default; {@code matchIfMissing=true}). The S3 artifact-store
+ * supplies its own bean when {@code cluster.artifactStore=s3}.
+ */
+@Configuration
+@ConditionalOnProperty(
+        prefix = "cluster",
+        name = "artifactStore",
+        havingValue = "local",
+        matchIfMissing = true)
+public class LocalDiskFileStoreConfiguration {
+
+    @Bean
+    @ConditionalOnMissingBean
+    public FileStore fileStore(@Value("${stirling.tempDir:/tmp/stirling-files}") String tempDir) {
+        return new LocalDiskFileStore(tempDir);
+    }
+}

app/common/src/main/java/stirling/software/common/configuration/AppConfig.java

@@ -3,7 +3,6 @@ package stirling.software.common.configuration;
 import java.io.IOException;
 import java.nio.file.Files;
 import java.nio.file.Path;
-import java.nio.file.Paths;
 import java.util.List;
 import java.util.Locale;
 import java.util.Properties;
@@ -122,17 +121,17 @@ public class AppConfig {
 
     @Bean(name = "RunningInDocker")
     public boolean runningInDocker() {
-        return Files.exists(Paths.get("/.dockerenv"));
+        return Files.exists(Path.of("/.dockerenv"));
     }
 
     @Bean(name = "configDirMounted")
     public boolean isRunningInDockerWithConfig() {
-        Path dockerEnv = Paths.get("/.dockerenv");
+        Path dockerEnv = Path.of("/.dockerenv");
         // default to true if not docker
         if (!Files.exists(dockerEnv)) {
             return true;
         }
-        Path mountInfo = Paths.get("/proc/1/mountinfo");
+        Path mountInfo = Path.of("/proc/1/mountinfo");
         // this should always exist, if not some unknown usecase
         if (!Files.exists(mountInfo)) {
             return true;

app/common/src/main/java/stirling/software/common/configuration/ConfigInitializer.java

@@ -7,7 +7,6 @@ import java.net.URISyntaxException;
 import java.net.URL;
 import java.nio.file.Files;
 import java.nio.file.Path;
-import java.nio.file.Paths;
 import java.nio.file.StandardCopyOption;
 import java.util.List;
 
@@ -27,7 +26,7 @@ public class ConfigInitializer {
 
     public void ensureConfigExists() throws IOException, URISyntaxException {
         // 1) If settings file doesn't exist, create from template
-        Path destPath = Paths.get(InstallationPathConfig.getSettingsPath());
+        Path destPath = Path.of(InstallationPathConfig.getSettingsPath());
 
         boolean settingsFileExists = Files.exists(destPath);
 
@@ -39,7 +38,7 @@ public class ConfigInitializer {
             if (settingsFileExists) {
                 // move settings.yml to settings.yml.{timestamp}.bak
                 Path backupPath =
-                        Paths.get(
+                        Path.of(
                                 InstallationPathConfig.getSettingsPath()
                                         + "."
                                         + System.currentTimeMillis()
@@ -80,6 +79,7 @@ public class ConfigInitializer {
             YamlHelper settingsFile = new YamlHelper(settingTempPath);
 
             migrateEnterpriseEditionToPremium(settingsFile, settingsTemplateFile);
+            migrateProFeaturesKeyCasing(settingsFile, settingsTemplateFile);
 
             boolean changesMade =
                     settingsTemplateFile.updateValuesFromYaml(settingsFile, settingsTemplateFile);
@@ -95,7 +95,7 @@ public class ConfigInitializer {
         }
 
         // 3) Ensure custom settings file exists
-        Path customSettingsPath = Paths.get(InstallationPathConfig.getCustomSettingsPath());
+        Path customSettingsPath = Path.of(InstallationPathConfig.getCustomSettingsPath());
         if (Files.notExists(customSettingsPath)) {
             Files.createFile(customSettingsPath);
             log.info("Created custom_settings file: {}", customSettingsPath);
@@ -116,31 +116,52 @@ public class ConfigInitializer {
         }
         if (yaml.getValueByExactKeyPath("enterpriseEdition", "SSOAutoLogin") != null) {
             template.updateValue(
-                    List.of("premium", "proFeatures", "SSOAutoLogin"),
+                    List.of("premium", "proFeatures", "ssoAutoLogin"),
                     yaml.getValueByExactKeyPath("enterpriseEdition", "SSOAutoLogin"));
         }
         if (yaml.getValueByExactKeyPath("enterpriseEdition", "CustomMetadata", "autoUpdateMetadata")
                 != null) {
             template.updateValue(
-                    List.of("premium", "proFeatures", "CustomMetadata", "autoUpdateMetadata"),
+                    List.of("premium", "proFeatures", "customMetadata", "autoUpdateMetadata"),
                     yaml.getValueByExactKeyPath(
                             "enterpriseEdition", "CustomMetadata", "autoUpdateMetadata"));
         }
         if (yaml.getValueByExactKeyPath("enterpriseEdition", "CustomMetadata", "author") != null) {
             template.updateValue(
-                    List.of("premium", "proFeatures", "CustomMetadata", "author"),
+                    List.of("premium", "proFeatures", "customMetadata", "author"),
                     yaml.getValueByExactKeyPath("enterpriseEdition", "CustomMetadata", "author"));
         }
         if (yaml.getValueByExactKeyPath("enterpriseEdition", "CustomMetadata", "creator") != null) {
             template.updateValue(
-                    List.of("premium", "proFeatures", "CustomMetadata", "creator"),
+                    List.of("premium", "proFeatures", "customMetadata", "creator"),
                     yaml.getValueByExactKeyPath("enterpriseEdition", "CustomMetadata", "creator"));
         }
         if (yaml.getValueByExactKeyPath("enterpriseEdition", "CustomMetadata", "producer")
                 != null) {
             template.updateValue(
-                    List.of("premium", "proFeatures", "CustomMetadata", "producer"),
+                    List.of("premium", "proFeatures", "customMetadata", "producer"),
                     yaml.getValueByExactKeyPath("enterpriseEdition", "CustomMetadata", "producer"));
         }
     }
+
+    // TODO: Remove post migration
+    // settings.yml.template renamed the two non-camelCase proFeatures keys
+    // ("SSOAutoLogin" -> "ssoAutoLogin", "CustomMetadata" -> "customMetadata") so the whole
+    // settings pipeline is consistent camelCase. The save path (YamlHelper.updateValue) matches
+    // keys case-sensitively, so without this carry-forward an existing install's values written
+    // under the old PascalCase keys would be dropped on upgrade and reset to template defaults.
+    void migrateProFeaturesKeyCasing(YamlHelper yaml, YamlHelper template) {
+        Object ssoAutoLogin = yaml.getValueByExactKeyPath("premium", "proFeatures", "SSOAutoLogin");
+        if (ssoAutoLogin != null) {
+            template.updateValue(List.of("premium", "proFeatures", "ssoAutoLogin"), ssoAutoLogin);
+        }
+        for (String field : List.of("autoUpdateMetadata", "author", "creator", "producer")) {
+            Object value =
+                    yaml.getValueByExactKeyPath("premium", "proFeatures", "CustomMetadata", field);
+            if (value != null) {
+                template.updateValue(
+                        List.of("premium", "proFeatures", "customMetadata", field), value);
+            }
+        }
+    }
 }

app/common/src/main/java/stirling/software/common/configuration/RuntimePathConfig.java

@@ -3,7 +3,6 @@ package stirling.software.common.configuration;
 import java.nio.file.Files;
 import java.nio.file.InvalidPathException;
 import java.nio.file.Path;
-import java.nio.file.Paths;
 import java.util.ArrayList;
 import java.util.Collections;
 import java.util.LinkedHashSet;
@@ -201,7 +200,7 @@ public class RuntimePathConfig {
 
             try {
                 // Normalize to absolute path
-                Path path = Paths.get(pathStr.trim()).toAbsolutePath().normalize();
+                Path path = Path.of(pathStr.trim()).toAbsolutePath().normalize();
                 String normalizedPath = path.toString();
 
                 // Check for duplicates
@@ -224,9 +223,9 @@ public class RuntimePathConfig {
 
     private void detectOverlappingPaths(List<String> paths) {
         for (int i = 0; i < paths.size(); i++) {
-            Path path1 = Paths.get(paths.get(i));
+            Path path1 = Path.of(paths.get(i));
             for (int j = i + 1; j < paths.size(); j++) {
-                Path path2 = Paths.get(paths.get(j));
+                Path path2 = Path.of(paths.get(j));
 
                 // Check if one path is a parent of the other
                 if (path1.startsWith(path2)) {
@@ -246,10 +245,10 @@ public class RuntimePathConfig {
 
     private void validatePipelinePaths() {
         try {
-            Path finishedPath = Paths.get(pipelineFinishedFoldersPath).toAbsolutePath().normalize();
+            Path finishedPath = Path.of(pipelineFinishedFoldersPath).toAbsolutePath().normalize();
 
             for (String watchedPathStr : pipelineWatchedFoldersPaths) {
-                Path watchedPath = Paths.get(watchedPathStr).toAbsolutePath().normalize();
+                Path watchedPath = Path.of(watchedPathStr).toAbsolutePath().normalize();
 
                 // Check if watched folder is same as finished folder
                 if (watchedPath.equals(finishedPath)) {

app/common/src/main/java/stirling/software/common/model/ApplicationProperties.java

@@ -13,6 +13,7 @@ import java.util.Arrays;
 import java.util.Collection;
 import java.util.List;
 import java.util.Locale;
+import java.util.UUID;
 
 import org.springframework.boot.context.properties.ConfigurationProperties;
 import org.springframework.context.annotation.Bean;
@@ -76,7 +77,10 @@ public class ApplicationProperties {
     private ProcessExecutor processExecutor = new ProcessExecutor();
     private PdfEditor pdfEditor = new PdfEditor();
     private AiEngine aiEngine = new AiEngine();
+    private Mcp mcp = new Mcp();
     private InternalApi internalApi = new InternalApi();
+    private Cluster cluster = new Cluster();
+    private Policies policies = new Policies();
 
     @Bean
     public PropertySource<?> dynamicYamlPropertySource(ConfigurableEnvironment environment)
@@ -200,6 +204,45 @@ public class ApplicationProperties {
         }
     }
 
+    @Data
+    public static class Policies {
+        /**
+         * Absolute directories that policy folder input sources and output sinks may read from or
+         * write to. Empty (the default) disables folder access entirely, so a policy can never be
+         * pointed at an arbitrary server path. Stirling's own config directory is always
+         * off-limits, and folder access is always disabled in SaaS mode regardless of this list.
+         */
+        private List<String> allowedFolderRoots = new java.util.ArrayList<>();
+
+        /** How often (seconds) the schedule trigger checks for policies whose schedule is due. */
+        private long scheduleSweepSeconds = 60;
+
+        /**
+         * How often (seconds) the folder-watch trigger reconciles its watch registrations and
+         * re-runs every folder-watch policy as a safety net for filesystem events that were missed
+         * (NFS, bind mounts, inotify-queue overflow).
+         */
+        private long watchReconcileSeconds = 300;
+
+        /**
+         * How long (milliseconds) the folder-watch trigger keeps draining filesystem events after
+         * the first, so a burst from a single file copy coalesces into one run instead of many.
+         */
+        private long watchQuietPeriodMs = 500;
+
+        /**
+         * SSE emitter timeout (milliseconds) for streamed runs; generous for long multi-step runs.
+         */
+        private long streamTimeoutMs = 1800000;
+
+        /**
+         * How long (minutes) a finished run's in-memory state is retained before eviction,
+         * mirroring the job-result expiry so rich run state does not outlive the process. Active
+         * and paused runs are kept regardless of age.
+         */
+        private int runExpiryMinutes = 30;
+    }
+
     @Data
     public static class PdfEditor {
         private Cache cache = new Cache();
@@ -254,6 +297,203 @@ public class ApplicationProperties {
         private int longRunningTimeoutSeconds = 600;
     }
 
+    /**
+     * Model Context Protocol (MCP) server configuration. All keys live under the top-level {@code
+     * mcp.*} prefix. {@link #enabled} defaults to {@code false}: when off, no MCP beans are wired,
+     * no /mcp endpoint exists, and no protected-resource metadata is published.
+     */
+    @Data
+    public static class Mcp {
+
+        /** Master switch. When {@code false} (default), no MCP beans are wired. */
+        private boolean enabled = false;
+
+        /**
+         * When {@code true} (default), invocations require an OAuth scope: {@code mcp.tools.read}
+         * for read-style operations and {@code mcp.tools.write} for write/destructive ones. When
+         * {@code false}, scope checks are skipped (use only if your IdP issues a single coarse
+         * scope).
+         */
+        private boolean scopesEnabled = true;
+
+        /** How often to refresh the AI capabilities manifest from the engine. */
+        private int engineCapabilityRefreshMinutes = 5;
+
+        /**
+         * Tool allow-list (operation ids, e.g. {@code compress-pdf}). When non-empty, ONLY these
+         * operations are exposed over MCP; everything else is hidden, undescribable, and
+         * uninvocable - on top of the global endpoint enable/disable config. Empty = allow all.
+         */
+        private List<String> allowedOperations = new ArrayList<>();
+
+        /**
+         * Tool deny-list (operation ids). Any operation listed here is removed from MCP even if it
+         * would otherwise be allowed. Applied after {@link #allowedOperations}.
+         */
+        private List<String> blockedOperations = new ArrayList<>();
+
+        /** Max MCP request body size in bytes; inline file uploads ride in the JSON-RPC body. */
+        private long maxRequestBytes = 10L * 1024 * 1024;
+
+        /** Results up to this size return inline as base64; larger ones return a fileId only. */
+        private long maxInlineResponseBytes = 10L * 1024 * 1024;
+
+        private Auth auth = new Auth();
+
+        @Data
+        public static class Auth {
+            /**
+             * Authentication mode for the MCP endpoint. {@code oauth} (default) runs a full OAuth2
+             * resource server (JWT, RFC 8707 audience, RFC 9728 metadata). {@code apikey} accepts a
+             * Stirling per-user API key via the {@code X-API-KEY} header (or {@code Authorization:
+             * Bearer <key>}) and binds the request to that user - the low-friction self-host path,
+             * no external IdP required.
+             */
+            private String mode = "oauth";
+
+            /** OAuth2 issuer URI, e.g. {@code http://localhost:9000}. Required when MCP is on. */
+            private String issuerUri = "";
+
+            /**
+             * JWKS URI. When blank, derived from the issuer's {@code
+             * /.well-known/openid-configuration} document.
+             */
+            private String jwksUri = "";
+
+            /**
+             * RFC 8707 resource identifier of THIS MCP server, e.g. {@code
+             * http://localhost:8080/mcp}. Tokens that do not list this id in their {@code aud}
+             * claim are rejected with HTTP 401.
+             */
+            private String resourceId = "";
+
+            /**
+             * Additional JWT audiences accepted at the MCP endpoint, on top of {@link #resourceId}.
+             * Empty (default) keeps strict RFC 8707 binding. Some IdPs cannot mint
+             * resource-specific audiences - e.g. Supabase's OAuth server always issues {@code
+             * aud=authenticated} - so operators list the audience their IdP actually emits here
+             * (env: {@code MCP_AUTH_ACCEPTEDAUDIENCES}, comma-separated).
+             */
+            private List<String> acceptedAudiences = new ArrayList<>();
+
+            /**
+             * JWT claim whose value is matched against a provisioned Stirling username. Defaults to
+             * {@code sub}; set to {@code email} or {@code preferred_username} to match how your IdP
+             * maps users to Stirling accounts.
+             */
+            private String usernameClaim = "sub";
+
+            /**
+             * When {@code true} (default), a validated token is accepted only if its {@link
+             * #usernameClaim} value resolves to an existing, enabled Stirling user account. Tokens
+             * whose subject has no Stirling account (or a disabled one) are rejected with HTTP 403.
+             * Set to {@code false} only if you intentionally want any IdP-valid token to use MCP
+             * without a local account.
+             */
+            private boolean requireExistingAccount = true;
+        }
+    }
+
+    /**
+     * Cluster backplane configuration. All keys live under the top-level {@code cluster.*} prefix
+     * (e.g. env var {@code CLUSTER_ENABLED}). The master switch is {@link #enabled} and defaults to
+     * off; when off the in-process backplane is wired and no other cluster keys are required.
+     */
+    @Data
+    public static class Cluster {
+
+        /** Master switch. When {@code false} (default) the in-process backplane is wired. */
+        private boolean enabled = false;
+
+        /** Backplane implementation selector. Valid values: {@code inprocess} | {@code valkey}. */
+        private String backplane = "inprocess";
+
+        /**
+         * Transient cluster job-artifact store selector. Valid values: {@code local} | {@code s3}.
+         *
+         * <p>This is distinct from {@code storage.provider}, which selects the backend for
+         * persistent user-uploaded files. The two switches exist because the user-facing storage
+         * feature is optional ({@code storage.enabled=false} is common) but every multi-node
+         * cluster still needs a shared artifact store to serve cross-node downloads. Both
+         * implementations share credentials from {@code storage.s3.*} when set to {@code s3}.
+         */
+        private String artifactStore = "local";
+
+        private Valkey valkey = new Valkey();
+        private Node node = new Node();
+
+        private transient String cachedNodeId;
+
+        public NodeRole resolvedRole() {
+            if (node == null || node.getRole() == null) {
+                return NodeRole.BOTH;
+            }
+            String value = node.getRole().trim().toUpperCase(Locale.ROOT);
+            try {
+                return NodeRole.valueOf(value);
+            } catch (IllegalArgumentException ex) {
+                return NodeRole.BOTH;
+            }
+        }
+
+        public synchronized String resolvedNodeId() {
+            if (node != null && node.getId() != null && !node.getId().isBlank()) {
+                return node.getId();
+            }
+            if (cachedNodeId == null) {
+                cachedNodeId = UUID.randomUUID().toString();
+            }
+            return cachedNodeId;
+        }
+
+        public enum NodeRole {
+            WEB,
+            WORKER,
+            BOTH
+        }
+
+        @Data
+        public static class Valkey {
+            /**
+             * {@code redis://host:6379} or {@code rediss://...} for TLS. Required when cluster mode
+             * is on and backplane is valkey.
+             */
+            private String url = "";
+
+            private Tls tls = new Tls();
+
+            @Data
+            public static class Tls {
+                /**
+                 * When {@code true}, skip Valkey/Redis TLS certificate verification (dev/test
+                 * only). Leave {@code false} in production.
+                 */
+                private boolean skipCertVerification = false;
+            }
+        }
+
+        @Data
+        public static class Node {
+            /** Optional explicit node id. Blank = auto-generated UUID at startup. */
+            private String id = "";
+
+            /** {@code web} | {@code worker} | {@code both}. */
+            private String role = "both";
+
+            /**
+             * Internal cluster address advertised in the instance registry (host:port). Blank =
+             * derived at startup.
+             */
+            private String internalAddress = "";
+
+            /** {@code http} | {@code https} - scheme used when peers call this node. */
+            private String scheme = "http";
+
+            /** Heartbeat publish interval for the instance registry, in milliseconds. */
+            private long heartbeatIntervalMs = 5000;
+        }
+    }
+
     /**
      * HTTP timeouts for loopback calls to internal Stirling API endpoints, used by the AI workflow
      * executor and the pipeline processor. A bounded read timeout prevents a hung tool (e.g. an
@@ -426,6 +666,16 @@ public class ApplicationProperties {
             private String provider;
             private Client client = new Client();
 
+            /**
+             * When true, the OAuth2/OIDC login flow logs the full set of ID token and UserInfo
+             * claims at INFO level (and again at ERROR level if the username attribute cannot be
+             * resolved). Used to diagnose provider misconfiguration (for example ADFS not returning
+             * an {@code email} claim). WARNING: writes PII (sub, email, name) to application logs.
+             * Leave disabled in production; enable only while actively troubleshooting and disable
+             * again afterwards.
+             */
+            private Boolean debugLogging = false;
+
             public void setScopes(String scopes) {
                 List<String> scopesList =
                         Arrays.stream(scopes.split(",")).map(String::trim).toList();
@@ -676,6 +926,7 @@ public class ApplicationProperties {
         private boolean enabled = false;
         private String provider = "local";
         private Local local = new Local();
+        private S3 s3 = new S3();
         private Quotas quotas = new Quotas();
         private Sharing sharing = new Sharing();
         private Signing signing = new Signing();
@@ -685,6 +936,57 @@ public class ApplicationProperties {
             private String basePath = InstallationPathConfig.getPath() + "storage";
         }
 
+        @Data
+        public static class S3 {
+            /**
+             * Optional custom endpoint (e.g. {@code https://<account>.r2.cloudflarestorage.com},
+             * {@code https://<project>.supabase.co/storage/v1/s3}, or {@code http://localhost:9000}
+             * for MinIO). Blank = use AWS regional default.
+             */
+            private String endpoint = "";
+
+            private String bucket = "";
+
+            private String region = "us-east-1";
+
+            private String accessKey = "";
+            private String secretKey = "";
+
+            /**
+             * When {@code true} use path-style URLs ({@code <endpoint>/<bucket>/<key>}) instead of
+             * virtual-hosted ({@code <bucket>.<endpoint>/<key>}). MinIO and most S3-compatible
+             * gateways require path-style; AWS S3 prefers virtual-hosted.
+             */
+            private boolean pathStyleAccess = false;
+
+            /**
+             * When {@code false} (default), {@code endpoint} hostnames that resolve to private,
+             * loopback, or link-local addresses are rejected at startup to block SSRF attacks via
+             * the cloud metadata service (e.g. {@code http://169.254.169.254/}). Set to {@code
+             * true} to opt in for MinIO / in-cluster S3 endpoints on private networks.
+             */
+            private boolean allowPrivateEndpoints = false;
+
+            /**
+             * Controls when the SDK adds an {@code x-amz-checksum-*} header on PUT/UploadPart.
+             * Default {@code WHEN_SUPPORTED} (the SDK default since 2.30) makes the SDK send a
+             * CRC32 checksum on every upload - this works on AWS S3, MinIO, current Supabase,
+             * Backblaze B2 (post-July-2025), and modern R2. Set to {@code WHEN_REQUIRED} to
+             * suppress the auto-checksum on vendors that reject unknown {@code x-amz-checksum-*}
+             * headers (older Backblaze B2, some R2 corner cases, GCS S3 endpoint). Invalid values
+             * fall back to {@code WHEN_SUPPORTED}.
+             */
+            private String requestChecksumCalculation = "WHEN_SUPPORTED";
+
+            /**
+             * Controls when the SDK validates returned {@code x-amz-checksum-*} headers on GET
+             * responses. Default {@code WHEN_SUPPORTED}. Set to {@code WHEN_REQUIRED} if your
+             * vendor never returns these headers and you see false-positive checksum-mismatch
+             * errors. Invalid values fall back to {@code WHEN_SUPPORTED}.
+             */
+            private String responseChecksumValidation = "WHEN_SUPPORTED";
+        }
+
         @Data
         public static class Sharing {
             private boolean enabled = false;
@@ -703,6 +1005,10 @@ public class ApplicationProperties {
         @Data
         public static class Signing {
             private boolean enabled = false;
+
+            // Signing user-picker scope: 'org' (default) = whole instance, anything else =
+            // caller's team only (fail-closed). The saas profile pins 'team'.
+            private String userListScope = "org";
         }
     }
 

app/common/src/main/java/stirling/software/common/model/FileInfo.java

@@ -1,7 +1,6 @@
 package stirling.software.common.model;
 
 import java.nio.file.Path;
-import java.nio.file.Paths;
 import java.time.LocalDateTime;
 import java.time.format.DateTimeFormatter;
 import java.util.Locale;
@@ -24,7 +23,7 @@ public class FileInfo {
 
     // Converts the file path string to a Path object.
     public Path getFilePathAsPath() {
-        return Paths.get(filePath);
+        return Path.of(filePath);
     }
 
     // Formats the file size into a human-readable string.

app/common/src/main/java/stirling/software/common/pdf/HeadingDetector.java

@@ -0,0 +1,191 @@
+package stirling.software.common.pdf;
+
+import java.util.ArrayList;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+import stirling.software.jpdfium.text.PageText;
+import stirling.software.jpdfium.text.TextChar;
+import stirling.software.jpdfium.text.TextLine;
+import stirling.software.jpdfium.text.TextWord;
+
+final class HeadingDetector {
+
+    private HeadingDetector() {}
+
+    /** A heading is at most this many words; longer lines are treated as body text. */
+    private static final int MAX_HEADING_WORDS = 12;
+
+    /**
+     * Returns the Markdown heading prefix for a line. The decision combines several signals, never
+     * text matching, so a plain line that merely shares text with a heading is never promoted:
+     *
+     * <ul>
+     *   <li><b>Size</b> — dominant glyph font size vs. the document body median (primary signal).
+     *       Some PDFs encode visual size in the text matrix, so every glyph reports ~1.0; for those
+     *       the line height is used as the proxy instead.
+     *   <li><b>Brevity</b> — headings are short labels; a line over {@value #MAX_HEADING_WORDS}
+     *       words is body text regardless of size.
+     *   <li><b>Not a sentence</b> — a line ending in {@code . ! ?} reads as prose, not a heading.
+     * </ul>
+     *
+     * <p>Boldness is deliberately <em>not</em> a heading signal — a bold-but-not-larger line is
+     * emphasis, not a heading (see {@link #isBoldLabel}); promoting it to {@code #}/{@code ##} is
+     * the main source of false-positive headings.
+     *
+     * <ul>
+     *   <li>size &gt; baseline * 1.4 → {@code "# "}
+     *   <li>size &gt; baseline * 1.2 → {@code "## "}
+     *   <li>otherwise → {@code ""}
+     * </ul>
+     */
+    static String headingPrefix(TextLine line, float medianBodySize, float medianBodyHeight) {
+        String text = line.text().strip();
+        if (text.isEmpty() || wordCount(text) > MAX_HEADING_WORDS || endsLikeSentence(text)) {
+            return "";
+        }
+
+        float dominant = dominantFontSize(line);
+        float value;
+        float baseline;
+        if (dominant > 2f && medianBodySize > 2f) {
+            value = dominant;
+            baseline = medianBodySize;
+        } else {
+            value = line.height();
+            baseline = medianBodyHeight;
+        }
+        if (baseline <= 0f) {
+            return "";
+        }
+
+        float ratio = value / baseline;
+        if (ratio > 1.4f) {
+            return "# ";
+        }
+        if (ratio > 1.2f) {
+            return "## ";
+        }
+        return "";
+    }
+
+    /**
+     * True when a line should be emphasised as bold (rendered {@code **like this**}) rather than
+     * promoted to a heading: it is bold, short, and not a full sentence. Used for bold labels that
+     * are not large enough to be headings.
+     */
+    static boolean isBoldLabel(TextLine line) {
+        String text = line.text().strip();
+        if (text.isEmpty() || wordCount(text) > MAX_HEADING_WORDS || endsLikeSentence(text)) {
+            return false;
+        }
+        return isBold(line);
+    }
+
+    private static int wordCount(String text) {
+        return text.split("\\s+").length;
+    }
+
+    private static boolean endsLikeSentence(String text) {
+        char last = text.charAt(text.length() - 1);
+        return last == '.' || last == '!' || last == '?';
+    }
+
+    /** True when the line's dominant font is bold, inferred from PostScript font names. */
+    private static boolean isBold(TextLine line) {
+        Map<String, Integer> counts = new HashMap<>();
+        for (TextWord word : line.words()) {
+            for (TextChar ch : word.chars()) {
+                if (ch.isWhitespace() || ch.isNewline()) {
+                    continue;
+                }
+                String name = ch.fontName();
+                if (name != null && !name.isBlank()) {
+                    counts.merge(name, 1, Integer::sum);
+                }
+            }
+        }
+        String dominantFont = "";
+        int max = -1;
+        for (Map.Entry<String, Integer> e : counts.entrySet()) {
+            if (e.getValue() > max) {
+                max = e.getValue();
+                dominantFont = e.getKey();
+            }
+        }
+        String lower = dominantFont.toLowerCase(java.util.Locale.ROOT);
+        return lower.contains("bold")
+                || lower.contains("black")
+                || lower.contains("heavy")
+                || lower.contains("semibold");
+    }
+
+    /** Computes the median glyph font size across all pages. */
+    static float medianFontSize(List<PageText> allPages) {
+        List<Float> sizes = new ArrayList<>();
+        for (PageText page : allPages) {
+            for (TextChar ch : page.chars()) {
+                if (!ch.isWhitespace() && !ch.isNewline() && ch.fontSize() > 0f) {
+                    sizes.add(ch.fontSize());
+                }
+            }
+        }
+        return median(sizes, 12f);
+    }
+
+    /** Computes the median TextLine height across all pages. Used when font size is degenerate. */
+    static float medianLineHeight(List<PageText> allPages) {
+        List<Float> heights = new ArrayList<>();
+        for (PageText page : allPages) {
+            for (TextLine line : page.lines()) {
+                if (line.height() > 0f && !line.text().isBlank()) {
+                    heights.add(line.height());
+                }
+            }
+        }
+        return median(heights, 12f);
+    }
+
+    private static float median(List<Float> values, float fallback) {
+        if (values.isEmpty()) {
+            return fallback;
+        }
+        Collections.sort(values);
+        int mid = values.size() / 2;
+        if (values.size() % 2 == 0) {
+            return (values.get(mid - 1) + values.get(mid)) / 2f;
+        }
+        return values.get(mid);
+    }
+
+    /**
+     * Returns the font size that appears most often (by character count) in the given line. Ties
+     * are broken in favour of the larger size.
+     */
+    private static float dominantFontSize(TextLine line) {
+        Map<Float, Integer> counts = new HashMap<>();
+        for (TextWord word : line.words()) {
+            for (TextChar ch : word.chars()) {
+                if (!ch.isWhitespace() && !ch.isNewline() && ch.fontSize() > 0f) {
+                    counts.merge(ch.fontSize(), 1, Integer::sum);
+                }
+            }
+        }
+        if (counts.isEmpty()) {
+            return 0f;
+        }
+        float dominant = 0f;
+        int maxCount = -1;
+        for (Map.Entry<Float, Integer> entry : counts.entrySet()) {
+            int count = entry.getValue();
+            float size = entry.getKey();
+            if (count > maxCount || (count == maxCount && size > dominant)) {
+                maxCount = count;
+                dominant = size;
+            }
+        }
+        return dominant;
+    }
+}

app/common/src/main/java/stirling/software/common/pdf/TableRenderer.java

@@ -0,0 +1,82 @@
+package stirling.software.common.pdf;
+
+import stirling.software.jpdfium.text.Table;
+
+final class TableRenderer {
+    private TableRenderer() {}
+
+    /** Renders a Table as a GitHub-Flavoured Markdown table string. */
+    static String render(Table table) {
+        if (table.rowCount() == 0) {
+            return "";
+        }
+
+        String[][] grid = table.asGrid();
+
+        if (table.rowCount() < 2) {
+            // No separator row possible — return plain lines
+            StringBuilder sb = new StringBuilder();
+            for (int c = 0; c < grid[0].length; c++) {
+                if (c > 0) sb.append('\n');
+                sb.append(escape(grid[0][c].trim()));
+            }
+            return sb.toString();
+        }
+
+        int cols = grid[0].length;
+
+        // Compute column widths: max(3, max content length across all rows)
+        int[] widths = new int[cols];
+        for (int c = 0; c < cols; c++) {
+            widths[c] = 3;
+        }
+        for (String[] row : grid) {
+            for (int c = 0; c < cols; c++) {
+                String cell = c < row.length ? row[c].trim() : "";
+                widths[c] = Math.max(widths[c], escape(cell).length());
+            }
+        }
+
+        StringBuilder sb = new StringBuilder();
+
+        // Header row
+        sb.append(buildRow(grid[0], widths, cols));
+        sb.append('\n');
+
+        // Separator row
+        sb.append('|');
+        for (int c = 0; c < cols; c++) {
+            sb.append('-').append("-".repeat(widths[c])).append('-').append('|');
+        }
+        sb.append('\n');
+
+        // Data rows
+        for (int r = 1; r < grid.length; r++) {
+            sb.append(buildRow(grid[r], widths, cols));
+            if (r < grid.length - 1) {
+                sb.append('\n');
+            }
+        }
+
+        return sb.toString();
+    }
+
+    private static String buildRow(String[] row, int[] widths, int cols) {
+        StringBuilder sb = new StringBuilder();
+        sb.append('|');
+        for (int c = 0; c < cols; c++) {
+            String cell = c < row.length ? escape(row[c].trim()) : "";
+            sb.append(' ').append(padRight(cell, widths[c])).append(' ').append('|');
+        }
+        return sb.toString();
+    }
+
+    private static String escape(String cell) {
+        return cell.replace("|", "\\|");
+    }
+
+    private static String padRight(String s, int width) {
+        if (s.length() >= width) return s;
+        return s + " ".repeat(width - s.length());
+    }
+}

app/common/src/main/java/stirling/software/common/service/FileStorage.java

@@ -1,15 +1,14 @@
 package stirling.software.common.service;
 
-import java.io.BufferedInputStream;
-import java.io.BufferedOutputStream;
+import java.io.ByteArrayInputStream;
 import java.io.IOException;
 import java.io.InputStream;
-import java.io.OutputStream;
-import java.nio.file.Files;
-import java.nio.file.Path;
-import java.util.UUID;
+import java.io.PipedInputStream;
+import java.io.PipedOutputStream;
+import java.util.Optional;
+import java.util.concurrent.Executors;
+import java.util.concurrent.atomic.AtomicReference;
 
-import org.springframework.beans.factory.annotation.Value;
 import org.springframework.core.io.Resource;
 import org.springframework.stereotype.Service;
 import org.springframework.web.multipart.MultipartFile;
@@ -18,9 +17,12 @@ import org.springframework.web.servlet.mvc.method.annotation.StreamingResponseBo
 import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
 
+import stirling.software.common.cluster.FileStore;
+import stirling.software.common.util.JobContext;
+
 /**
  * Service for storing and retrieving files with unique file IDs. Used by the AutoJobPostMapping
- * system to handle file references.
+ * system to handle file references. Disk I/O is delegated to the injected {@link FileStore} bean.
  */
 @Service
 @RequiredArgsConstructor
@@ -30,251 +32,198 @@ public class FileStorage {
     /** Holds the result of a stream-to-disk store operation: the file ID and the bytes written. */
     public record StoredFile(String fileId, long size) {}
 
-    @Value("${stirling.tempDir:/tmp/stirling-files}")
-    private String tempDirPath;
-
     private final FileOrUploadService fileOrUploadService;
+    private final FileStore fileStore;
+    private final Optional<JobOwnershipService> jobOwnershipService;
 
-    /**
-     * Store a file and return its unique ID
-     *
-     * @param file The file to store
-     * @return The unique ID assigned to the file
-     * @throws IOException If there is an error storing the file
-     */
     public String storeFile(MultipartFile file) throws IOException {
-        String fileId = generateFileId();
-        Path filePath = getFilePath(fileId);
-
-        // Ensure the directory exists
-        Files.createDirectories(filePath.getParent());
-
-        // Transfer the file to the storage location
-        file.transferTo(filePath.toFile());
-
-        log.debug("Stored file with ID: {}", fileId);
-        return fileId;
-    }
-
-    /**
-     * Store a byte array as a file and return its unique ID
-     *
-     * @param bytes The byte array to store
-     * @param originalName The original name of the file (for extension)
-     * @return The unique ID assigned to the file
-     * @throws IOException If there is an error storing the file
-     */
-    public String storeBytes(byte[] bytes, String originalName) throws IOException {
-        String fileId = generateFileId();
-        Path filePath = getFilePath(fileId);
-
-        // Ensure the directory exists
-        Files.createDirectories(filePath.getParent());
-
-        // Write the bytes to the file
-        Files.write(filePath, bytes);
-
-        log.debug("Stored byte array with ID: {}", fileId);
-        return fileId;
-    }
-
-    /**
-     * Retrieve a file by its ID as a MultipartFile
-     *
-     * @param fileId The ID of the file to retrieve
-     * @return The file as a MultipartFile
-     * @throws IOException If the file doesn't exist or can't be read
-     */
-    public MultipartFile retrieveFile(String fileId) throws IOException {
-        Path filePath = getFilePath(fileId);
-
-        if (!Files.exists(filePath)) {
-            throw new IOException("File not found with ID: " + fileId);
+        String owner = resolveOwner();
+        // Fast path: when Spring buffered the multipart to disk (typical for large uploads), the
+        // backing Resource exposes a real File. Hand the Path to the FileStore so it can do a
+        // file-to-file copy (Linux sendfile, no copy through Java heap) rather than streaming
+        // the bytes through an 8K buffer. Falls back to the InputStream path for in-memory
+        // multiparts, exotic Resource impls, and anything that does not back onto a File.
+        Resource res;
+        try {
+            res = file.getResource();
+        } catch (RuntimeException ignored) {
+            res = null;
         }
+        if (res != null && res.isFile()) {
+            try {
+                FileStore.Stored stored =
+                        fileStore.store(res.getFile().toPath(), file.getOriginalFilename(), owner);
+                log.debug("Stored file with ID: {} (fast path)", stored.fileId());
+                return stored.fileId();
+            } catch (IOException ex) {
+                // Some Resource impls advertise isFile()=true but throw on getFile(); fall through.
+                log.debug("Resource fast path failed, falling back to stream copy", ex);
+            }
+        }
+        try (InputStream in = file.getInputStream()) {
+            FileStore.Stored stored = fileStore.store(in, file.getOriginalFilename(), owner);
+            log.debug("Stored file with ID: {}", stored.fileId());
+            return stored.fileId();
+        }
+    }
 
-        byte[] fileData = Files.readAllBytes(filePath);
+    public String storeBytes(byte[] bytes, String originalName) throws IOException {
+        FileStore.Stored stored =
+                fileStore.store(new ByteArrayInputStream(bytes), originalName, resolveOwner());
+        log.debug("Stored byte array with ID: {}", stored.fileId());
+        return stored.fileId();
+    }
+
+    public MultipartFile retrieveFile(String fileId) throws IOException {
+        enforceOwnership(fileId);
+        byte[] fileData = fileStore.retrieveBytes(fileId);
         return fileOrUploadService.toMockMultipartFile(fileId, fileData);
     }
 
-    /**
-     * Retrieve a file by its ID as a byte array
-     *
-     * @param fileId The ID of the file to retrieve
-     * @return The file as a byte array
-     * @throws IOException If the file doesn't exist or can't be read
-     */
     public byte[] retrieveBytes(String fileId) throws IOException {
-        Path filePath = getFilePath(fileId);
-
-        if (!Files.exists(filePath)) {
-            throw new IOException("File not found with ID: " + fileId);
-        }
-
-        return Files.readAllBytes(filePath);
+        enforceOwnership(fileId);
+        return fileStore.retrieveBytes(fileId);
     }
 
-    /**
-     * Retrieve a file by its ID as a streaming InputStream. The caller is responsible for closing
-     * the returned stream.
-     *
-     * @param fileId The ID of the file to retrieve
-     * @return A buffered InputStream for the file
-     * @throws IOException If the file doesn't exist or can't be read
-     */
     public InputStream retrieveInputStream(String fileId) throws IOException {
-        Path filePath = getFilePath(fileId);
-        // Let Files.newInputStream throw NoSuchFileException naturally — avoids TOCTOU race
-        // between exists-check and open when another thread may delete concurrently.
-        return new BufferedInputStream(Files.newInputStream(filePath));
+        enforceOwnership(fileId);
+        return fileStore.retrieve(fileId);
     }
 
-    /**
-     * Store data from an InputStream as a file and return its unique ID and byte count. Streams
-     * directly to disk without buffering the entire content in heap.
-     *
-     * @param inputStream The input stream to read from
-     * @param originalName The original name of the file (unused, kept for API symmetry)
-     * @return A {@link StoredFile} containing the file ID and the number of bytes written
-     * @throws IOException If there is an error storing the file
-     */
     public StoredFile storeInputStream(InputStream inputStream, String originalName)
             throws IOException {
-        String fileId = generateFileId();
-        Path filePath = getFilePath(fileId);
-        Files.createDirectories(filePath.getParent());
-        long size = Files.copy(inputStream, filePath);
-        log.debug("Stored input stream with ID: {}", fileId);
-        return new StoredFile(fileId, size);
+        FileStore.Stored stored = fileStore.store(inputStream, originalName, resolveOwner());
+        log.debug("Stored input stream with ID: {}", stored.fileId());
+        return new StoredFile(stored.fileId(), stored.size());
     }
 
     public String storeFromStreamingBody(StreamingResponseBody body, String originalName)
             throws IOException {
-        String fileId = generateFileId();
-        Path filePath = getFilePath(fileId);
-        Files.createDirectories(filePath.getParent());
-        boolean success = false;
-        try (OutputStream os = new BufferedOutputStream(Files.newOutputStream(filePath))) {
-            body.writeTo(os);
-            success = true;
-        } finally {
-            if (!success) {
+        String owner = resolveOwner();
+        // Hold Throwable not IOException: an unchecked failure (NPE, IllegalState, OOM, etc.)
+        // from the body writer would otherwise close the pipe with EOF and the consumer would
+        // return a truncated file with no error surfaced to the caller.
+        AtomicReference<Throwable> bodyError = new AtomicReference<>();
+        try (PipedOutputStream out = new PipedOutputStream();
+                PipedInputStream in = new PipedInputStream(out, 8192)) {
+            var executor = Executors.newSingleThreadExecutor(Thread.ofVirtual().factory());
+            java.util.concurrent.Future<?> task = null;
+            try {
+                task =
+                        executor.submit(
+                                () -> {
+                                    try {
+                                        body.writeTo(out);
+                                    } catch (Throwable ex) {
+                                        bodyError.set(ex);
+                                    } finally {
+                                        try {
+                                            out.close();
+                                        } catch (IOException ignored) {
+                                            // closed on the consumer side too
+                                        }
+                                    }
+                                });
+                FileStore.Stored stored = fileStore.store(in, originalName, owner);
+                Throwable writerErr = bodyError.get();
+                if (writerErr != null) {
+                    // Body failed mid-write: the FileStore persisted a truncated entry.
+                    // Best-effort delete so we don't leak partial files; never let cleanup
+                    // mask the original writer error.
+                    try {
+                        fileStore.delete(stored.fileId());
+                    } catch (RuntimeException cleanupEx) {
+                        log.warn(
+                                "Failed to delete partial file {} after writer error: {}",
+                                stored.fileId(),
+                                cleanupEx.getMessage());
+                    }
+                    if (writerErr instanceof IOException ioe) {
+                        throw ioe;
+                    }
+                    throw new IOException(
+                            "StreamingResponseBody writer failed: " + writerErr.getMessage(),
+                            writerErr);
+                }
+                log.debug("Stored StreamingResponseBody with ID: {}", stored.fileId());
+                return stored.fileId();
+            } finally {
+                // Interrupt and join the writer task: shutdown() alone returns immediately and a
+                // failed store leaves the writer running, leaking a thread per failed upload.
+                if (task != null && !task.isDone()) {
+                    task.cancel(true);
+                }
+                executor.shutdown();
                 try {
-                    Files.deleteIfExists(filePath);
-                } catch (IOException cleanupEx) {
-                    log.warn(
-                            "Failed to clean up partial file {} after store failure",
-                            filePath,
-                            cleanupEx);
+                    if (!executor.awaitTermination(5, java.util.concurrent.TimeUnit.SECONDS)) {
+                        executor.shutdownNow();
+                    }
+                } catch (InterruptedException ie) {
+                    executor.shutdownNow();
+                    Thread.currentThread().interrupt();
                 }
             }
         }
-        log.debug("Stored StreamingResponseBody with ID: {}", fileId);
-        return fileId;
     }
 
-    /**
-     * Persist a {@link Resource} body to disk, returning the generated file ID. Used by the async
-     * job pipeline to capture {@code ResponseEntity<Resource>} results produced by controllers.
-     */
     public String storeFromResource(Resource resource, String originalName) throws IOException {
-        String fileId = generateFileId();
-        Path filePath = getFilePath(fileId);
-        Files.createDirectories(filePath.getParent());
-        boolean success = false;
         try (InputStream in = resource.getInputStream()) {
-            Files.copy(in, filePath);
-            success = true;
-        } finally {
-            if (!success) {
-                try {
-                    Files.deleteIfExists(filePath);
-                } catch (IOException cleanupEx) {
-                    log.warn(
-                            "Failed to clean up partial file {} after store failure",
-                            filePath,
-                            cleanupEx);
-                }
-            }
+            FileStore.Stored stored = fileStore.store(in, originalName, resolveOwner());
+            log.debug("Stored Resource with ID: {}", stored.fileId());
+            return stored.fileId();
         }
-        log.debug("Stored Resource with ID: {}", fileId);
-        return fileId;
     }
 
-    /**
-     * Delete a file by its ID
-     *
-     * @param fileId The ID of the file to delete
-     * @return true if the file was deleted, false otherwise
-     */
     public boolean deleteFile(String fileId) {
-        try {
-            Path filePath = getFilePath(fileId);
-            return Files.deleteIfExists(filePath);
-        } catch (IOException e) {
-            log.error("Error deleting file with ID: {}", fileId, e);
-            return false;
-        }
+        enforceOwnership(fileId);
+        return fileStore.delete(fileId);
     }
 
-    /**
-     * Check if a file exists by its ID
-     *
-     * @param fileId The ID of the file to check
-     * @return true if the file exists, false otherwise
-     */
     public boolean fileExists(String fileId) {
-        Path filePath = getFilePath(fileId);
-        return Files.exists(filePath);
+        enforceOwnership(fileId);
+        return fileStore.exists(fileId);
     }
 
-    /**
-     * Get the size of a file by its ID without loading the content into memory
-     *
-     * @param fileId The ID of the file
-     * @return The size of the file in bytes
-     * @throws IOException If the file doesn't exist or can't be read
-     */
     public long getFileSize(String fileId) throws IOException {
-        Path filePath = getFilePath(fileId);
-
-        if (!Files.exists(filePath)) {
-            throw new IOException("File not found with ID: " + fileId);
-        }
-
-        return Files.size(filePath);
+        enforceOwnership(fileId);
+        return fileStore.size(fileId);
     }
 
-    /**
-     * Get the path for a file ID
-     *
-     * @param fileId The ID of the file
-     * @return The path to the file
-     * @throws IllegalArgumentException if fileId contains path traversal characters or resolves
-     *     outside base directory
-     */
-    private Path getFilePath(String fileId) {
-        // Validate fileId to prevent path traversal
-        if (fileId.contains("..") || fileId.contains("/") || fileId.contains("\\")) {
-            throw new IllegalArgumentException("Invalid file ID");
+    private String resolveOwner() {
+        String propagated = JobContext.getOwner();
+        if (propagated != null) {
+            return propagated;
         }
-
-        Path basePath = Path.of(tempDirPath).normalize().toAbsolutePath();
-        Path resolvedPath = basePath.resolve(fileId).normalize();
-
-        // Ensure resolved path is within the base directory
-        if (!resolvedPath.startsWith(basePath)) {
-            throw new IllegalArgumentException("File ID resolves to an invalid path");
-        }
-
-        return resolvedPath;
+        return jobOwnershipService.flatMap(JobOwnershipService::getCurrentUserId).orElse(null);
     }
 
-    /**
-     * Generate a unique file ID
-     *
-     * @return A unique file ID
-     */
-    private String generateFileId() {
-        return UUID.randomUUID().toString();
+    private void enforceOwnership(String fileId) {
+        if (jobOwnershipService.isEmpty()) {
+            return;
+        }
+        Optional<String> currentUser = jobOwnershipService.get().getCurrentUserId();
+        if (currentUser.isEmpty()) {
+            return;
+        }
+        String owner;
+        try {
+            owner = fileStore.getOwner(fileId);
+        } catch (IOException e) {
+            log.warn("Failed to read owner for file {}: {}", fileId, e.getMessage());
+            throw new SecurityException(
+                    "Access denied: could not verify ownership of the requested file");
+        }
+        if (owner == null) {
+            return;
+        }
+        if (!owner.equals(currentUser.get())) {
+            log.warn(
+                    "Access denied: user {} attempted to access file {} owned by {}",
+                    currentUser.get(),
+                    fileId,
+                    owner);
+            throw new SecurityException(
+                    "Access denied: you do not have permission to access this file");
+        }
     }
 }

app/common/src/main/java/stirling/software/common/service/InternalApiClient.java

@@ -50,6 +50,16 @@ public class InternalApiClient {
                     "^/api/v1/(general|misc|security|convert|filter)(/[A-Za-z0-9_-]+)+$"
                             + "|^/api/v1/ai/tools(/[A-Za-z0-9_-]+)+$");
 
+    /**
+     * Marker propagated on every internal sub-step dispatch so the saas PAYG interceptor classifies
+     * the call as {@code BillingCategory.AUTOMATION}. By construction every {@link
+     * InternalApiClient#post} caller is an automation surface (pipeline executor, AI workflow,
+     * policy runner) running a child tool inside a parent automation flow — see the saas {@code
+     * PaygChargeInterceptor.determineCategory} precedence chain, where this header dominates any
+     * per-tool {@code @RequiresFeature} annotation.
+     */
+    public static final String AUTOMATION_HEADER = "X-Stirling-Automation";
+
     private final ServletContext servletContext;
     private final UserServiceInterface userService;
     private final TempFileManager tempFileManager;
@@ -96,7 +106,23 @@ public class InternalApiClient {
         if (apiKey != null && !apiKey.isEmpty()) {
             headers.add("X-API-KEY", apiKey);
         }
+        // Tag the sub-step as automation so PAYG bills it under AUTOMATION regardless of which
+        // tool-level @RequiresFeature annotation the dispatched controller carries (e.g. an AI-OCR
+        // step inside a policy run must bill as AUTOMATION, not AI). Set unconditionally because
+        // every caller of this dispatcher is an automation surface by design.
+        headers.add(AUTOMATION_HEADER, "true");
 
+        // A no-file ai/tools call (e.g. create-pdf-from-html-agent) sends only string params, so
+        // without this RestTemplate would use urlencoded instead of the multipart the controller
+        // expects. File-bearing calls get the right multipart content-type from RestTemplate.
+        boolean isAiTool = endpointPath.startsWith("/api/v1/ai/tools/");
+        boolean hasFilePart =
+                body.values().stream()
+                        .flatMap(java.util.List::stream)
+                        .anyMatch(v -> v instanceof Resource);
+        if (isAiTool && !hasFilePart) {
+            headers.setContentType(MediaType.MULTIPART_FORM_DATA);
+        }
         HttpEntity<MultiValueMap<String, Object>> entity = new HttpEntity<>(body, headers);
         RequestCallback requestCallback = restTemplate.httpEntityCallback(entity, Resource.class);
 

app/common/src/main/java/stirling/software/common/service/JobExecutorService.java

@@ -57,85 +57,43 @@ public class JobExecutorService {
         this.resourceMonitor = resourceMonitor;
         this.jobQueue = jobQueue;
 
-        // Parse session timeout and calculate effective timeout once during initialization
         long sessionTimeoutMs = parseSessionTimeout(sessionTimeout);
         this.effectiveTimeoutMs = Math.min(asyncRequestTimeoutMs, sessionTimeoutMs);
         log.debug(
                 "Job executor configured with effective timeout of {} ms", this.effectiveTimeoutMs);
     }
 
-    /**
-     * Run a job either asynchronously or synchronously
-     *
-     * @param async Whether to run the job asynchronously
-     * @param work The work to be done
-     * @return The response
-     */
     public ResponseEntity<?> runJobGeneric(boolean async, Supplier<Object> work) {
         return runJobGeneric(async, work, -1);
     }
 
-    /**
-     * Run a job either asynchronously or synchronously with a custom timeout
-     *
-     * @param async Whether to run the job asynchronously
-     * @param work The work to be done
-     * @param customTimeoutMs Custom timeout in milliseconds, or -1 to use the default
-     * @return The response
-     */
     public ResponseEntity<?> runJobGeneric(
             boolean async, Supplier<Object> work, long customTimeoutMs) {
         return runJobGeneric(async, work, customTimeoutMs, false, 50);
     }
 
-    /**
-     * Run a job either asynchronously or synchronously with custom parameters
-     *
-     * @param async Whether to run the job asynchronously
-     * @param work The work to be done
-     * @param customTimeoutMs Custom timeout in milliseconds, or -1 to use the default
-     * @param queueable Whether this job can be queued when system resources are limited
-     * @param resourceWeight The resource weight of this job (1-100)
-     * @return The response
-     */
     public ResponseEntity<?> runJobGeneric(
             boolean async,
             Supplier<Object> work,
             long customTimeoutMs,
             boolean queueable,
             int resourceWeight) {
-        // Generate base UUID
         String baseJobId = UUID.randomUUID().toString();
-
-        // Scope job to authenticated user if security is enabled
         String scopedJobKey = getScopedJobKey(baseJobId);
 
         log.debug("Generated jobId: {} (base: {})", scopedJobKey, baseJobId);
 
-        // Store the scoped job ID in the request for potential use by other components
         if (request != null) {
             request.setAttribute("jobId", scopedJobKey);
-
-            // Also track this job ID in the user's session for authorization purposes
-            // This ensures users can only cancel their own jobs
-            if (request.getSession() != null) {
-                @SuppressWarnings("unchecked")
-                java.util.Set<String> userJobIds =
-                        (java.util.Set<String>) request.getSession().getAttribute("userJobIds");
-
-                if (userJobIds == null) {
-                    userJobIds = new java.util.concurrent.ConcurrentSkipListSet<>();
-                    request.getSession().setAttribute("userJobIds", userJobIds);
-                }
-
-                userJobIds.add(scopedJobKey);
-                log.debug("Added scoped job ID {} to user session", scopedJobKey);
-            }
         }
 
         String jobId = scopedJobKey;
 
-        // Determine which timeout to use
+        final String jobOwner =
+                jobOwnershipService != null
+                        ? jobOwnershipService.getCurrentUserId().orElse(null)
+                        : null;
+
         long timeoutToUse = customTimeoutMs > 0 ? customTimeoutMs : effectiveTimeoutMs;
 
         log.debug(
@@ -146,7 +104,6 @@ public class JobExecutorService {
                 queueable,
                 resourceWeight);
 
-        // Check if we need to queue this job based on resource availability
         boolean shouldQueue =
                 queueable
                         && async
@@ -154,7 +111,6 @@ public class JobExecutorService {
                         resourceMonitor.shouldQueueJob(resourceWeight);
 
         if (shouldQueue) {
-            // Queue the job instead of executing immediately
             log.debug(
                     "Queueing job {} due to resource constraints (weight: {})",
                     jobId,
@@ -162,18 +118,13 @@ public class JobExecutorService {
 
             taskManager.createTask(jobId);
 
-            // Create a specialized wrapper that updates the TaskManager
             final String capturedJobIdForQueue = jobId;
             Supplier<Object> wrappedWork =
                     () -> {
                         try {
-                            // Set jobId in ThreadLocal context for the queued job
                             stirling.software.common.util.JobContext.setJobId(
                                     capturedJobIdForQueue);
-                            log.debug(
-                                    "Set jobId {} in JobContext for queued job execution",
-                                    capturedJobIdForQueue);
-
+                            stirling.software.common.util.JobContext.setOwner(jobOwner);
                             Object result = work.get();
                             processJobResult(capturedJobIdForQueue, result);
                             return result;
@@ -186,21 +137,17 @@ public class JobExecutorService {
                             taskManager.setError(capturedJobIdForQueue, e.getMessage());
                             throw e;
                         } finally {
-                            // Clean up ThreadLocal to avoid memory leaks
                             stirling.software.common.util.JobContext.clear();
                         }
                     };
 
-            // Queue the job and get the future
             CompletableFuture<ResponseEntity<?>> future =
                     jobQueue.queueJob(jobId, resourceWeight, wrappedWork, timeoutToUse);
 
-            // Return immediately with job ID
             return ResponseEntity.ok().body(new JobResponse<>(true, jobId, null));
         } else if (async) {
             taskManager.createTask(jobId);
 
-            // Capture the jobId for the async thread
             final String capturedJobId = jobId;
 
             executor.execute(
@@ -211,13 +158,8 @@ public class JobExecutorService {
                                     capturedJobId,
                                     timeoutToUse);
 
-                            // Set jobId in ThreadLocal context for the async thread
                             stirling.software.common.util.JobContext.setJobId(capturedJobId);
-                            log.debug(
-                                    "Set jobId {} in JobContext for async execution",
-                                    capturedJobId);
-
-                            // Execute with timeout
+                            stirling.software.common.util.JobContext.setOwner(jobOwner);
                             Object result = executeWithTimeout(() -> work.get(), timeoutToUse);
                             processJobResult(capturedJobId, result);
                         } catch (TimeoutException te) {
@@ -227,7 +169,6 @@ public class JobExecutorService {
                             log.error("Error executing job {}: {}", jobId, e.getMessage(), e);
                             taskManager.setError(jobId, e.getMessage());
                         } finally {
-                            // Clean up ThreadLocal to avoid memory leaks
                             stirling.software.common.util.JobContext.clear();
                         }
                     });
@@ -237,27 +178,19 @@ public class JobExecutorService {
             try {
                 log.debug("Running sync job with timeout {} ms", timeoutToUse);
 
-                // Make jobId available to downstream components on the worker thread
                 stirling.software.common.util.JobContext.setJobId(jobId);
-                log.debug("Set jobId {} in JobContext for sync execution", jobId);
-
-                // Execute with timeout
                 Object result = executeWithTimeout(() -> work.get(), timeoutToUse);
 
-                // If the result is already a ResponseEntity, return it directly
                 if (result instanceof ResponseEntity) {
                     return (ResponseEntity<?>) result;
                 }
 
-                // Process different result types
                 return handleResultForSyncJob(result);
             } catch (TimeoutException te) {
                 log.error("Synchronous job timed out after {} ms", timeoutToUse);
                 return ResponseEntity.internalServerError()
                         .body(Map.of("error", "Job timed out after " + timeoutToUse + " ms"));
             } catch (RuntimeException e) {
-                // Check if this is a typed exception that should be handled by
-                // GlobalExceptionHandler (either directly or wrapped)
                 Throwable cause = e.getCause();
                 if (e instanceof IllegalArgumentException
                         || cause
@@ -267,16 +200,13 @@ public class JobExecutorService {
                                 instanceof
                                 stirling.software.common.util.ExceptionUtils
                                         .BaseValidationException) {
-                    // Rethrow so GlobalExceptionHandler can handle with proper HTTP status codes
                     throw e;
                 }
-                // Handle other RuntimeExceptions as generic errors
                 log.error("Error executing synchronous job: {}", e.getMessage(), e);
                 return ResponseEntity.internalServerError()
                         .body(Map.of("error", "Job failed: " + e.getMessage()));
             } catch (Exception e) {
                 log.error("Error executing synchronous job: {}", e.getMessage(), e);
-                // Construct a JSON error response
                 return ResponseEntity.internalServerError()
                         .body(Map.of("error", "Job failed: " + e.getMessage()));
             } finally {
@@ -285,23 +215,13 @@ public class JobExecutorService {
         }
     }
 
-    /**
-     * Process the result of an asynchronous job
-     *
-     * @param jobId The job ID
-     * @param result The result
-     */
     private void processJobResult(String jobId, Object result) {
         try {
             if (result instanceof byte[]) {
-                // Store byte array directly to disk to avoid double memory consumption
                 String fileId = fileStorage.storeBytes((byte[]) result, "result.pdf");
                 taskManager.setFileResult(
                         jobId, fileId, "result.pdf", MediaType.APPLICATION_PDF_VALUE);
                 log.debug("Stored byte[] result with fileId: {}", fileId);
-
-                // Let the byte array get collected naturally in the next GC cycle
-                // We don't need to force System.gc() which can be harmful
             } else if (result instanceof ResponseEntity) {
                 ResponseEntity<?> response = (ResponseEntity<?>) result;
                 Object body = response.getBody();
@@ -330,16 +250,13 @@ public class JobExecutorService {
                     taskManager.setFileResult(jobId, fileId, filename, contentType);
                     log.debug("Stored ResponseEntity<Resource> result with fileId: {}", fileId);
                 } else {
-                    // Check if the response body contains a fileId
                     if (body != null && body.toString().contains("fileId")) {
                         try {
-                            // Try to extract fileId using reflection
                             java.lang.reflect.Method getFileId =
                                     body.getClass().getMethod("getFileId");
                             String fileId = (String) getFileId.invoke(body);
 
                             if (fileId != null && !fileId.isEmpty()) {
-                                // Try to get filename and content type
                                 String filename = "result.pdf";
                                 String contentType = MediaType.APPLICATION_PDF_VALUE;
 
@@ -379,7 +296,6 @@ public class JobExecutorService {
                         }
                     }
 
-                    // Store generic result
                     taskManager.setResult(jobId, body);
                 }
             } else if (result instanceof MultipartFile file) {
@@ -388,16 +304,13 @@ public class JobExecutorService {
                         jobId, fileId, file.getOriginalFilename(), file.getContentType());
                 log.debug("Stored MultipartFile result with fileId: {}", fileId);
             } else {
-                // Check if result has a fileId field
                 if (result != null) {
                     try {
-                        // Try to extract fileId using reflection
                         java.lang.reflect.Method getFileId =
                                 result.getClass().getMethod("getFileId");
                         String fileId = (String) getFileId.invoke(result);
 
                         if (fileId != null && !fileId.isEmpty()) {
-                            // Try to get filename and content type
                             String filename = "result.pdf";
                             String contentType = MediaType.APPLICATION_PDF_VALUE;
 
@@ -435,7 +348,6 @@ public class JobExecutorService {
                     }
                 }
 
-                // Default case: store the result as is
                 taskManager.setResult(jobId, result);
             }
 
@@ -446,16 +358,8 @@ public class JobExecutorService {
         }
     }
 
-    /**
-     * Handle different result types for synchronous jobs
-     *
-     * @param result The result object
-     * @return The appropriate ResponseEntity
-     * @throws IOException If there is an error processing the result
-     */
     private ResponseEntity<?> handleResultForSyncJob(Object result) throws IOException {
         if (result instanceof byte[]) {
-            // Return byte array as PDF
             return ResponseEntity.ok()
                     .contentType(MediaType.APPLICATION_PDF)
                     .header(
@@ -463,7 +367,6 @@ public class JobExecutorService {
                             "form-data; name=\"attachment\"; filename=\"result.pdf\"")
                     .body(result);
         } else if (result instanceof MultipartFile file) {
-            // Return MultipartFile content
             return ResponseEntity.ok()
                     .contentType(MediaType.parseMediaType(file.getContentType()))
                     .header(
@@ -473,7 +376,6 @@ public class JobExecutorService {
                                     + "\"")
                     .body(file.getBytes());
         } else {
-            // Default case: return as JSON
             return ResponseEntity.ok(result);
         }
     }
@@ -493,15 +395,9 @@ public class JobExecutorService {
         return mediaType != null ? mediaType.toString() : MediaType.APPLICATION_PDF_VALUE;
     }
 
-    /**
-     * Parse session timeout string (e.g., "30m", "1h") to milliseconds
-     *
-     * @param timeout The timeout string
-     * @return The timeout in milliseconds
-     */
     private long parseSessionTimeout(String timeout) {
         if (timeout == null || timeout.isEmpty()) {
-            return 30 * 60 * 1000; // Default: 30 minutes
+            return 30 * 60 * 1000;
         }
 
         try {
@@ -523,27 +419,16 @@ public class JobExecutorService {
                 case "m" -> (long) (numericValue * 60 * 1000);
                 case "h" -> (long) (numericValue * 60 * 60 * 1000);
                 case "d" -> (long) (numericValue * 24 * 60 * 60 * 1000);
-                default -> (long) (numericValue * 60 * 1000); // Default to minutes
+                default -> (long) (numericValue * 60 * 1000);
             };
         } catch (Exception e) {
             log.warn("Could not parse session timeout '{}', using default", timeout);
-            return 30 * 60 * 1000; // Default: 30 minutes
+            return 30 * 60 * 1000;
         }
     }
 
-    /**
-     * Execute a supplier with a timeout
-     *
-     * @param supplier The supplier to execute
-     * @param timeoutMs The timeout in milliseconds
-     * @return The result from the supplier
-     * @throws TimeoutException If the execution times out
-     * @throws Exception If the supplier throws an exception
-     */
     private <T> T executeWithTimeout(Supplier<T> supplier, long timeoutMs)
             throws TimeoutException, Exception {
-        // Use the same executor as other async jobs for consistency
-        // This ensures all operations run on the same thread pool
         String currentJobId = stirling.software.common.util.JobContext.getJobId();
 
         java.util.concurrent.CompletableFuture<T> future =
@@ -577,17 +462,10 @@ public class JobExecutorService {
         }
     }
 
-    /**
-     * Get a scoped job key that includes user ownership when security is enabled.
-     *
-     * @param baseJobId the base job identifier
-     * @return scoped job key, or just baseJobId if no ownership service available
-     */
     private String getScopedJobKey(String baseJobId) {
         if (jobOwnershipService != null) {
             return jobOwnershipService.createScopedJobKey(baseJobId);
         }
-        // Security disabled, return unsecured job key
         return baseJobId;
     }
 }

app/common/src/main/java/stirling/software/common/service/MobileScannerService.java

@@ -3,7 +3,6 @@ package stirling.software.common.service;
 import java.io.IOException;
 import java.nio.file.Files;
 import java.nio.file.Path;
-import java.nio.file.Paths;
 import java.util.ArrayList;
 import java.util.HashMap;
 import java.util.List;
@@ -35,7 +34,7 @@ public class MobileScannerService {
     public MobileScannerService() throws IOException {
         // Create temp directory for mobile scanner uploads
         this.tempDirectory =
-                Paths.get(System.getProperty("java.io.tmpdir"), "stirling-mobile-scanner");
+                Path.of(System.getProperty("java.io.tmpdir"), "stirling-mobile-scanner");
         Files.createDirectories(tempDirectory);
         log.info("Mobile scanner temp directory: {}", tempDirectory);
     }

app/common/src/main/java/stirling/software/common/service/PostHogService.java

@@ -8,7 +8,7 @@ import java.lang.management.OperatingSystemMXBean;
 import java.lang.management.RuntimeMXBean;
 import java.lang.management.ThreadMXBean;
 import java.nio.file.Files;
-import java.nio.file.Paths;
+import java.nio.file.Path;
 import java.util.HashMap;
 import java.util.Locale;
 import java.util.Map;
@@ -160,7 +160,7 @@ public class PostHogService {
     }
 
     private boolean isRunningInDocker() {
-        return Files.exists(Paths.get("/.dockerenv"));
+        return Files.exists(Path.of("/.dockerenv"));
     }
 
     private Map<String, Object> getDockerMetrics() {

app/common/src/main/java/stirling/software/common/service/TaskManager.java

@@ -3,9 +3,13 @@ package stirling.software.common.service;
 import java.io.BufferedInputStream;
 import java.io.IOException;
 import java.io.InputStream;
+import java.time.Duration;
+import java.time.Instant;
 import java.time.LocalDateTime;
+import java.time.ZoneId;
 import java.time.temporal.ChronoUnit;
 import java.util.ArrayList;
+import java.util.HashMap;
 import java.util.List;
 import java.util.Locale;
 import java.util.Map;
@@ -16,6 +20,7 @@ import java.util.concurrent.TimeUnit;
 import java.util.zip.ZipEntry;
 import java.util.zip.ZipInputStream;
 
+import org.springframework.beans.factory.annotation.Autowired;
 import org.springframework.beans.factory.annotation.Value;
 import org.springframework.http.MediaType;
 import org.springframework.stereotype.Service;
@@ -26,6 +31,10 @@ import jakarta.annotation.PreDestroy;
 
 import lombok.extern.slf4j.Slf4j;
 
+import stirling.software.common.cluster.ClusterBackplane;
+import stirling.software.common.cluster.JobStore;
+import stirling.software.common.cluster.JobStoreEntry;
+import stirling.software.common.cluster.JobStoreEntry.JobState;
 import stirling.software.common.model.job.JobResult;
 import stirling.software.common.model.job.JobStats;
 import stirling.software.common.model.job.ResultFile;
@@ -40,20 +49,20 @@ public class TaskManager {
     private int jobResultExpiryMinutes = 30;
 
     private final FileStorage fileStorage;
+    private final JobStore jobStore;
+    private final ClusterBackplane clusterBackplane;
     private final ScheduledExecutorService cleanupExecutor =
             Executors.newSingleThreadScheduledExecutor(
                     Thread.ofVirtual().name("task-cleanup-", 0).factory());
 
-    /** Initialize the task manager and start the cleanup scheduler */
-    public TaskManager(FileStorage fileStorage) {
+    @Autowired
+    public TaskManager(
+            FileStorage fileStorage, JobStore jobStore, ClusterBackplane clusterBackplane) {
         this.fileStorage = fileStorage;
+        this.jobStore = jobStore;
+        this.clusterBackplane = clusterBackplane;
 
-        // Schedule periodic cleanup of old job results
-        cleanupExecutor.scheduleAtFixedRate(
-                this::cleanupOldJobs,
-                10, // Initial delay
-                10, // Interval
-                TimeUnit.MINUTES);
+        cleanupExecutor.scheduleAtFixedRate(this::cleanupOldJobs, 10, 10, TimeUnit.MINUTES);
 
         log.debug(
                 "Task manager initialized with job result expiry of {} minutes",
@@ -66,7 +75,9 @@ public class TaskManager {
      * @param jobId The job ID
      */
     public void createTask(String jobId) {
-        jobResults.put(jobId, JobResult.createNew(jobId));
+        JobResult result = JobResult.createNew(jobId);
+        jobResults.put(jobId, result);
+        writeThrough(jobId, result);
         log.debug("Created task with job ID: {}", jobId);
     }
 
@@ -79,6 +90,7 @@ public class TaskManager {
     public void setResult(String jobId, Object result) {
         JobResult jobResult = getOrCreateJobResult(jobId);
         jobResult.completeWithResult(result);
+        writeThrough(jobId, jobResult);
         log.debug("Set result for job ID: {}", jobId);
     }
 
@@ -101,6 +113,7 @@ public class TaskManager {
                         extractZipToIndividualFiles(fileId, originalFileName);
                 if (!extractedFiles.isEmpty()) {
                     jobResult.completeWithFiles(extractedFiles);
+                    writeThrough(jobId, jobResult);
                     log.debug(
                             "Set multiple file results for job ID: {} with {} files extracted from"
                                     + " ZIP",
@@ -127,6 +140,7 @@ public class TaskManager {
                     "Failed to get file size for job {}: {}. Using size 0.", jobId, e.getMessage());
             jobResult.completeWithSingleFile(fileId, originalFileName, contentType, 0);
         }
+        writeThrough(jobId, jobResult);
     }
 
     /**
@@ -138,6 +152,7 @@ public class TaskManager {
     public void setMultipleFileResults(String jobId, List<ResultFile> resultFiles) {
         JobResult jobResult = getOrCreateJobResult(jobId);
         jobResult.completeWithFiles(resultFiles);
+        writeThrough(jobId, jobResult);
         log.debug(
                 "Set multiple file results for job ID: {} with {} files",
                 jobId,
@@ -153,6 +168,7 @@ public class TaskManager {
     public void setError(String jobId, String error) {
         JobResult jobResult = getOrCreateJobResult(jobId);
         jobResult.failWithError(error);
+        writeThrough(jobId, jobResult);
         log.debug("Set error for job ID: {}: {}", jobId, error);
     }
 
@@ -169,6 +185,7 @@ public class TaskManager {
             // If no result or error has been set, mark it as complete with an empty result
             jobResult.completeWithResult("Task completed successfully");
         }
+        writeThrough(jobId, jobResult);
         log.debug("Marked job ID: {} as complete", jobId);
     }
 
@@ -205,6 +222,7 @@ public class TaskManager {
         JobResult jobResult = jobResults.get(jobId);
         if (jobResult != null) {
             jobResult.addNote(note);
+            writeThrough(jobId, jobResult);
             log.debug("Added note to job ID: {}: {}", jobId, note);
             return true;
         }
@@ -295,8 +313,11 @@ public class TaskManager {
         return jobResults.computeIfAbsent(jobId, JobResult::createNew);
     }
 
-    /** Clean up old completed job results */
+    /** Clean up old completed job results. No-op in cluster mode; the backplane TTL owns expiry. */
     public void cleanupOldJobs() {
+        if (clusterBackplane != null && !clusterBackplane.shouldRunLocalCleanup()) {
+            return;
+        }
         LocalDateTime expiryThreshold =
                 LocalDateTime.now().minus(jobResultExpiryMinutes, ChronoUnit.MINUTES);
         int removedCount = 0;
@@ -315,6 +336,9 @@ public class TaskManager {
 
                     // Remove the job result
                     jobResults.remove(entry.getKey());
+                    if (jobStore != null) {
+                        jobStore.delete(entry.getKey());
+                    }
                     removedCount++;
                 }
             }
@@ -327,6 +351,53 @@ public class TaskManager {
         }
     }
 
+    /** Mirror the in-memory {@code JobResult} into the cluster-visible {@link JobStore}. */
+    private void writeThrough(String jobId, JobResult result) {
+        if (jobStore == null) {
+            return;
+        }
+        try {
+            jobStore.put(toEntry(jobId, result), Duration.ofMinutes(jobResultExpiryMinutes));
+        } catch (RuntimeException ex) {
+            log.warn("JobStore write-through failed for job {}: {}", jobId, ex.getMessage());
+        }
+    }
+
+    private JobStoreEntry toEntry(String jobId, JobResult result) {
+        JobState state;
+        if (result.isComplete()) {
+            state = result.getError() != null ? JobState.FAILED : JobState.COMPLETE;
+        } else {
+            state = JobState.PENDING;
+        }
+        Instant createdAt = toInstant(result.getCreatedAt());
+        Instant completedAt = toInstant(result.getCompletedAt());
+        List<String> fileIds = new ArrayList<>();
+        if (result.hasFiles()) {
+            for (ResultFile rf : result.getAllResultFiles()) {
+                fileIds.add(rf.getFileId());
+            }
+        }
+        Map<String, String> meta = new HashMap<>();
+        if (result.getNotes() != null && !result.getNotes().isEmpty()) {
+            meta.put("notesCount", Integer.toString(result.getNotes().size()));
+        }
+        String owningNodeId = clusterBackplane == null ? "local" : clusterBackplane.localNodeId();
+        return new JobStoreEntry(
+                jobId,
+                state,
+                owningNodeId,
+                createdAt,
+                completedAt,
+                result.getError(),
+                fileIds,
+                meta);
+    }
+
+    private Instant toInstant(LocalDateTime ldt) {
+        return ldt == null ? null : ldt.atZone(ZoneId.systemDefault()).toInstant();
+    }
+
     /** Shutdown the cleanup executor */
     @PreDestroy
     public void shutdown() {
@@ -370,7 +441,7 @@ public class TaskManager {
             while ((entry = zipIn.getNextEntry()) != null) {
                 if (!entry.isDirectory()) {
                     String contentType = determineContentType(entry.getName());
-                    // storeInputStream returns the fileId and byte count — no extra stat needed
+                    // storeInputStream returns the fileId and byte count - no extra stat needed
                     FileStorage.StoredFile stored =
                             fileStorage.storeInputStream(zipIn, entry.getName());
 
@@ -458,7 +529,8 @@ public class TaskManager {
     }
 
     /**
-     * Find the job key that owns a given file ID.
+     * Find the job key that owns a given file ID. Checks the local in-memory map first, then falls
+     * back to the cluster-visible {@link JobStore}.
      *
      * @param fileId file identifier to look up
      * @return scoped job key if found, otherwise null
@@ -474,6 +546,18 @@ public class TaskManager {
                 }
             }
         }
+        if (jobStore != null) {
+            // Propagate JobStore failures: returning null on a backplane outage would conflate
+            // "no such file" with "lookup unavailable" and the caller would respond 404 to a
+            // transient blip that should be retried. Let Spring's exception handler surface a
+            // 5xx so clients know to retry.
+            try {
+                return jobStore.findJobIdByFileId(fileId).orElse(null);
+            } catch (RuntimeException e) {
+                log.warn("JobStore findJobIdByFileId failed for {}: {}", fileId, e.getMessage());
+                throw e;
+            }
+        }
         return null;
     }
 }

app/common/src/main/java/stirling/software/common/service/ToolMetadataService.java

@@ -1,11 +1,21 @@
 package stirling.software.common.service;
 
+import java.util.List;
+
 /** Provides metadata about tool endpoints for internal dispatch. */
 public interface ToolMetadataService {
 
     /** Returns true if the given operation path accepts multiple input files. */
     boolean isMultiInput(String operationPath);
 
+    /**
+     * Returns the file extensions (lowercase, no leading dot, e.g. {@code "pdf"}) that the
+     * operation accepts as input ({@code output=false}) or produces as output ({@code
+     * output=true}), derived from the endpoint's declared type. Returns {@code null} when the
+     * endpoint declares no specific type, which callers should treat as "any type accepted".
+     */
+    List<String> getExtensionTypes(boolean output, String operationPath);
+
     /**
      * Returns true when the endpoint's ZIP response is a transport for multiple typed results and
      * should be unpacked: multi-output endpoints (Type:SIMO / Type:MIMO) and wrapper declarations

app/common/src/main/java/stirling/software/common/util/GeneralUtils.java

@@ -255,7 +255,7 @@ public class GeneralUtils {
         String pattern = locationPattern;
         if (pattern.startsWith("file:")) {
             String rawPath = pattern.substring(5).replace("\\*", "").replace("/*", "");
-            Path normalizePath = Paths.get(rawPath).normalize();
+            Path normalizePath = Path.of(rawPath).normalize();
             pattern = "file:" + normalizePath.toString().replace("\\", "/") + "/*";
         }
         return ResourcePatternUtils.getResourcePatternResolver(resourceLoader)
@@ -837,7 +837,7 @@ public class GeneralUtils {
     }
 
     public boolean createDir(String path) {
-        Path folder = Paths.get(path);
+        Path folder = Path.of(path);
         if (!Files.exists(folder)) {
             try {
                 Files.createDirectories(folder);
@@ -867,7 +867,7 @@ public class GeneralUtils {
 
     public void saveKeyToSettings(String key, Object newValue) throws IOException {
         String[] keyArray = key.split("\\.");
-        Path settingsPath = Paths.get(InstallationPathConfig.getSettingsPath());
+        Path settingsPath = Path.of(InstallationPathConfig.getSettingsPath());
         YamlHelper settingsYaml = new YamlHelper(settingsPath);
         settingsYaml.updateValue(Arrays.asList(keyArray), newValue);
         settingsYaml.saveOverride(settingsPath);
@@ -888,7 +888,7 @@ public class GeneralUtils {
             return;
         }
 
-        Path settingsPath = Paths.get(InstallationPathConfig.getSettingsPath());
+        Path settingsPath = Path.of(InstallationPathConfig.getSettingsPath());
         YamlHelper settingsYaml = new YamlHelper(settingsPath);
 
         // Apply all updates to the same YamlHelper instance
@@ -974,11 +974,11 @@ public class GeneralUtils {
      */
     public void extractPipeline() throws IOException {
         Path pipelineDir =
-                Paths.get(InstallationPathConfig.getPipelinePath(), DEFAULT_WEBUI_CONFIGS_DIR);
+                Path.of(InstallationPathConfig.getPipelinePath(), DEFAULT_WEBUI_CONFIGS_DIR);
         Files.createDirectories(pipelineDir);
 
         for (String name : DEFAULT_VALID_PIPELINE) {
-            if (!Paths.get(name).getFileName().toString().equals(name)) {
+            if (!Path.of(name).getFileName().toString().equals(name)) {
                 log.error("Invalid pipeline file name: {}", name);
                 throw new IllegalArgumentException("Invalid pipeline file name: " + name);
             }
@@ -1014,7 +1014,7 @@ public class GeneralUtils {
             throw new IllegalArgumentException(
                     "scriptName must not contain path traversal characters");
         }
-        if (!Paths.get(scriptName).getFileName().toString().equals(scriptName)) {
+        if (!Path.of(scriptName).getFileName().toString().equals(scriptName)) {
             throw new IllegalArgumentException(
                     "scriptName must not contain path traversal characters");
         }
@@ -1024,7 +1024,7 @@ public class GeneralUtils {
                     "scriptName must be either 'png_to_webp.py' or 'split_photos.py'");
         }
 
-        Path scriptsDir = Paths.get(InstallationPathConfig.getScriptsPath(), PYTHON_SCRIPTS_DIR);
+        Path scriptsDir = Path.of(InstallationPathConfig.getScriptsPath(), PYTHON_SCRIPTS_DIR);
         Files.createDirectories(scriptsDir);
 
         Path target = scriptsDir.resolve(scriptName);

app/common/src/main/java/stirling/software/common/util/JarPathUtil.java

@@ -4,7 +4,6 @@ import java.io.File;
 import java.net.URISyntaxException;
 import java.nio.file.Files;
 import java.nio.file.Path;
-import java.nio.file.Paths;
 
 import lombok.extern.slf4j.Slf4j;
 
@@ -20,7 +19,7 @@ public class JarPathUtil {
     public static Path currentJar() {
         try {
             Path jar =
-                    Paths.get(
+                    Path.of(
                                     JarPathUtil.class
                                             .getProtectionDomain()
                                             .getCodeSource()
@@ -61,14 +60,14 @@ public class JarPathUtil {
         }
 
         // Location 2: ./build/libs/ (development build)
-        possibleLocations[1] = Paths.get("build", "libs", "restart-helper.jar").toAbsolutePath();
+        possibleLocations[1] = Path.of("build", "libs", "restart-helper.jar").toAbsolutePath();
 
         // Location 3: app/common/build/libs/ (multi-module build)
         possibleLocations[2] =
-                Paths.get("app", "common", "build", "libs", "restart-helper.jar").toAbsolutePath();
+                Path.of("app", "common", "build", "libs", "restart-helper.jar").toAbsolutePath();
 
         // Location 4: Current working directory
-        possibleLocations[3] = Paths.get("restart-helper.jar").toAbsolutePath();
+        possibleLocations[3] = Path.of("restart-helper.jar").toAbsolutePath();
 
         // Check each location
         for (Path location : possibleLocations) {

app/common/src/main/java/stirling/software/common/util/JobContext.java

@@ -1,8 +1,9 @@
 package stirling.software.common.util;
 
-/** Thread-local context for passing job ID across async boundaries */
+/** Thread-local context for passing job ID and owner across async boundaries */
 public class JobContext {
     private static final ThreadLocal<String> CURRENT_JOB_ID = new ThreadLocal<>();
+    private static final ThreadLocal<String> CURRENT_OWNER = new ThreadLocal<>();
 
     public static void setJobId(String jobId) {
         CURRENT_JOB_ID.set(jobId);
@@ -12,7 +13,16 @@ public class JobContext {
         return CURRENT_JOB_ID.get();
     }
 
+    public static void setOwner(String owner) {
+        CURRENT_OWNER.set(owner);
+    }
+
+    public static String getOwner() {
+        return CURRENT_OWNER.get();
+    }
+
     public static void clear() {
         CURRENT_JOB_ID.remove();
+        CURRENT_OWNER.remove();
     }
 }

app/common/src/main/java/stirling/software/common/util/OfficeDocumentSanitizer.java

@@ -0,0 +1,310 @@
+package stirling.software.common.util;
+
+import java.io.ByteArrayInputStream;
+import java.io.ByteArrayOutputStream;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Locale;
+import java.util.Set;
+import java.util.zip.ZipEntry;
+import java.util.zip.ZipInputStream;
+import java.util.zip.ZipOutputStream;
+
+import javax.xml.XMLConstants;
+import javax.xml.parsers.DocumentBuilder;
+import javax.xml.parsers.DocumentBuilderFactory;
+import javax.xml.parsers.ParserConfigurationException;
+import javax.xml.transform.OutputKeys;
+import javax.xml.transform.Transformer;
+import javax.xml.transform.TransformerException;
+import javax.xml.transform.TransformerFactory;
+import javax.xml.transform.dom.DOMSource;
+import javax.xml.transform.stream.StreamResult;
+
+import org.springframework.stereotype.Component;
+import org.w3c.dom.Document;
+import org.w3c.dom.Element;
+import org.w3c.dom.NamedNodeMap;
+import org.w3c.dom.Node;
+import org.w3c.dom.NodeList;
+import org.xml.sax.SAXException;
+
+import io.github.pixee.security.ZipSecurity;
+
+import lombok.extern.slf4j.Slf4j;
+
+import stirling.software.common.model.ApplicationProperties;
+import stirling.software.common.service.SsrfProtectionService;
+
+// Strips external refs from OOXML/ODF uploads so LibreOffice can't be made to fetch them.
+@Component
+@Slf4j
+public class OfficeDocumentSanitizer {
+
+    private static final Set<String> OOXML_EXTENSIONS =
+            Set.of(
+                    "docx", "docm", "dotx", "dotm", "xlsx", "xlsm", "xltx", "xltm", "pptx", "pptm",
+                    "potx", "potm", "ppsx", "ppsm");
+
+    private static final Set<String> ODF_EXTENSIONS =
+            Set.of(
+                    "odt", "ott", "ods", "ots", "odp", "otp", "odg", "otg", "odf", "odc", "odi",
+                    "odm");
+
+    private static final Set<String> ODF_XML_PARTS =
+            Set.of("content.xml", "styles.xml", "meta.xml", "settings.xml");
+
+    private final SsrfProtectionService ssrfProtectionService;
+    private final ApplicationProperties applicationProperties;
+
+    public OfficeDocumentSanitizer(
+            SsrfProtectionService ssrfProtectionService,
+            ApplicationProperties applicationProperties) {
+        this.ssrfProtectionService = ssrfProtectionService;
+        this.applicationProperties = applicationProperties;
+    }
+
+    public boolean isSanitizableExtension(String extension) {
+        if (extension == null) {
+            return false;
+        }
+        String lower = extension.toLowerCase(Locale.ROOT);
+        return OOXML_EXTENSIONS.contains(lower) || ODF_EXTENSIONS.contains(lower);
+    }
+
+    public byte[] sanitize(byte[] documentBytes, String extension) throws IOException {
+        if (documentBytes == null || documentBytes.length == 0) {
+            throw new IOException("Office document input is empty or null");
+        }
+        if (applicationProperties.getSystem().isDisableSanitize()) {
+            log.debug("Office document sanitization disabled by configuration");
+            return documentBytes;
+        }
+        if (!isSanitizableExtension(extension)) {
+            return documentBytes;
+        }
+
+        ByteArrayOutputStream out = new ByteArrayOutputStream(documentBytes.length);
+        try (ZipInputStream zipIn =
+                        ZipSecurity.createHardenedInputStream(
+                                new ByteArrayInputStream(documentBytes));
+                ZipOutputStream zipOut = new ZipOutputStream(out)) {
+
+            ZipEntry entry;
+            while ((entry = zipIn.getNextEntry()) != null) {
+                String name = entry.getName();
+                byte[] bytes = entry.isDirectory() ? new byte[0] : zipIn.readAllBytes();
+
+                if (!entry.isDirectory()) {
+                    bytes = sanitizeEntry(name, bytes);
+                }
+
+                ZipEntry outEntry = new ZipEntry(name);
+                if (entry.getComment() != null) {
+                    outEntry.setComment(entry.getComment());
+                }
+                if (entry.getExtra() != null) {
+                    outEntry.setExtra(entry.getExtra());
+                }
+                zipOut.putNextEntry(outEntry);
+                if (!entry.isDirectory()) {
+                    zipOut.write(bytes);
+                }
+                zipOut.closeEntry();
+            }
+        }
+        return out.toByteArray();
+    }
+
+    private byte[] sanitizeEntry(String entryName, byte[] entryBytes) {
+        String lower = entryName.toLowerCase(Locale.ROOT);
+        try {
+            if (lower.endsWith(".rels")) {
+                return sanitizeOoxmlRels(entryBytes);
+            }
+            if (isOdfXmlPart(lower)) {
+                return sanitizeOdfXml(entryBytes);
+            }
+        } catch (ParserConfigurationException
+                | SAXException
+                | IOException
+                | TransformerException e) {
+            log.warn(
+                    "Failed to parse XML part '{}' for sanitization, leaving as-is: {}",
+                    entryName,
+                    e.getMessage());
+        }
+        return entryBytes;
+    }
+
+    private boolean isOdfXmlPart(String lowerName) {
+        int slash = lowerName.lastIndexOf('/');
+        String base = slash >= 0 ? lowerName.substring(slash + 1) : lowerName;
+        return ODF_XML_PARTS.contains(base);
+    }
+
+    private byte[] sanitizeOoxmlRels(byte[] xmlBytes)
+            throws IOException, ParserConfigurationException, SAXException, TransformerException {
+        Document doc = parseSecurely(xmlBytes);
+        Element root = doc.getDocumentElement();
+        if (root == null) {
+            return xmlBytes;
+        }
+        NodeList relationships = root.getElementsByTagNameNS("*", "Relationship");
+        List<Node> toRemove = new ArrayList<>();
+        for (int i = 0; i < relationships.getLength(); i++) {
+            Node node = relationships.item(i);
+            NamedNodeMap attrs = node.getAttributes();
+            if (attrs == null) {
+                continue;
+            }
+            Node targetMode = attrs.getNamedItem("TargetMode");
+            if (targetMode == null || !"external".equalsIgnoreCase(targetMode.getNodeValue())) {
+                continue;
+            }
+            Node target = attrs.getNamedItem("Target");
+            String targetValue = target == null ? "" : target.getNodeValue();
+            if (isAdminAllowed(targetValue)) {
+                continue;
+            }
+            log.warn(
+                    "Stripping OOXML external relationship target: {}",
+                    truncateForLog(targetValue));
+            toRemove.add(node);
+        }
+        if (toRemove.isEmpty()) {
+            return xmlBytes;
+        }
+        for (Node n : toRemove) {
+            n.getParentNode().removeChild(n);
+        }
+        return serializeDocument(doc);
+    }
+
+    private byte[] sanitizeOdfXml(byte[] xmlBytes)
+            throws IOException, ParserConfigurationException, SAXException, TransformerException {
+        Document doc = parseSecurely(xmlBytes);
+        Element root = doc.getDocumentElement();
+        if (root == null) {
+            return xmlBytes;
+        }
+        boolean modified = stripExternalHrefs(root);
+        if (!modified) {
+            return xmlBytes;
+        }
+        return serializeDocument(doc);
+    }
+
+    private boolean stripExternalHrefs(Node node) {
+        boolean modified = false;
+        if (node.getNodeType() == Node.ELEMENT_NODE) {
+            NamedNodeMap attrs = node.getAttributes();
+            List<String> hrefAttrsToRemove = new ArrayList<>();
+            for (int i = 0; i < attrs.getLength(); i++) {
+                Node attr = attrs.item(i);
+                String name = attr.getNodeName();
+                if (name == null) {
+                    continue;
+                }
+                String lower = name.toLowerCase(Locale.ROOT);
+                if (!(lower.equals("xlink:href")
+                        || lower.endsWith(":href")
+                        || lower.equals("href"))) {
+                    continue;
+                }
+                String value = attr.getNodeValue();
+                if (!isExternalUrl(value)) {
+                    continue;
+                }
+                if (isAdminAllowed(value)) {
+                    continue;
+                }
+                log.warn(
+                        "Stripping ODF external href attribute ({}): {}",
+                        name,
+                        truncateForLog(value));
+                hrefAttrsToRemove.add(name);
+            }
+            Element element = (Element) node;
+            for (String attrName : hrefAttrsToRemove) {
+                element.removeAttribute(attrName);
+                modified = true;
+            }
+        }
+        NodeList children = node.getChildNodes();
+        for (int i = 0; i < children.getLength(); i++) {
+            if (stripExternalHrefs(children.item(i))) {
+                modified = true;
+            }
+        }
+        return modified;
+    }
+
+    private boolean isExternalUrl(String url) {
+        if (url == null) {
+            return false;
+        }
+        String trimmed = url.trim().toLowerCase(Locale.ROOT);
+        if (trimmed.isEmpty() || trimmed.startsWith("#") || trimmed.startsWith("../")) {
+            return false;
+        }
+        return trimmed.startsWith("http://")
+                || trimmed.startsWith("https://")
+                || trimmed.startsWith("ftp://")
+                || trimmed.startsWith("ftps://")
+                || trimmed.startsWith("file:")
+                || trimmed.startsWith("smb:")
+                || trimmed.startsWith("\\\\")
+                || trimmed.startsWith("//");
+    }
+
+    // Preserved only with an explicit allowedDomains entry; MEDIUM default would admit public URLs.
+    private boolean isAdminAllowed(String url) {
+        if (ssrfProtectionService == null || url == null || url.isBlank()) {
+            return false;
+        }
+        ApplicationProperties.Html.UrlSecurity config =
+                applicationProperties.getSystem().getHtml().getUrlSecurity();
+        if (config == null
+                || config.getAllowedDomains() == null
+                || config.getAllowedDomains().isEmpty()) {
+            return false;
+        }
+        return ssrfProtectionService.isUrlAllowed(url);
+    }
+
+    private Document parseSecurely(byte[] xmlBytes)
+            throws ParserConfigurationException, SAXException, IOException {
+        DocumentBuilderFactory factory = DocumentBuilderFactory.newInstance();
+        factory.setFeature(XMLConstants.FEATURE_SECURE_PROCESSING, true);
+        factory.setFeature("http://apache.org/xml/features/disallow-doctype-decl", true);
+        factory.setFeature("http://xml.org/sax/features/external-general-entities", false);
+        factory.setFeature("http://xml.org/sax/features/external-parameter-entities", false);
+        factory.setFeature("http://apache.org/xml/features/nonvalidating/load-external-dtd", false);
+        factory.setXIncludeAware(false);
+        factory.setExpandEntityReferences(false);
+        factory.setNamespaceAware(true);
+        DocumentBuilder builder = factory.newDocumentBuilder();
+        return builder.parse(new ByteArrayInputStream(xmlBytes));
+    }
+
+    private byte[] serializeDocument(Document doc) throws TransformerException {
+        TransformerFactory tf = TransformerFactory.newInstance();
+        tf.setFeature(XMLConstants.FEATURE_SECURE_PROCESSING, true);
+        Transformer transformer = tf.newTransformer();
+        transformer.setOutputProperty(OutputKeys.ENCODING, "UTF-8");
+        transformer.setOutputProperty(OutputKeys.INDENT, "no");
+        transformer.setOutputProperty(OutputKeys.OMIT_XML_DECLARATION, "no");
+        ByteArrayOutputStream baos = new ByteArrayOutputStream();
+        transformer.transform(new DOMSource(doc), new StreamResult(baos));
+        return baos.toByteArray();
+    }
+
+    private String truncateForLog(String value) {
+        if (value == null) {
+            return "null";
+        }
+        return value.length() > 80 ? value.substring(0, 80) + "..." : value;
+    }
+}

app/common/src/main/java/stirling/software/common/util/RequestUriUtils.java

@@ -83,7 +83,16 @@ public class RequestUriUtils {
             return false;
         }
 
-        // Blocklist of backend/non-frontend paths that should still go through filters
+        // Blocklist of backend/non-frontend paths that should still go through filters.
+        //
+        // `/files` was historically a backend route; it is now a frontend route
+        // owned by HomePage / FileManagerView. Direct-nav or refresh on /files
+        // (or /files/<folder-uuid>) was returning the Spring auth filter's 401
+        // JSON instead of serving index.html, so the SPA never got a chance to
+        // mount and the user saw a raw error response. There are no `/files`
+        // backend mappings at the servlet root - the real storage endpoints
+        // live under `/api/v1/storage/files`, which is filtered out a few lines
+        // up by the `startsWith("/api/")` guard.
         String[] backendOnlyPrefixes = {
             "/register",
             "/pipeline",
@@ -91,7 +100,6 @@ public class RequestUriUtils {
             "/pdfjs-legacy",
             "/fonts",
             "/images",
-            "/files",
             "/css",
             "/js",
             "/swagger",
@@ -181,7 +189,7 @@ public class RequestUriUtils {
                 || trimmedUri.startsWith(
                         "/api/v1/mobile-scanner/") // Mobile scanner endpoints (no auth)
                 || trimmedUri.startsWith("/v1/api-docs")
-                // Workflow participant endpoints — access controlled by share tokens, not login
+                // Workflow participant endpoints - access controlled by share tokens, not login
                 || trimmedUri.startsWith("/api/v1/workflow/participant/")
                 // Share-link SPA bootstrap; data APIs remain protected
                 || trimmedUri.matches("^/share/[^/]+/?$");

app/common/src/main/java/stirling/software/common/util/propertyeditor/JsonListPropertyEditor.java

@@ -0,0 +1,48 @@
+package stirling.software.common.util.propertyeditor;
+
+import java.beans.PropertyEditorSupport;
+import java.util.ArrayList;
+import java.util.List;
+
+import lombok.extern.slf4j.Slf4j;
+
+import tools.jackson.core.type.TypeReference;
+import tools.jackson.databind.DeserializationFeature;
+import tools.jackson.databind.ObjectMapper;
+import tools.jackson.databind.json.JsonMapper;
+
+/**
+ * Binds a multipart form value containing a JSON array into a typed {@code List<T>}. Used for
+ * endpoints that accept structured list parameters via {@code @ModelAttribute} — the form field
+ * carries the full JSON array as its value and the editor parses it once.
+ */
+@Slf4j
+public class JsonListPropertyEditor<T> extends PropertyEditorSupport {
+
+    private static final ObjectMapper OBJECT_MAPPER =
+            JsonMapper.builder()
+                    .enable(DeserializationFeature.ACCEPT_SINGLE_VALUE_AS_ARRAY)
+                    .disable(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES)
+                    .build();
+
+    private final TypeReference<? extends List<T>> typeRef;
+
+    public JsonListPropertyEditor(TypeReference<? extends List<T>> typeRef) {
+        this.typeRef = typeRef;
+    }
+
+    @Override
+    public void setAsText(String text) throws IllegalArgumentException {
+        if (text == null || text.trim().isEmpty()) {
+            setValue(new ArrayList<T>());
+            return;
+        }
+        try {
+            setValue(OBJECT_MAPPER.readValue(text, typeRef));
+        } catch (Exception e) {
+            log.error("Failed to parse JSON list value", e);
+            throw new IllegalArgumentException(
+                    "Expected a JSON array but could not parse: " + e.getMessage());
+        }
+    }
+}