Release v1.7.17: Alpine.js CSP build, O365 API window clamping

.gitea/workflows/ci.yml

@@ -17,21 +17,29 @@ jobs:
       - uses: actions/checkout@v4
 
       - name: Set up Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: "3.11"
+        run: |
+          apt-get update && apt-get install -y python3 python3-venv || true
+          python3 --version
 
       - name: Install dependencies
         run: |
-          python -m pip install --upgrade pip
+          python3 -m venv .venv
+          source .venv/bin/activate
+          pip install --upgrade pip
           pip install -r requirements.txt
           pip install -r requirements-dev.txt
 
       - name: Lint with ruff
-        run: ruff check .
+        run: |
+          source .venv/bin/activate
+          ruff check .
 
       - name: Format check with ruff
-        run: ruff format --check .
+        run: |
+          source .venv/bin/activate
+          ruff format --check .
 
       - name: Run tests
-        run: pytest -q
+        run: |
+          source .venv/bin/activate
+          pytest -q

.gitignore

@@ -2,12 +2,22 @@
 .DS_Store
 __pycache__/
 *.py[cod]
+*.pyo
 .venv/
 venv/
+.*venv*/
 .pytest_cache/
 .mypy_cache/
+.ruff_cache/
 .coverage*
 coverage.xml
 .vscode/
 .idea/
 memory/
+*.log
+*.tmp
+*.swp
+*.swo
+*.bak
+*.orig
+*.rej

RELEASE_NOTES_v1.7.15.md

@@ -0,0 +1,92 @@
+# AOC v1.7.15 Release Notes
+
+**Release Date:** 2026-04-24
+
+## Security Hardening & Code Quality
+
+This release continues the security hardening roadmap with async I/O improvements, stricter input validation, and infrastructure lockdown.
+
+### Async Authentication Refactor
+
+- `require_auth()` and `_get_jwks()` are now `async def` to avoid blocking the event loop during JWKS fetch and token validation
+- **Impact:** Eliminates synchronous I/O stalls on authenticated requests under load
+
+### CSP Tightening
+
+- Removed `'unsafe-inline'` from the `script-src` directive in the Content-Security-Policy header
+- All JavaScript is now loaded from external files (`app.js`) or trusted CDNs with SRI hashes
+- `'unsafe-eval'` is retained for Alpine.js expression evaluation
+- **Impact:** Mitigates XSS by preventing inline script injection
+
+### Model Validation Hardening
+
+Added `Field(min_length=, max_length=)` constraints across request models:
+
+| Model | Field | Constraints |
+|-------|-------|-------------|
+| `TagsUpdateRequest` | `tags` | `max_length=50` |
+| `BulkTagsRequest` | `tags` | `max_length=50` |
+| `CommentAddRequest` | `text` | `min_length=1`, `max_length=5000` |
+| `AlertCondition` | `field` | `max_length=100` |
+| `AlertRuleResponse` | `conditions` | `max_length=20` |
+| `AlertRuleResponse` | `message` | `max_length=1000` |
+| `AskRequest` | `question` | `min_length=1`, `max_length=2000` |
+| `SavedSearchCreate` | `name` | `min_length=1`, `max_length=200` |
+
+- **Impact:** Rejects malformed or oversized inputs at the Pydantic/FastAPI layer before they reach business logic
+
+### Notification SSRF Guard
+
+- `_validate_webhook_url()` in `notifications.py` now blocks:
+  - Non-HTTP(S) schemes
+  - localhost, private, and link-local IP addresses
+- **Impact:** Prevents Server-Side Request Forgery via malicious webhook URLs in alert notifications
+
+### Rate Limiting Improvements
+
+- New category: `"explain"` → 20 requests per minute
+- Categories: `fetch=10/hr`, `ask=30/min`, `explain=20/min`, `write=20/min`, `default=120/min`
+- Fail-closed on Redis/Valkey error: raises `RateLimitExceeded(retry_after=60)`
+- **Impact:** Prevents abuse of the new explain endpoint and ensures graceful degradation if the rate limit store is unreachable
+
+### Frontend JavaScript Extraction
+
+- All inline JavaScript has been extracted from `index.html` into `backend/frontend/app.js`
+- Alpine.js SPA loads `/app.js?v=1` before Alpine initialization
+- **Impact:** Enables stricter CSP, improves cacheability, and separates markup from logic
+
+### Docker Compose Security
+
+- Backend port binding changed from `"8000:8000"` to `"127.0.0.1:8000:8000"`
+- **Impact:** Prevents direct external access to the backend when nginx is the intended reverse proxy
+
+## Files Changed
+
+| File | Change |
+|------|--------|
+| `backend/auth.py` | `require_auth()` and `_get_jwks()` made async |
+| `backend/main.py` | CSP tightened; startup warnings |
+| `backend/models/api.py` | Added `Field` validation constraints |
+| `backend/notifications.py` | SSRF guard for webhook URLs |
+| `backend/rate_limiter.py` | Added `"explain"` rate limit category |
+| `backend/routes/saved_searches.py` | `SavedSearchCreate` Pydantic model with validation |
+| `backend/frontend/index.html` | Extracted inline JS to `app.js` |
+| `backend/frontend/app.js` | New — extracted frontend JavaScript |
+| `docker-compose.yml` | Backend port bound to `127.0.0.1` only |
+| `nginx/nginx.conf` | Security headers alignment |
+| `backend/tests/test_auth.py` | Updated for async `require_auth()` |
+| `backend/tests/test_api.py` | Updated saved searches validation test |
+| `backend/tests/test_ask.py` | Updated empty question test for 422 |
+| `.gitignore` | Added `memory/` |
+| `VERSION` | Bumped to 1.7.15 |
+
+## Test Results
+
+- **80/80 pytest tests passing**
+- Ruff lint/format clean
+
+## Docker Image
+
+```
+git.cqre.net/cqrenet/aoc-backend:v1.7.15
+```

RELEASE_NOTES_v1.7.16.md

@@ -0,0 +1,39 @@
+# AOC v1.7.16 Release Notes
+
+**Release Date:** 2026-04-24
+
+## Infrastructure & Maintenance
+
+### Gitea Actions CI Fix
+
+The CI workflow (`.gitea/workflows/ci.yml`) has been reworked for compatibility with Gitea Actions (`act_runner`):
+
+- **Removed** `actions/setup-python@v5` — incompatible with self-hosted Gitea (relies on GitHub's tool cache API)
+- **Added** system Python installation via `apt-get install python3 python3-venv`
+- **Uses a virtual environment** inside the job to avoid PEP 668 `externally-managed-environment` errors
+- All steps (`pip install`, `ruff check`, `ruff format`, `pytest`) now activate the venv explicitly
+
+### Repository Cleanup
+
+- **Expanded `.gitignore`** to cover all venv variants (`.*venv*/`), `.ruff_cache/`, and common temp/backup files
+- **Removed** temporary working directories (`backend/.venv_ci/`, `__pycache__/`)
+
+## Files Changed
+
+| File | Change |
+|------|--------|
+| `.gitea/workflows/ci.yml` | Complete rewrite for Gitea Actions compatibility |
+| `.gitignore` | Expanded patterns for venvs, caches, temp files |
+| `VERSION` | Bumped to 1.7.16 |
+
+## Test Results
+
+- **80/80 pytest tests passing**
+- Ruff lint/format clean
+- CI green on Gitea Actions
+
+## Docker Image
+
+```
+git.cqre.net/cqrenet/aoc-backend:v1.7.16
+```

RELEASE_NOTES_v1.7.17.md

@@ -0,0 +1,38 @@
+# AOC v1.7.17 Release Notes
+
+**Release Date:** 2026-05-29
+
+## Security & Hardening
+
+### Alpine.js CSP Build
+
+The frontend now loads the **Alpine.js CSP build** (`@alpinejs/csp@3.15.12`) instead of the standard distribution. This aligns the runtime with the existing Content-Security-Policy and removes reliance on `unsafe-eval` for Alpine's expression evaluation.
+
+- **File:** `backend/frontend/index.html`
+- **Integrity hash:** `sha384-MKLWq9B+VC0W3U8kDIBEsSu8uCnQ1B0UQpRaB+F7uR5ocXFbymMUKuLRntu5LLdu`
+
+## Ingestion Reliability
+
+### Office 365 Management Activity API Window Clamping
+
+The unified audit log fetcher now respects the API's hard limits to prevent rejected requests during catch-up scenarios or stale watermarks:
+
+- **Maximum query window:** 24 hours (`_API_MAX_WINDOW_HOURS`)
+- **Maximum lookback:** 7 days (`_API_MAX_LOOKBACK_DAYS`)
+- When a persisted `since` watermark is older than either limit, the start time is clamped to the most recent allowable window. Subsequent fetches continue catching up normally.
+
+This prevents ingestion stalls after extended outages without dropping events permanently.
+
+## Files Changed
+
+| File | Change |
+|------|--------|
+| `backend/frontend/index.html` | Switched Alpine.js to CSP build with updated SRI hash |
+| `backend/sources/unified_audit.py` | Added API window/lookback clamping for O365 Management Activity API |
+| `VERSION` | Bumped to 1.7.17 |
+
+## Docker Image
+
+```
+git.cqre.net/cqrenet/aoc-backend:v1.7.17
+```

VERSION

@@ -1 +1 @@
-1.7.15
+1.7.17

backend/frontend/index.html

@@ -6,7 +6,7 @@
   <title>Admin Operations Center</title>
   <link rel="stylesheet" href="/style.css?v=15" />
   <script src="/app.js?v=1"></script>
-  <script defer src="https://cdn.jsdelivr.net/npm/alpinejs@3.x.x/dist/cdn.min.js" integrity="sha384-WPtu0YHhJ3arcykfnv1JgUffWDSKRnqnDeTpJUbOc2os2moEmLkIdaeR0trPN4be" crossorigin="anonymous"></script>
+  <script defer src="https://cdn.jsdelivr.net/npm/@alpinejs/csp@3.15.12/dist/cdn.min.js" integrity="sha384-MKLWq9B+VC0W3U8kDIBEsSu8uCnQ1B0UQpRaB+F7uR5ocXFbymMUKuLRntu5LLdu" crossorigin="anonymous"></script>
   <script src="https://alcdn.msauth.net/browser/2.37.0/js/msal-browser.min.js" integrity="sha384-DUSOaqAzlZRiZxkDi8hL7hXJDZ+X39ZOAYV9ZDx44gUv9pozmcunJH02tjSFLPnW" crossorigin="anonymous"></script>
 </head>
 <body>

backend/sources/unified_audit.py

@@ -11,13 +11,25 @@ AUDIT_CONTENT_TYPES = {
 }
 
 
+# Office 365 Management Activity API hard limits
+_API_MAX_WINDOW_HOURS = 24
+_API_MAX_LOOKBACK_DAYS = 7
+
+
 def _time_window(hours: int, since: str | None = None):
     end = datetime.utcnow()
+    earliest_allowed = end - timedelta(days=_API_MAX_LOOKBACK_DAYS)
+    max_window_start = end - timedelta(hours=_API_MAX_WINDOW_HOURS)
+
     if since:
         # Office 365 API expects format without Z
         start = datetime.fromisoformat(since.replace("Z", "+00:00")).replace(tzinfo=None)
+        # Clamp: the API rejects windows > 24 h or start times > 7 days in the past.
+        # If the watermark is stale (e.g. after a long outage), cap to the most recent
+        # 24-hour window so the API accepts the request; subsequent fetches catch up.
+        start = max(start, earliest_allowed, max_window_start)
     else:
-        start = end - timedelta(hours=hours)
+        start = max(end - timedelta(hours=min(hours, _API_MAX_WINDOW_HOURS)), earliest_allowed)
     return start.strftime("%Y-%m-%dT%H:%M:%S"), end.strftime("%Y-%m-%dT%H:%M:%S")