feat: bump to v2.1.0 — Phase 3 multi-modal depth

- VideoFrameAnalyzer (Layer 12): ffmpeg frame sampling, temporal correlation
- Cross-modal context fusion: escalates corroborating sub-threshold signals across text/visual/agentic domains
- images= parameter on guardian.analyze() for direct image passing
- SVG attack surface deepened: script injection, event handlers, external refs, foreignObject, CSS injection
- QR-in-stego bridge: LSB payload extraction dispatched to QRCodeAnalyzer on BLOCK-level chi-square
- New optional extras: vision2 (scipy, imagehash, opencv), video (ffmpeg-python)
- Exports VideoFrameAnalyzer, VideoFrameResult, FrameThreat from analyzers package

Author: Oracles Technologies LLC

ethicore_guardian/__init__.py

@@ -7,7 +7,7 @@
 """
 
 # Version information
-__version__ = "2.0.0"
+__version__ = "2.1.0"
 __author__ = "Oracles Technologies LLC"
 
 # Core exports

ethicore_guardian/versions.py

@@ -2,7 +2,7 @@
 Ethicore Engine™ - Guardian SDK - Version Information
 """
 
-__version__ = "2.0.0"
+__version__ = "2.1.0"
 __version_info__ = tuple(map(int, __version__.split('.')))
 
 # Build information

pyproject.toml

@@ -110,6 +110,15 @@ vision2 = [
     "opencv-python-headless>=4.8.0",
 ]
 
+# video: enables Layer 12 (VideoFrameAnalyzer — frame sampling + temporal correlation).
+# Also requires the ffmpeg binary in PATH:
+#   macOS : brew install ffmpeg
+#   Ubuntu: apt install ffmpeg
+#   Windows: https://ffmpeg.org/download.html
+video = [
+    "ffmpeg-python>=0.2.0",
+]
+
 # ml: enables full transformer-based ML inference in MLInferenceEngine.
 # Without these packages the engine runs its built-in heuristic fallback,
 # which is still effective but less accurate than a fine-tuned classifier.