refactor: clean up model roster comments and remove broken model references

FIX.md

@@ -13,13 +13,8 @@ Copy `fingerprint_engine.py` (from outputs) to:
 src/engines/fingerprint/engine.py
 ```
 
-This removes ALL broken models:
-
-- `yermandy/deepfake-detection` — gone
-- `yermandy/GenD_CLIP_L_14` — gone
-- `yermandy/GenD_DINOv3_L` — gone
-- `Wvolf/ViT_Deepfake_Detection` — gone
-- `trust_remote_code` kwarg bug — fixed
+This removes all broken legacy models and fixes the `trust_remote_code`
+duplicate kwarg bug.
 
 Replaces with 3 working models:
 
@@ -37,10 +32,9 @@ Copy `coherence_engine.py` (from outputs) to:
 src/engines/coherence/engine.py
 ```
 
-This removes the broken wav2vec model
-(`nii-yamagishilab/wav2vec-large-anti-deepfake-nda`) which has incompatible
-weights and was producing random output anyway. Coherence now runs visual-only
-(FaceNet + MediaPipe).
+This removes the broken wav2vec-based audio model path with incompatible
+weights and random output behavior. Coherence now runs visual-only (FaceNet +
+MediaPipe).
 
 ---
 
@@ -52,7 +46,8 @@ Copy `sstgnn_engine.py` (from outputs) to:
 src/engines/sstgnn/engine.py
 ```
 
-Removes `Wvolf/ViT_Deepfake_Detection`. Uses `dima806` + `prithivMLmods` only.
+Removes legacy ViT `.pt` checkpoint usage. Uses `dima806` + `prithivMLmods`
+only.
 
 ---
 
@@ -110,10 +105,9 @@ torchvision>=0.21.0
 torchaudio>=2.6.0
 ```
 
-Torch < 2.6 blocks loading `.pt` files due to CVE-2025-32434.
-`Wvolf/ViT_Deepfake_Detection` uses `.pt` — it will NEVER load on torch < 2.6.
-Since you're removing that model anyway, this is a safety measure for other
-models.
+Torch < 2.6 blocks loading `.pt` files due to CVE-2025-32434. Since legacy
+`.pt` checkpoints are removed, this is still a safety measure for other
+dependencies that may call `torch.load`.
 
 ---
 
@@ -167,8 +161,8 @@ SSTGNN engine ready: 2 detectors, mediapipe=ok
 
 | Error                             | Cause                                       | Fix                               |
 | --------------------------------- | ------------------------------------------- | --------------------------------- |
-| `yermandy/*` warnings             | custom GenD arch                            | removed from engine               |
-| `Wvolf/*` torch CVE error         | .pt file + torch < 2.6                      | removed from engine               |
+| Legacy GenD warnings              | custom architecture                          | removed from engine               |
+| Legacy ViT torch CVE error        | `.pt` file + torch < 2.6                    | removed from engine               |
 | `trust_remote_code` TypeError     | duplicate kwarg in \_build_image_classifier | removed from all pipeline() calls |
 | `wav2vec` MISSING/UNEXPECTED keys | custom m_ssl.\* namespace, incompatible     | removed from engine               |
 | `libGLESv2.so.2 not found`        | missing apt packages in Docker              | add libgles2 libegl1 libgbm1      |

src/engines/fingerprint/engine.py

@@ -24,20 +24,7 @@ from src.types import EngineResult
 logger = logging.getLogger(__name__)
 CACHE = os.environ.get("MODEL_CACHE_DIR", "/tmp/models")
 
-# ---------------------------------------------------------------------------
-# Model roster — all confirmed safetensors, standard HF architectures
-# ---------------------------------------------------------------------------
-# Primary: Organika/sdxl-detector — ViT-L/14 fine-tuned on 2M real/AI images
-# Secondary: haywoodsloan/ai-image-detector-deploy — EfficientNet, robust
-# Tertiary: dima806/deepfake_vs_real_image_detection — ResNet50, 70k faces
-#
-# REMOVED (broken):
-#   yermandy/deepfake-detection       — missing model_type in config.json
-#   yermandy/GenD_CLIP_L_14           — custom GenD arch, not in Transformers
-#   yermandy/GenD_DINOv3_L            — same GenD arch issue
-#   Wvolf/ViT_Deepfake_Detection      — .pt file + trust_remote_code conflict +
-#                                       torch CVE-2025-32434 block
-# ---------------------------------------------------------------------------
+
 DETECTOR_CANDIDATES = [
     "Organika/sdxl-detector",
     "haywoodsloan/ai-image-detector-deploy",