Upload folder using huggingface_hub

app.py

@@ -201,8 +201,13 @@ def analyze_forensics(
     verdict_html += "</div>"
 
     # Overlay Image (Original + Heatmap)
-    overlay_img = explainability.get("overlay")
-    if overlay_img is None:
+    heatmap_bytes = explainability.get("heatmap_bytes")
+    if heatmap_bytes:
+        try:
+            overlay_img = Image.open(io.BytesIO(heatmap_bytes))
+        except Exception:
+            overlay_img = image
+    else:
         overlay_img = image
 
     # Consensus Table
@@ -252,15 +257,18 @@ def analyze_forensics(
 # ─────────────────────────────────────────────────────────────────────────────
 # Interactive Forensic Investigator Chat (Multi-turn)
 # ─────────────────────────────────────────────────────────────────────────────
-async def chat_investigator_fn(message: str, history: List[Dict[str, str]], context: Dict[str, Any]):
-    if not message.strip():
-        return history, ""
+async def chat_investigator_fn(message: str, history: Optional[List[Any]], context: Dict[str, Any]):
+    if not message or not message.strip():
+        return history or [], ""
 
     messages = []
     if history:
         for msg in history:
             if isinstance(msg, dict):
                 messages.append({"role": msg.get("role", "user"), "content": msg.get("content", "")})
+            elif isinstance(msg, (list, tuple)) and len(msg) >= 2:
+                messages.append({"role": "user", "content": str(msg[0])})
+                messages.append({"role": "assistant", "content": str(msg[1])})
     messages.append({"role": "user", "content": message})
 
     result = await chat_with_forensic_context(
@@ -366,7 +374,7 @@ with gr.Blocks(title="DeepGuard — Forensic Detection") as demo:
     gr.Markdown("### 🕵️ Interactive Forensic Investigator Chat")
     gr.Markdown("Inquire deeper into model decision boundaries, specific attention zones, or contradictory detector signals.")
 
-    chatbot = gr.Chatbot(height=320, label="DeepGuard Forensic Investigator")
+    chatbot = gr.Chatbot(height=320, label="DeepGuard Forensic Investigator", type="messages")
     with gr.Row():
         chat_input = gr.Textbox(
             placeholder="Ask about this image's verdict, Grad-CAM peaks, or provenance flags...",
@@ -395,47 +403,6 @@ with gr.Blocks(title="DeepGuard — Forensic Detection") as demo:
 
 
 # ─────────────────────────────────────────────────────────────────────────────
-# FastAPI REST API mounting (Allows direct HTTP /predict & /chat requests)
+# Launch ZeroGPU Space
 # ─────────────────────────────────────────────────────────────────────────────
-api_app = FastAPI(title="DeepGuard REST API", version="1.0.0")
-api_app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-
-@api_app.get("/health")
-async def health():
-    return {
-        "status": "ok",
-        "service": "DeepGuard HF Space (ZeroGPU)",
-        "device": str(DEVICE),
-        "zerogpu": has_spaces,
-    }
-
-
-class ChatRequest(BaseModel):
-    message: str
-    history: Optional[List[Dict[str, str]]] = []
-    forensic_context: Dict[str, Any]
-
-
-@api_app.post("/chat")
-async def api_chat(req: ChatRequest):
-    messages = list(req.history or [])
-    messages.append({"role": "user", "content": req.message})
-    result = await chat_with_forensic_context(
-        messages=messages,
-        forensic_context=req.forensic_context,
-    )
-    return {"reply": result.get("reply", "")}
-
-
-# Mount Gradio onto FastAPI
-app = gr.mount_gradio_app(api_app, demo, path="/")
-
-if __name__ == "__main__":
-    demo.launch()
+demo.queue(default_concurrency_limit=5).launch()

detectors/itsnotai.py

@@ -160,5 +160,13 @@ class ItsNotAIDetector(BaseDetector):
 
   def get_gradcam_target_layer(self) -> torch.nn.Module:
     self._ensure_loaded()
-    # In HuggingFace BeitModel, transformer blocks are stored in beit.layers
-    return self._model.beit.layers[-1].mlp
+    if hasattr(self._model, "beit"):
+      beit = self._model.beit
+      if hasattr(beit, "encoder") and hasattr(beit.encoder, "layer") and len(beit.encoder.layer) > 0:
+        layer = beit.encoder.layer[-1]
+        return getattr(layer, "output", layer)
+      if hasattr(beit, "layers") and len(beit.layers) > 0:
+        layer = beit.layers[-1]
+        return getattr(layer, "mlp", layer)
+    children = list(self._model.children())
+    return children[-2] if len(children) >= 2 else children[-1]

requirements.txt

@@ -1,7 +1,5 @@
 gradio>=5.0.0
 transformers
-torch
-torchvision
 Pillow
 numpy
 matplotlib
@@ -13,6 +11,3 @@ scikit-learn
 grad-cam
 scipy
 peft
-fastapi
-python-multipart
-uvicorn