Update app.py

app.py

@@ -1,13 +1,14 @@
 from fastapi import FastAPI, File, UploadFile, Body
-from fastapi.responses import RedirectResponse
+from fastapi.responses import RedirectResponse, StreamingResponse
 from fastapi.middleware.cors import CORSMiddleware
 from PIL import Image
 import io
 import numpy as np
 from structure import  TextPredictionRequest, PredictionResponse
-from detector import detect_embedding
+from detector import detect_embedding, _active_layer
 from text_embedder import get_text_embedding
 from image_embedder import get_image_embedding
+from kuwahara import apply_kuwahara
 
 origins=[
     "http://localhost:5173",
@@ -36,6 +37,26 @@ async def root():
     # Redirect to the automatic Swagger UI provided by FastAPI
     return RedirectResponse(url="/docs")
 
+@app.post(
+    "/debug/kuwahara",
+    summary="Debug: Get raw Kuwahara output",
+    description="Upload an image to see exactly how the Kuwahara filter transforms it in memory.",
+)
+async def debug_kuwahara(image: UploadFile = File(...)):
+    """Accept an image upload and return the Kuwahara processed image directly."""
+    try:
+        image_data = await image.read()
+        pil_img = Image.open(io.BytesIO(image_data)).convert("RGB")
+        processed_img = apply_kuwahara(pil_img)
+        # Save to memory buffer and stream
+        buf = io.BytesIO()
+        processed_img.save(buf, format="PNG")
+        buf.seek(0)
+        return StreamingResponse(buf, media_type="image/png")
+    except Exception as e:
+        print(f"Error in debug kuwahara: {e}")
+        return {"error": str(e)}
+
 @app.post(
     "/predict/image",
     response_model=PredictionResponse,
@@ -49,12 +70,12 @@ async def predict(image: UploadFile = File(...)):
         image_data = await image.read()
         pil_img = Image.open(io.BytesIO(image_data)).convert("RGB")
         emb = get_image_embedding(pil_img)
-        prediction = detect_embedding(emb)
+        prediction = detect_embedding(emb, threshold=0.5)
         print(f"Image prediction: {prediction['predicted_class']} with confidence {prediction['confidence']:.4f}")
         return PredictionResponse(predicted_class=prediction["predicted_class"], confidence=prediction["confidence"])
     except Exception as e:
         print(f"Error in image prediction: {e}")
-        return PredictionResponse(predicted_class=0, confidence=0)
+        return PredictionResponse(predicted_class=0, confidence=0.0)
 @app.post(
     "/predict/text",
     response_model=PredictionResponse,
@@ -64,7 +85,8 @@ async def predict(image: UploadFile = File(...)):
 async def predict_text_endpoint(payload: TextPredictionRequest = Body(...)):
     """Accept a text string and return a prediction of whether it's human or AI-generated."""
     try:
-        emb    = get_text_embedding(payload.text)
+        layer = _active_layer if _active_layer is not None else -1
+        emb    = get_text_embedding(payload.text, layer=layer)
         result = detect_embedding(emb)
         
         return PredictionResponse(
@@ -74,4 +96,4 @@ async def predict_text_endpoint(payload: TextPredictionRequest = Body(...)):
     except Exception as e:
         # Return a fallback response in case of error
         print(f"Error in text prediction: {e}")
-        return PredictionResponse(predicted_class=0, confidence=0)
+        return PredictionResponse(predicted_class=0, confidence=0.0)