Create app.py

app.py

@@ -1,30 +1,161 @@
-import gradio as gr
+# app.py
+
+import uvicorn
+import base64
+import io
+from pathlib import Path
+
+# Fastai imports
 from fastai.vision.all import *
-import skimage
-
-
-learn = load_learner('model.pkl')
-#from huggingface_hub import from_pretrained_fastai
-#learn = from_pretrained_fastai("devdatanalytics/commonbean")
-
-labels = learn.dls.vocab
-def predict(img):
-    img = PILImage.create(img)
-    pred,pred_idx,probs = learn.predict(img)
-    return {labels[i]: float(probs[i]) for i in range(len(labels))}
-
-#title = "Common beans diseases classfier"
-#description = "An app for Common beans diseases Classisfication"
-#article="<p style='text-align: center'>The app identifies and classifies common beans diseases: Anthracnose and Bean rust.</p>"
-# Create the Gradio interface
-interface = gr.Interface(
-    fn=predict,
-    inputs=gr.Image(),
-    outputs=gr.Label(num_top_classes=3)
+
+# FastHTML imports
+from fasthtml.common import * # Imports common HTML tags, App, etc.
+from fasthtml.core import FastHTML # Base class if needed, but App is usually sufficient
+from fasthtml.components import FileInput # Specific component for file input
+from fastcore.utils import * # For Upload class
+
+# --- Configuration ---
+# Ensure the path to your exported model is correct
+# When deploying to HF Spaces, this relative path should work if model.pkl is in the same directory
+MODEL_PATH = Path(__file__).parent / 'model.pkl'
+# Set device (CPU is usually the default/safest for HF free tier)
+defaults.device = torch.device('cpu')
+
+# --- Load Fastai Learner ---
+try:
+    print(f"Loading model from: {MODEL_PATH}")
+    learn = load_learner(MODEL_PATH)
+    print("Model loaded successfully.")
+    # Get class names (vocab) from the learner's dataloaders
+    CLASS_NAMES = learn.dls.vocab
+    print(f"Model Classes: {CLASS_NAMES}")
+except FileNotFoundError:
+    print(f"Error: Model file not found at {MODEL_PATH}")
+    print("Please make sure 'model.pkl' is in the same directory or update MODEL_PATH.")
+    # In a deployed environment, you might want to raise the error or handle it differently
+    # For now, we exit if the model isn't found on startup.
+    raise SystemExit(f"Error: Model file not found at {MODEL_PATH}")
+except Exception as e:
+    print(f"Error loading the model: {e}")
+    raise SystemExit(f"Error loading the model: {e}")
+
+# --- FastHTML App Setup ---
+# FastHTML automatically finds this 'app' object when run with uvicorn app:app
+app = FastHTML()
+rt = app.route # Route decorator
+
+# --- Helper Function for Prediction ---
+def predict_image(img_bytes: bytes):
+    """Takes image bytes, predicts using the fastai model."""
+    try:
+        img = PILImage.create(img_bytes)
+        pred_class, pred_idx, probs = learn.predict(img)
+        confidence = probs[pred_idx].item() # Get the probability of the predicted class
+        return pred_class, confidence
+    except Exception as e:
+        print(f"Error during prediction: {e}")
+        # Return a user-friendly error message and neutral confidence
+        return f"Prediction Error: {e}", 0.0
+
+# --- Define Routes ---
+
+@rt("/")
+async def get(request):
+    """Serves the main page with the upload form."""
+    # Using Bootstrap classes for basic styling
+    return Titled("Fastai Image Classifier",
+        Main(cls="container mt-4",
+            H1("Upload an Image for Classification"),
+            # Form for uploading the image
+            Form(
+                Div(cls="mb-3",
+                    # Label("Choose Image", fr="fileInput", cls="form-label"), # Optional label
+                    FileInput(name="file", id="fileInput", cls="form-control", required=True), # Added required
+                ),
+                Button("Classify Image", type="submit", cls="btn btn-primary"), # Submit button
+                # HTMX attributes for form submission
+                hx_post="/predict",      # POST request to /predict
+                hx_target="#results",    # Put the response into the #results div
+                hx_swap="innerHTML",     # Replace the content of #results
+                hx_encoding="multipart/form-data", # Needed for file uploads
+                # Add indicator for user feedback during processing
+                hx_indicator="#loading-spinner",
+                id="upload-form"
+            ),
+            # Loading indicator (hidden by default)
+            Div(id="loading-spinner", cls="htmx-indicator spinner-border mt-3", role="status",
+                Span("Loading...", cls="visually-hidden")
+            ),
+            # Div where results will be displayed
+            Div(id="results", cls="mt-4")
+        )
+    )
+
+@rt("/predict", methods=["POST"])
+async def post(request, file: Upload):
+    """Handles image upload, prediction, and returns results."""
+    if not file or not file.filename:
+        return P("No file uploaded. Please select a file.", cls="alert alert-warning")
+
+    # Check if it's likely an image file (basic check)
+    allowed_extensions = {'.png', '.jpg', '.jpeg', '.gif', '.bmp', '.webp'}
+    file_ext = Path(file.filename).suffix.lower()
+    if file_ext not in allowed_extensions:
+         return P(f"Invalid file type: {file_ext}. Please upload an image (png, jpg, jpeg, gif, bmp, webp).", cls="alert alert-danger")
+
+    print(f"Received file: {file.filename}, Content-Type: {file.content_type}")
+
+    # Read image bytes
+    try:
+        img_bytes = await file.read() # Use await for async file reading
+        if not img_bytes:
+             return P("Uploaded file is empty.", cls="alert alert-warning")
+    except Exception as e:
+        print(f"Error reading file: {e}")
+        return P(f"Error reading uploaded file: {e}", cls="alert alert-danger")
+
+    # Perform prediction
+    prediction, confidence = predict_image(img_bytes)
+
+    # Encode image to base64 to display it back
+    img_src = None
+    if "Error" not in prediction: # Only try to display image if prediction didn't fail critically
+        try:
+            img_base64 = base64.b64encode(img_bytes).decode('utf-8')
+            # Try to use the provided content type, default if necessary
+            content_type = file.content_type if file.content_type and file.content_type.startswith('image/') else 'image/jpeg'
+            img_src = f"data:{content_type};base64,{img_base64}"
+        except Exception as e:
+            print(f"Error encoding image to base64: {e}")
+            # Don't display image if encoding fails, but still show prediction
+
+    # Return the results as HTML fragment
+    # Using Bootstrap alert classes for results
+    result_cls = "alert alert-danger" if "Error" in prediction else "alert alert-success"
+
+    return Div(
+        (Img(src=img_src, alt="Uploaded Image", style="max-width: 300px; max-height: 300px; margin-top: 15px; margin-bottom: 10px; display: block;") if img_src else P("Preview not available.")),
+        Div(cls=f"{result_cls} mt-3", role="alert",
+            P(Strong("Prediction: "), f"{prediction}"),
+            P(Strong("Confidence: "), f"{confidence:.4f}") if "Error" not in prediction else ""
+        )
+    )
+
+# --- Add CSS/JS ---
+# Add Bootstrap CSS and JS for styling and components (like the spinner)
+app.hdrs.append(
+    Script(src="https://cdn.jsdelivr.net/npm/[email protected]/dist/js/bootstrap.bundle.min.js", integrity="sha384-C6RzsynM9kWDrMNeT87bh95OGNyZPhcTNXj1NW7RuBCsyN/o0jlpcV8Qyq46cDfL", crossorigin="anonymous"),
+)
+app.sheets.append(
+    Link(href="https://cdn.jsdelivr.net/npm/[email protected]/dist/css/bootstrap.min.css", rel="stylesheet", integrity="sha384-T3c6CoIi6uLrA9TneNEoa7RxnatzjcDSCmG1MXxSR1GAsXEV/Dwwykc2MPK8M2HN", crossorigin="anonymous")
 )
+# Add HTMX itself (FastHTML includes a minimal version, but sometimes the full CDN is useful)
+# app.hdrs.append(Script(src="https://unpkg.com/[email protected]/dist/htmx.min.js"))
 
-# Enable the queue to handle POST requests
-interface.queue(api_open=True)
 
-# Launch the interface
-interface.launch()
+# --- Run the App (for local testing) ---
+# This part is mainly for running locally with `python app.py`
+# When deployed on Hugging Face Spaces, Uvicorn is run automatically based on the README config
+if __name__ == "__main__":
+    # Use port 8000 which is often standard for these deployments
+    uvicorn.run("app:app", host="0.0.0.0", port=8000, reload=True)