feat: implement custom JSON encoder for numpy types and ensure type consistency in inference results

app_mcp.py

@@ -38,6 +38,13 @@ os.environ['HF_HUB_CACHE'] = './models'
 load_dotenv()
 # print(os.getenv("HF_HUB_CACHE"))
 
+# Custom JSON Encoder to handle numpy types
+class NumpyEncoder(json.JSONEncoder):
+    def default(self, obj):
+        if isinstance(obj, np.float32):
+            return float(obj)
+        return json.JSONEncoder.default(self, obj)
+
 # Ensure using GPU if available
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
@@ -221,8 +228,8 @@ def infer(image: Image.Image, model_id: str, confidence_threshold: float = 0.75)
         result = entry.model(img)
         scores = entry.postprocess(result, entry.class_names)
         # Flatten output for Dataframe: include metadata and both class scores
-        ai_score = scores.get(entry.class_names[0], 0.0)
-        real_score = scores.get(entry.class_names[1], 0.0)
+        ai_score = float(scores.get(entry.class_names[0], 0.0))
+        real_score = float(scores.get(entry.class_names[1], 0.0))
         label = "AI" if ai_score >= confidence_threshold else ("REAL" if real_score >= confidence_threshold else "UNCERTAIN")
         return {
             "Model": entry.display_name,
@@ -462,8 +469,16 @@ def predict_image_with_json(img, confidence_threshold, augment_methods, rotate_d
 
     logger.info(f"Cleaned forensic images types: {[type(img) for img in cleaned_forensics_images]}")
 
+    # Ensure numerical values in results are standard Python floats before JSON serialization
+    for i, res_dict in enumerate(results):
+        for key in ["AI Score", "Real Score"]:
+            value = res_dict.get(key)
+            if isinstance(value, np.float32):
+                res_dict[key] = float(value)
+                logger.info(f"Converted {key} for result {i} from numpy.float32 to float.")
+
     # Return raw model results as JSON string for debug_json component
-    json_results = json.dumps(results)
+    json_results = json.dumps(results, cls=NumpyEncoder)
 
     return img_pil, cleaned_forensics_images, table_rows, json_results, consensus_html
 

utils/hf_logger.py

@@ -7,11 +7,19 @@ from PIL import Image
 import logging
 from datasets import Dataset, load_dataset, Features, Value, Sequence
 import copy
+import numpy as np
 
 logger = logging.getLogger(__name__)
 
 HF_DATASET_NAME = "aiwithoutborders-xyz/degentic_rd0" # TODO: Replace with your actual HF username and dataset name
 
+# Custom JSON Encoder to handle numpy types (copy from app_mcp.py if it's identical)
+class NumpyEncoder(json.JSONEncoder):
+    def default(self, obj):
+        if isinstance(obj, np.float32):
+            return float(obj)
+        return json.JSONEncoder.default(self, obj)
+
 def _pil_to_base64(image: Image.Image) -> str:
     """Converts a PIL Image to a base64 string."""
     # Explicitly check if the input is a PIL Image
@@ -91,12 +99,12 @@ def log_inference_data(
         new_entry = {
             "timestamp": datetime.datetime.now().isoformat(),
             "image": original_image_b64,
-            "inference_request": json.dumps(inference_params),
-            "model_predictions": json.dumps(model_predictions),
-            "ensemble_output": json.dumps(ensemble_output),
+            "inference_request": json.dumps(inference_params, cls=NumpyEncoder),
+            "model_predictions": json.dumps(model_predictions, cls=NumpyEncoder),
+            "ensemble_output": json.dumps(ensemble_output, cls=NumpyEncoder),
             "forensic_outputs": forensic_images_b64, # This is already a list of strings
-            "agent_monitoring_data": json.dumps(agent_monitoring_data),
-            "human_feedback": json.dumps(human_feedback if human_feedback is not None else {})
+            "agent_monitoring_data": json.dumps(agent_monitoring_data, cls=NumpyEncoder),
+            "human_feedback": json.dumps(human_feedback if human_feedback is not None else {}, cls=NumpyEncoder)
         }
         
         # Get current dataset features