Update app.py

app.py

@@ -1,31 +1,29 @@
 import gradio as gr
 import json
-import requests
+import os
 import numpy as np
 import matplotlib.pyplot as plt
 import seaborn as sns
 from io import BytesIO
 
 # -------------------------------
-# 1. Configuration and Data Loading
+# 1. Load Results from Local File
 # -------------------------------
-# URL to the JSON file (the URL below resolves to the raw file)
-DATA_URL = "https://huggingface.co/spaces/alielfilali01/3C3H-HeatMap/resolve/main/files/aragen_v1_results.json"
-
-# Define the metrics order (6 dimensions)
-METRICS = ["Correctness", "Completeness", "Conciseness", "Helpfulness", "Honesty", "Harmlessness"]
-
-def load_data(url=DATA_URL):
-    response = requests.get(url)
-    data = response.json()
-    # Filter out any non-model entries (e.g. timestamp entries)
+def load_results():
+    # Get the directory of the current file
+    current_dir = os.path.dirname(os.path.abspath(__file__))
+    # Construct the path to the JSON file (assumes file is stored in "files/aragen_v1_results.json")
+    results_file = os.path.join(current_dir, "files", "aragen_v1_results.json")
+    with open(results_file, "r") as f:
+        data = json.load(f)
+    # Filter out any non-model entries (e.g., timestamp entries)
     model_data = [entry for entry in data if "Meta" in entry]
     return model_data
 
 # Load the JSON data once when the app starts
-DATA = load_data()
+DATA = load_results()
 
-# Extract model names for the dropdown based on the JSON "Meta" field
+# Extract model names for the dropdown from the JSON "Meta" field
 def get_model_names(data):
     model_names = [entry["Meta"]["Model Name"] for entry in data]
     return model_names
@@ -33,38 +31,39 @@ def get_model_names(data):
 MODEL_NAMES = get_model_names(DATA)
 
 # -------------------------------
-# 2. Heatmap Generation Functions
+# 2. Define Metrics and Heatmap Generation Functions
 # -------------------------------
+# Define the six metrics in the desired order.
+METRICS = ["Correctness", "Completeness", "Conciseness", "Helpfulness", "Honesty", "Harmlessness"]
+
 def generate_heatmap_image(model_entry):
     """
-    Given a model entry from the JSON data, this function extracts the 6 metrics,
-    computes a 6x6 similarity matrix using the definition: similarity = 1 - |v_i - v_j|,
-    and returns the heatmap image as bytes.
+    For a given model entry, extract the six metrics and compute a 6x6 similarity matrix
+    using the definition: similarity = 1 - |v_i - v_j|, then return the heatmap image as bytes.
     """
     scores = model_entry["claude-3.5-sonnet Scores"]["3C3H Scores"]
-    # Create a vector with the metrics in the defined order
+    # Create a vector with the metrics in the defined order.
     v = np.array([scores[m] for m in METRICS])
-    # Compute the 6x6 similarity matrix
+    # Compute the 6x6 similarity matrix.
     matrix = 1 - np.abs(np.subtract.outer(v, v))
-    
-    # Create a mask for the upper triangle (diagonal remains visible)
+    # Create a mask for the upper triangle (keeping the diagonal visible).
     mask = np.triu(np.ones_like(matrix, dtype=bool), k=1)
     
     plt.figure(figsize=(6, 5))
-    ax = sns.heatmap(matrix,
-                     mask=mask,
-                     annot=True,
-                     fmt=".2f",
-                     cmap="viridis",
-                     xticklabels=METRICS,
-                     yticklabels=METRICS,
-                     cbar_kws={"label": "Similarity"})
+    sns.heatmap(matrix,
+                mask=mask,
+                annot=True,
+                fmt=".2f",
+                cmap="viridis",
+                xticklabels=METRICS,
+                yticklabels=METRICS,
+                cbar_kws={"label": "Similarity"})
     plt.title(f"Confusion Matrix for Model: {model_entry['Meta']['Model Name']}")
     plt.xlabel("Metrics")
     plt.ylabel("Metrics")
     plt.tight_layout()
     
-    # Save the figure to a bytes buffer
+    # Save the plot to a bytes buffer.
     buf = BytesIO()
     plt.savefig(buf, format="png")
     plt.close()
@@ -73,8 +72,8 @@ def generate_heatmap_image(model_entry):
 
 def generate_heatmaps(selected_model_names):
     """
-    Filters the global DATA for entries matching the selected model names,
-    generates a heatmap for each one, and returns a list of image bytes.
+    Filter the global DATA for entries matching the selected model names,
+    generate a heatmap for each, and return a list of image bytes.
     """
     filtered_entries = [entry for entry in DATA if entry["Meta"]["Model Name"] in selected_model_names]
     images = []