Upload 3 files

app.py

@@ -64,20 +64,10 @@ if "time_taken_for_response" not in st.session_state:
 if "metrics" not in st.session_state:
     st.session_state.metrics = {}
 
-# Streamlit Sidebar for Recent Questions
-st.sidebar.title("Recent Questions")
-
 recent_data = load_recent_questions()
-for q in reversed(recent_data["questions"]):  # Show latest first
-    st.sidebar.write(f"🔹 {q['question']}")
-    st.json({q['metrics']})
-
-st.sidebar.markdown("---")  # Separator
-
 
 import matplotlib.pyplot as plt
 
-
 # for visualization
 st.sidebar.title("Analytics")
 
@@ -93,6 +83,14 @@ ax.set_ylabel("Time Taken for Response")
 ax.legend()
 st.sidebar.pyplot(fig)
 
+st.sidebar.markdown("---")  # Separator
+
+# Streamlit Sidebar for Recent Questions
+st.sidebar.title("Recent Questions")
+for q in reversed(recent_data["questions"]):  # Show latest first
+    with st.expander(f"🔹 {q['question']}"):
+        st.json(q["metrics"])
+
 # Submit Button
 # if st.button("Submit"):
 #     start_time = time.time()

data_processing.py

@@ -95,11 +95,11 @@ def load_ragbench():
 
 def load_query_dataset(query_dataset):
     global query_dataset_data
-    if query_dataset_data:
-        return query_dataset_data  
+    if query_dataset_data[query_dataset]:
+        return query_dataset_data[query_dataset] 
     else:
-        query_dataset_data = load_dataset("rungalileo/ragbench", query_dataset)
-    return query_dataset_data
+        query_dataset_data[query_dataset] = load_dataset("rungalileo/ragbench", query_dataset)
+    return query_dataset_data[query_dataset]
 
 def load_faiss(query_dataset):
     global index
@@ -149,5 +149,27 @@ def save_recent_question(question, metrics):
     data["questions"]  = data["questions"][-5:]
 
     # Write back to file
+    with open(RECENT_QUESTIONS_FILE, "w") as file:
+        json.dump(data, file, indent=4)
+
+# Function to save/update a question in the recent list
+def save_recent_question(question, metrics):
+    data = load_recent_questions()
+
+    # Check if the question already exists
+    existing_questions = {q["question"]: q for q in data["questions"]}
+    
+    if question in existing_questions:
+        # Update metrics & move to the latest position
+        existing_questions[question]["metrics"] = metrics
+        data["questions"].remove(existing_questions[question])  # Remove old entry
+
+    # Add the updated/new question at the latest position
+    data["questions"].append({"question": question, "metrics": metrics})
+
+    # Keep only the last 10 questions
+    data["questions"] = data["questions"][-5:]
+
+    # Write back to the file
     with open(RECENT_QUESTIONS_FILE, "w") as file:
         json.dump(data, file, indent=4)

evaluation.py

@@ -111,22 +111,21 @@ def calculate_metrics(question, response, docs, time_taken):
     }
     return predicted_metrics
 
-def retrieve_ground_truths(question, ragbench_set):
-    for dataset_name in ragbench_set.keys():
-        for split_name, instances in ragbench_set[dataset_name].items():
-            print(f"Processing {split_name} split")
-            for instance in instances:
-                if instance['question'] == question:
-                    instance_id = instance['id']
-                    instance_response = instance['response']
-                    # ground_truth_metrics = {
-                    #     "context_relevance": instance['relevance_score'],
-                    #     "context_utilization": instance['utilization_score'],
-                    #     "completeness": instance['completeness_score'],
-                    #     "adherence": instance['adherence_score']
-                    # }
-                    print(f"Match found in {split_name} split!")
-                    print(f"ID: {instance_id}, Response: {instance_response}")
-                    return instance_response  # Return ground truth response immediately
+def retrieve_ground_truths(question, dataset):
+    for split_name, instances in dataset.items():
+        print(f"Processing {split_name} split")
+        for instance in instances:
+            if instance['question'] == question:
+                instance_id = instance['id']
+                instance_response = instance['response']
+                # ground_truth_metrics = {
+                #     "context_relevance": instance['relevance_score'],
+                #     "context_utilization": instance['utilization_score'],
+                #     "completeness": instance['completeness_score'],
+                #     "adherence": instance['adherence_score']
+                # }
+                print(f"Match found in {split_name} split!")
+                print(f"ID: {instance_id}, Response: {instance_response}")
+                return instance_response  # Return ground truth response immediately
 
     return None  # Return None if no match is found