Upload 4 files

app.py

@@ -71,25 +71,30 @@ import matplotlib.pyplot as plt
 # for visualization
 st.sidebar.title("Analytics")
 
-#context_relevance = [q["metrics"]["context_relevance"] for q in recent_data["questions"]]
-response_time = [q["metrics"]["response_time"] for q in recent_data["questions"]]
+# response_time = [q["response_time"] for q in recent_data["questions"]]
+# labels = [f"Q{i+1}" for i in range(len(response_time))]  # Labels for X-axis
+
+# fig, ax = plt.subplots()
+# ax.set_xlabel("Recent Questions")
+# ax.set_ylabel("Time Taken for Response")
+# ax.legend()
+# st.sidebar.pyplot(fig)
+response_time = [q["response_time"] for q in recent_data["questions"]]
 labels = [f"Q{i+1}" for i in range(len(response_time))]  # Labels for X-axis
 
 fig, ax = plt.subplots()
-#ax.plot(labels, context_relevance, marker="o", label="Context Relevance")
-#ax.plot(labels, response_time, marker="s", label="Response Time (sec)")
+ax.plot(labels, response_time, color="skyblue")
 ax.set_xlabel("Recent Questions")
-ax.set_ylabel("Time Taken for Response")
-ax.legend()
-st.sidebar.pyplot(fig)
+ax.set_ylabel("Time Taken for Response (seconds)")
+ax.set_title("Response Time Analysis")
+
 
 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"])
+    st.sidebar.write(f"🔹 {q}")
 
 # Submit Button
 # if st.button("Submit"):
@@ -105,13 +110,7 @@ if st.button("Submit"):
     st.session_state.response = generate_response_from_document(question, st.session_state.retrieved_documents)
     end_time = time.time()
     st.session_state.time_taken_for_response = end_time - start_time
-
-    # Calculate metrics
-    st.session_state.metrics = calculate_metrics(question, st.session_state.response, st.session_state.retrieved_documents, st.session_state.time_taken_for_response)
-
-    # Save question & metrics
-    save_recent_question(question, st.session_state.metrics)
-
+    save_recent_question(question, st.session_state.time_taken_for_response)
 
 # Display stored response
 st.subheader("Response")
@@ -134,10 +133,10 @@ col1, col2 = st.columns([1, 3])  # Creating two columns for button and metrics d
 # Calculate Metrics Button
 with col1:
     if st.button("Show Metrics"):
-        metrics_ = st.session_state.metrics
+        st.session_state.metrics = calculate_metrics(question, st.session_state.response, st.session_state.retrieved_documents, st.session_state.time_taken_for_response)
     else:
         metrics_ = {}
 
 with col2:
     #st.text_area("Metrics:", value=metrics, height=100, disabled=True)
-    st.json(metrics_)
+    st.json(st.session_state.metrics )

data_processing.py

@@ -93,13 +93,13 @@ def load_ragbench():
         ragbench[dataset] = load_dataset("rungalileo/ragbench", dataset)
     return ragbench
 
-def load_query_dataset(query_dataset):
+def load_query_dataset(q_dataset):
     global query_dataset_data
-    if query_dataset_data[query_dataset]:
-        return query_dataset_data[query_dataset] 
+    if query_dataset_data[q_dataset]:
+        return query_dataset_data[q_dataset] 
     else:
-        query_dataset_data[query_dataset] = load_dataset("rungalileo/ragbench", query_dataset)
-    return query_dataset_data[query_dataset]
+        query_dataset_data[q_dataset] = load_dataset("rungalileo/ragbench", q_dataset)
+    return query_dataset_data[q_dataset]
 
 def load_faiss(query_dataset):
     global index
@@ -136,40 +136,18 @@ def load_recent_questions():
             return json.load(file)
     return {"questions": []}  # Default structure if file doesn't exist
 
-def save_recent_question(question, metrics):
+def save_recent_question(question, response_time):
     data = load_recent_questions()
 
     # Append new question & metrics
     data["questions"].append({
         "question": question,
-        "metrics": metrics
+        "response_time": response_time
     })
     
     # Keep only the last 5 questions
     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

@@ -102,6 +102,7 @@ def calculate_metrics(question, response, docs, time_taken):
 
     # Predicted metrics
     predicted_metrics = {
+        "RAG_model_response": response,
         "ground_truth": ground_truth_answer,
         "context_relevance": context_relevance(question, docs),
         "context_utilization": context_utilization(response, docs),

retrieval.py

@@ -9,8 +9,6 @@ from sentence_transformers import CrossEncoder
 reranker = CrossEncoder("cross-encoder/ms-marco-MiniLM-L-6-v2")
 
 retrieved_docs = None
-query_dataset = 'hotpotqa'
-
 
 def retrieve_documents_hybrid(query, top_k=5):
     global query_dataset