Upload 2 files

app.py

@@ -1,196 +1,144 @@
-# import streamlit as st
-# from generator import generate_response_from_document
-# from retrieval import retrieve_documents_hybrid
-# from evaluation import calculate_metrics
-# #from data_processing import load_data_from_faiss
-# import time
-
-# # Page Title
-# st.title("RAG7 - Real World RAG System")
-
-# # global retrieved_documents
-# # retrieved_documents = []
-
-# # global response 
-# # response = ""
-
-# # global time_taken_for_response
-# # time_taken_for_response = 'N/A'
-
-# # @st.cache_data
-# # def load_data():
-# #     load_data_from_faiss()
-
-# # data_status = load_data()
-
-# # Question Section
-# st.subheader("Hi, What do you want to know today?")
-# question = st.text_area("Enter your question:", placeholder="Type your question here...", height=100)
-
-# # # Submit Button
-# # if st.button("Submit"):
-# #     start_time = time.time()
-# #     retrieved_documents = retrieve_documents_hybrid(question, 10)  
-# #     response = generate_response_from_document(question, retrieved_documents)
-# #     end_time = time.time()
-# #     time_taken_for_response = end_time-start_time
-# # else:
-# #     response = ""
-
-# # # Response Section
-# # st.subheader("Response")
-# # st.text_area("Generated Response:", value=response, height=150, disabled=True)
-
-# # # Metrics Section
-# # st.subheader("Metrics")
-
-# # col1, col2 = st.columns([1, 3])  # Creating two columns for button and metrics display
-
-# # with col1:
-# #     if st.button("Calculate Metrics"):
-# #         metrics = calculate_metrics(question, response, retrieved_documents, time_taken_for_response)
-# #     else:
-# #         metrics = ""
-
-# # with col2:
-# #     st.text_area("Metrics:", value=metrics, height=100, disabled=True)
-
-# if "retrieved_documents" not in st.session_state:
-#     st.session_state.retrieved_documents = []
-# if "response" not in st.session_state:
-#     st.session_state.response = ""
-# if "time_taken_for_response" not in st.session_state:
-#     st.session_state.time_taken_for_response = "N/A"
+import streamlit as st
+from generator import generate_response_from_document
+from retrieval import retrieve_documents_hybrid
+from evaluation import calculate_metrics
+from data_processing import load_recent_questions, save_recent_question
+import time
+
+# Page Title
+st.title("RAG7 - Real World RAG System")
+
+# global retrieved_documents
+# retrieved_documents = []
+
+# global response 
+# response = ""
+
+# global time_taken_for_response
+# time_taken_for_response = 'N/A'
+
+# @st.cache_data
+# def load_data():
+#     load_data_from_faiss()
+
+# data_status = load_data()
+
+# Question Section
+st.subheader("Hi, What do you want to know today?")
+question = st.text_area("Enter your question:", placeholder="Type your question here...", height=100)
 
 # # Submit Button
 # if st.button("Submit"):
 #     start_time = time.time()
-#     st.session_state.retrieved_documents = retrieve_documents_hybrid(question, 10)  
-#     st.session_state.response = generate_response_from_document(question, st.session_state.retrieved_documents)
+#     retrieved_documents = retrieve_documents_hybrid(question, 10)  
+#     response = generate_response_from_document(question, retrieved_documents)
 #     end_time = time.time()
-#     st.session_state.time_taken_for_response = end_time - start_time
+#     time_taken_for_response = end_time-start_time
+# else:
+#     response = ""
 
-# # Display stored response
+# # Response Section
 # st.subheader("Response")
-# st.text_area("Generated Response:", value=st.session_state.response, height=150, disabled=True)
+# st.text_area("Generated Response:", value=response, height=150, disabled=True)
+
+# # Metrics Section
+# st.subheader("Metrics")
 
 # col1, col2 = st.columns([1, 3])  # Creating two columns for button and metrics display
 
-# # Calculate Metrics Button
 # with col1:
 #     if st.button("Calculate Metrics"):
-#         metrics = calculate_metrics(question, st.session_state.response, st.session_state.retrieved_documents, st.session_state.time_taken_for_response)
+#         metrics = calculate_metrics(question, response, retrieved_documents, time_taken_for_response)
 #     else:
-#         metrics = {}
+#         metrics = ""
 
 # with col2:
-#     #st.text_area("Metrics:", value=metrics, height=100, disabled=True)
-#     st.json(metrics)
+#     st.text_area("Metrics:", value=metrics, height=100, disabled=True)
 
-import streamlit as st
-import plotly.express as px
-from datasets import load_dataset, Dataset, DatasetDict
-from generator import generate_response_from_document
-from retrieval import retrieve_documents_hybrid
-from evaluation import calculate_metrics
-import time
+if "retrieved_documents" not in st.session_state:
+    st.session_state.retrieved_documents = []
+if "response" not in st.session_state:
+    st.session_state.response = ""
+if "time_taken_for_response" not in st.session_state:
+    st.session_state.time_taken_for_response = "N/A"
+if "metrics" not in st.session_state:
+    st.session_state.metrics = {}
 
-# Hugging Face Dataset Details
-HF_DATASET_REPO = "cb1716pics/23RAG7_recent_questions"  # Hugging Face repo
-
-# Load Dataset from Hugging Face
-@st.cache_resource
-def load_hf_dataset():
-    try:
-        return load_dataset(HF_DATASET_REPO)
-    except:
-        return DatasetDict({"recent": Dataset.from_dict({"question": [], "response": [], "metrics": []})})
-
-dataset = load_hf_dataset()
-
-# Function to Save Data to Hugging Face Dataset
-def save_to_hf_dataset(question, response, metrics):
-    global dataset
-    new_data = {
-        "question": [question],
-        "response": [response],
-        "metrics": [metrics]
-    }
-    
-    # Convert existing dataset to a list and append new data
-    dataset_dict = dataset["recent"].to_dict()
-    for key in new_data.keys():
-        dataset_dict[key] = dataset_dict.get(key, []) + new_data[key]
-    
-    # Keep only the last 10 entries
-    for key in dataset_dict.keys():
-        dataset_dict[key] = dataset_dict[key][-10:]
-
-    # Convert back to dataset and push to Hugging Face
-    dataset["recent"] = Dataset.from_dict(dataset_dict)
-    dataset.push_to_hub(HF_DATASET_REPO)
-
-# Streamlit UI
-st.title("🔍 RAG7 - Real World RAG System")
-
-# Sidebar - Recent Questions
-st.sidebar.header("📌 Recent Questions")
-if len(dataset["recent"]) > 0:
-    for q in dataset["recent"]["question"][-10:]:
-        st.sidebar.write(f"🔹 {q}")
-
-# Sidebar - Analytics with Graph
-st.sidebar.header("📊 Analytics Overview")
-if len(dataset["recent"]) > 0:
-    # Extract recent metrics for visualization
-    metrics_data = dataset["recent"]["metrics"][-10:]
-    metrics_keys = ["context_relevance", "context_utilization", "completeness", "adherence"]
-    
-    # Prepare a dictionary for graphing
-    graph_data = {key: [m[key] for m in metrics_data] for key in metrics_keys}
-    graph_data["Question #"] = list(range(1, len(metrics_data) + 1))
-
-    # Convert to DataFrame for Plotly
-    import pandas as pd
-    df = pd.DataFrame(graph_data)
-
-    # Plot Metrics Over Time
-    fig = px.line(df, x="Question #", y=metrics_keys, 
-                  labels={"value": "Score", "variable": "Metric"},
-                  title="📈 Model Performance Over Recent Questions")
-    st.sidebar.plotly_chart(fig, use_container_width=True)
-
-# Evaluate Button
-if st.sidebar.button("⚡ Evaluate RAG Model"):
-    st.sidebar.success("✅ Model Evaluation Triggered!")
-
-# Main Section - User Input
-st.subheader("💬 Ask a Question")
-question = st.text_area("Enter your question:", placeholder="Type your question here...", height=100)
+# 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.sidebar.markdown("---")  # Separator
+
+
+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"]]
+labels = [f"Q{i+1}" for i in range(len(context_relevance))]  # 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.set_xlabel("Recent Questions")
+ax.set_ylabel("Scores")
+ax.legend()
+st.sidebar.pyplot(fig)
 
 # Submit Button
-if st.button("🚀 Submit"):
+# if st.button("Submit"):
+#     start_time = time.time()
+#     st.session_state.retrieved_documents = retrieve_documents_hybrid(question, 10)  
+#     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
+
+if st.button("Submit"):
     start_time = time.time()
-    retrieved_documents = retrieve_documents_hybrid(question, 10)  
-    response = generate_response_from_document(question, retrieved_documents)
+    st.session_state.retrieved_documents = retrieve_documents_hybrid(question, 10)  
+    st.session_state.response = generate_response_from_document(question, st.session_state.retrieved_documents)
     end_time = time.time()
-    time_taken_for_response = end_time - start_time
+    st.session_state.time_taken_for_response = end_time - start_time
 
-    # Calculate Metrics
-    metrics = calculate_metrics(question, response, retrieved_documents, time_taken_for_response)
+    # 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 Data
-    save_to_hf_dataset(question, response, metrics)
+    # Save question & metrics
+    save_recent_question(question, st.session_state.metrics)
+
+
+# Display stored response
+st.subheader("Response")
+st.text_area("Generated Response:", value=st.session_state.response, height=150, disabled=True)
+
+col1, col2 = st.columns([1, 3])  # Creating two columns for button and metrics display
+
+# # Calculate Metrics Button
+# with col1:
+#     if st.button("Calculate Metrics"):
+#         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)
 
-    # Display Response
-    st.subheader("💡 Response")
-    st.text_area("Generated Response:", value=response, height=150, disabled=True)
 
-    # Display Metrics with Bar Chart
-    st.subheader("📊 Metrics")
-    st.json(metrics)
+# Calculate Metrics Button
+with col1:
+    if st.button("Show Metrics"):
+        metrics_ = st.session_state.metrics
+    else:
+        metrics_ = {}
 
-    # Plot Bar Chart for Metrics
-    metric_df = pd.DataFrame({"Metric": list(metrics.keys()), "Score": list(metrics.values())})
-    fig2 = px.bar(metric_df, x="Metric", y="Score", title="📊 Current Query Metrics")
-    st.plotly_chart(fig2, use_container_width=True)
+with col2:
+    #st.text_area("Metrics:", value=metrics, height=100, disabled=True)
+    st.json(metrics_)

data_processing.py

@@ -20,6 +20,14 @@ embedding_model = HuggingFaceEmbeddings(
 
 reranker = CrossEncoder("cross-encoder/ms-marco-MiniLM-L-6-v2")
 
+# File path for storing recently asked questions and metrics
+RECENT_QUESTIONS_FILE = "data_local/recent_questions.json"
+
+# Ensure the file exists and initialize if empty
+if not os.path.exists(RECENT_QUESTIONS_FILE):
+    with open(RECENT_QUESTIONS_FILE, "w") as file:
+        json.dump({"questions": []}, file, indent=4)
+
 all_documents = []
 ragbench = {}
 index = None  
@@ -106,7 +114,6 @@ def load_chunks(query_dataset):
 def load_data_from_faiss(query_dataset):
     load_faiss(query_dataset)
     load_chunks(query_dataset)
-    #return index_, chunks_  
 
 def rerank_documents(query, retrieved_docs):
     doc_texts = [doc for doc in retrieved_docs]
@@ -114,3 +121,24 @@ def rerank_documents(query, retrieved_docs):
     ranked_docs = [doc for _, doc in sorted(zip(scores, retrieved_docs), reverse=True)]
     return ranked_docs[:5]  # Return top 5 most relevant
 
+def load_recent_questions():
+    if os.path.exists(RECENT_QUESTIONS_FILE):
+        with open(RECENT_QUESTIONS_FILE, "r") as file:
+            return json.load(file)
+    return {"questions": []}  # Default structure if file doesn't exist
+
+def save_recent_question(question, metrics):
+    data = load_recent_questions()
+
+    # Append new question & metrics
+    data["questions"].append({
+        "question": question,
+        "metrics": metrics
+    })
+
+    # 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)