Update app.py

app.py

@@ -1,135 +1,104 @@
-import os
 import json
+import os
 import difflib
-import streamlit as st
+import gradio as gr
 
-# ===============================
-# Load quotes
-# ===============================
+# Path to your JSON data
 DATA_PATH = "quotes.json"
 
+# Load quotes
 def load_quotes():
     if os.path.exists(DATA_PATH):
-        try:
-            with open(DATA_PATH, "r", encoding="utf-8") as f:
-                data = json.load(f)
-            categories = data.get("categories", {})
-            return categories
-        except Exception as e:
-            st.error(f"Error loading data: {e}")
-            return {}
-    else:
-        st.error("quotes.json not found")
-        return {}
+        with open(DATA_PATH, "r", encoding="utf-8") as f:
+            return json.load(f)
+    return {}
 
 QUOTES = load_quotes()
 
-# ===============================
-# Fuzzy Matching
-# ===============================
-def find_best_match(user_input, category_quotes):
-    all_quotes = []
-    for sentiment, quotes in category_quotes.items():
-        for quote in quotes:
-            all_quotes.append((quote, sentiment))
-
-    if not all_quotes:
-        return None, None
-
-    # Match user input against quotes
-    quotes_only = [q[0] for q in all_quotes]
-    match = difflib.get_close_matches(user_input, quotes_only, n=1, cutoff=0.3)
-
-    if match:
-        for q, s in all_quotes:
-            if q == match[0]:
-                return q, s
-
-    return None, None
-
-# ===============================
-# Response Formatter
-# ===============================
-def generate_response(user_input, category, sentiment, domain):
+# Fuzzy match logic (improved with tokenization)
+def fuzzy_match(query, choices):
+    tokens = query.lower().split()
+    best_score = 0
+    best_match = None
+    for choice in choices:
+        score = difflib.SequenceMatcher(None, query.lower(), choice.lower()).ratio()
+        token_overlap = len(set(tokens) & set(choice.lower().split())) / max(len(tokens), 1)
+        combined_score = (score + token_overlap) / 2
+        if combined_score > best_score:
+            best_score = combined_score
+            best_match = choice
+    return best_match, best_score
+
+# Handle user query
+def handle_query(user_message, category, sentiment, url_restrict):
+    responses = []
+
+    # Validate category
     if category not in QUOTES:
-        return "Summary not available.", "No data in this category.", "No related articles."
-
-    category_quotes = QUOTES[category]
+        return [{"role": "assistant", "content": "Category not found in dataset."}]
 
-    # Handle auto sentiment
+    # Collect relevant quotes
+    choices = []
     if sentiment == "auto":
-        all_quotes = []
-        for quotes in category_quotes.values():
-            all_quotes.extend(quotes)
-        if not all_quotes:
-            return "Summary not available.", "No quotes available.", "No related articles."
+        for sent in QUOTES[category]:
+            choices.extend(QUOTES[category][sent])
     else:
-        all_quotes = category_quotes.get(sentiment, [])
-        if not all_quotes:
-            return "Summary not available.", f"No {sentiment} quotes found.", "No related articles."
+        choices = QUOTES[category].get(sentiment, [])
 
-    # Best match
-    best_quote, best_sentiment = find_best_match(user_input, category_quotes)
+    if not choices:
+        return [{"role": "assistant", "content": "No quotes available for this category/sentiment."}]
 
-    # 3-tier response
-    summary_response = f"The topic seems to be about **{category.lower()}**."
-    if best_quote:
-        details_response = best_quote
-    else:
-        details_response = "No exact matching quotes found, but here’s what people say in general."
-
-    # Fusion = simple join of first two available quotes
-    fusion_response = ""
-    if len(all_quotes) >= 2:
-        fusion_response = all_quotes[0] + " " + all_quotes[1]
-    elif all_quotes:
-        fusion_response = all_quotes[0]
-    else:
-        fusion_response = "No quotes available for fusion."
+    # Fuzzy match query against quotes
+    best_match, score = fuzzy_match(user_message, choices)
+
+    # === 3-tier response ===
+    # Tier 1: Summary (simple echo of category/topic)
+    summary = f"**Topic Summary:** This question seems related to *{category.replace('_', ' ')}*."
 
-    # Related article (placeholder link with domain search)
-    if domain.strip():
-        related_link = f"[Search related articles on {domain}](https://www.google.com/search?q=site:{domain}+{category})"
+    # Tier 2: "What real people say"
+    if best_match and score > 0.3:
+        details = f"**What real people say:**\n{best_match}"
     else:
-        related_link = f"[Search related articles online](https://www.google.com/search?q={category})"
+        details = "**What real people say:**\nSorry, no close match found."
 
-    return summary_response, details_response, fusion_response, related_link
+    # Tier 3: External article reference (stub)
+    if url_restrict.strip():
+        external = f"**Similar articles (restricted to {url_restrict}):**\n[Search results on {url_restrict}](https://www.google.com/search?q={user_message}+site:{url_restrict})"
+    else:
+        external = "**Similar articles:**\n[Search on Google](https://www.google.com/search?q=" + user_message.replace(" ", "+") + ")"
 
-# ===============================
-# Streamlit UI
-# ===============================
-st.set_page_config(page_title="Campus Life Quotes", layout="centered")
+    responses.extend([
+        {"role": "assistant", "content": summary},
+        {"role": "assistant", "content": details},
+        {"role": "assistant", "content": external}
+    ])
 
-st.title("🎓 Campus Life Quotes Explorer")
+    return responses
 
-# Align dropdowns side by side
-categories = sorted(list(QUOTES.keys()))
-col1, col2 = st.columns(2)
-with col1:
-    category = st.selectbox("Category", categories)
-with col2:
-    sentiment = st.selectbox("Sentiment", ["auto", "positive", "negative"])
+# Build Gradio UI
+with gr.Blocks() as demo:
+    gr.Markdown("## 🎓 College Life Q&A Chatbot")
 
-user_input = st.text_input("Ask a question:", "")
+    with gr.Row():
+        category = gr.Dropdown(choices=list(QUOTES.keys()), label="Category", interactive=True)
+        sentiment = gr.Dropdown(choices=["auto", "positive", "negative"], value="auto", label="Sentiment", interactive=True)
 
-# New: input domain for related articles
-domain = st.text_input("Enter a domain for related articles (optional)", "")
+    url_restrict = gr.Textbox(label="Restrict search to domain (optional)", placeholder="e.g., nytimes.com")
 
-if st.button("Get Response"):
-    if user_input.strip() == "":
-        st.warning("Please enter a question.")
-    else:
-        summary, details, fusion, related = generate_response(user_input, category, sentiment, domain)
+    chatbot = gr.Chatbot(label="Conversation", type="messages", height=400)
+    msg = gr.Textbox(label="Ask a question")
 
-        # New formatted outputs
-        st.markdown("**Summary:**")
-        st.write(summary)
+    with gr.Row():
+        clear_btn = gr.Button("Clear Chat")
 
-        st.markdown("**What real people say:**")
-        st.write(details)
+    def respond(message, chat_history, category, sentiment, url_restrict):
+        bot_msgs = handle_query(message, category, sentiment, url_restrict)
+        chat_history.append({"role": "user", "content": message})
+        chat_history.extend(bot_msgs)
+        return "", chat_history
 
-        st.markdown("**Fusion:**")
-        st.write(fusion)
+    msg.submit(respond, [msg, chatbot, category, sentiment, url_restrict], [msg, chatbot])
+    clear_btn.click(lambda: [], None, chatbot)
 
-        st.markdown("**Related Articles:**")
-        st.markdown(related)
+if __name__ == "__main__":
+    demo.launch()