Update app.py

app.py

@@ -3,55 +3,69 @@ import requests
 import gradio as gr
 from huggingface_hub import InferenceClient
 
-# Initialize Inference Client
 client = InferenceClient("meta-llama/Llama-3.3-70B-Instruct", token=os.getenv("HF_TOKEN"))
 
-def fetch_catalog(query):
+def fetch_catalog_detailed(query):
     try:
-        url = f"https://openlibrary.org/search.json?q={query}&limit=3"
+        url = f"https://openlibrary.org/search.json?q={query}&limit=5"
         response = requests.get(url, timeout=5)
-        data = response.json().get("docs", [])
-        if not data:
-            return "No matches found."
-        return "\n".join([f"- {d.get('title')} by {', '.join(d.get('author_name', ['Unknown']))}" for d in data])
+        docs = response.json().get("docs", [])
+        
+        books = []
+        for d in docs:
+            books.append({
+                "title": d.get("title", "Unknown"),
+                "author": ", ".join(d.get("author_name", ["Unknown"])),
+                "year": d.get("first_publish_year", "N/A"),
+                "popularity": d.get("edition_count", 0)
+            })
+        return books
     except:
-        return "Search service offline."
+        return []
 
-def librarian_agent(message, history, session_state):
-    # Standardize state initialization
+def librarian_agent(message, history, session_state, sort_by):
     if session_state is None:
         session_state = {"verified": False, "age": None}
 
-    # --- PHASE 1: ETHICAL GATE ---
+    # --- ETHICAL GATE ---
     if not session_state["verified"]:
         age_input = "".join(filter(str.isdigit, message))
         if age_input:
-            age = int(age_input)
-            if age < 13:
-                reply = "🛡️ Safety: Access restricted for users under 13."
-            else:
-                session_state["age"] = age
-                session_state["verified"] = True
-                reply = f"✅ Verified (Age {age}). How can I help you today?"
+            session_state["age"] = int(age_input)
+            session_state["verified"] = True
+            reply = "✅ Verified. What topic are you interested in?"
         else:
-            reply = "👋 Please enter your age as a number to start."
-        
-        # history is a list of {"role": "...", "content": "..."}
-        history.append({"role": "user", "content": message})
+            reply = "👋 Please enter your age to start."
         history.append({"role": "assistant", "content": reply})
         return history, session_state
 
-    # --- PHASE 2: RAG PIPELINE ---
-    catalog_context = fetch_catalog(message)
-    
-    # Building the internal LLM prompt
+    # --- RETRIEVAL & SORTING ---
+    raw_books = fetch_catalog_detailed(message)
+    if sort_by == "Newest First":
+        raw_books.sort(key=lambda x: x['year'] if isinstance(x['year'], int) else 0, reverse=True)
+    elif sort_by == "Popularity":
+        raw_books.sort(key=lambda x: x['popularity'], reverse=True)
+
+    # Creating the data string for the LLM
+    catalog_str = ""
+    for b in raw_books:
+        catalog_str += f"Title: {b['title']}, Author: {b['author']}, Year: {b['year']}\n"
+
+    # --- INSTRUCTING THE LLM FOR TABULAR SUMMARY ---
     llm_messages = [
-        {"role": "system", "content": f"You are a librarian for a {session_state['age']} year old. Use: {catalog_context}"},
-        {"role": "user", "content": message}
+        {
+            "role": "system", 
+            "content": (
+                f"You are a helpful librarian for a {session_state['age']}-year-old. "
+                "Output a Markdown table with EXACTLY these columns: | Book & Author | Year | Summary |."
+                "The Summary must be 1-2 concise sentences. Be professional and age-appropriate."
+            )
+        },
+        {"role": "user", "content": f"Books found:\n{catalog_str}\n\nUser Question: {message}"}
     ]
 
     try:
-        output = client.chat_completion(messages=llm_messages, max_tokens=1000)
+        output = client.chat_completion(messages=llm_messages, max_tokens=1200)
         bot_res = output.choices[0].message.content
     except Exception as e:
         bot_res = f"Service Error: {str(e)}"
@@ -60,27 +74,26 @@ def librarian_agent(message, history, session_state):
     history.append({"role": "assistant", "content": bot_res})
     return history, session_state
 
-# --- UI BLOCK (Minimalist) ---
+# --- UI CONFIGURATION ---
 with gr.Blocks() as demo:
-    gr.Markdown("## 📚 AI Librarian Agent")
-    
+    gr.Markdown("## ⚡ Fast-Track Librarian with AI Summaries")
     state = gr.State(None)
     
-    # We initialize with a system message in the new dict format
-    # We REMOVE 'type="messages"' to avoid the keyword error
+    with gr.Row():
+        sort_option = gr.Radio(
+            ["Relevance", "Newest First", "Popularity"], 
+            label="Sort by", 
+            value="Relevance"
+        )
+    
     chatbot = gr.Chatbot(
-        label="Librarian Consultation",
-        value=[{"role": "assistant", "content": "👋 Welcome! Please enter your **age** to begin."}]
+        label="Quick Catalog",
+        value=[{"role": "assistant", "content": "👋 Enter your **age** to begin."}]
     )
     
-    msg = gr.Textbox(label="Input", placeholder="Enter age (e.g. 25)")
-
-    def respond(user_input, chat_history, current_state):
-        # Gradio automatically detects the format if 'type' is omitted
-        updated_history, updated_state = librarian_agent(user_input, chat_history, current_state)
-        return "", updated_history, updated_state
+    msg = gr.Textbox(label="Search Query", placeholder="e.g. Science Fiction or History of Rome")
 
-    msg.submit(respond, [msg, chatbot, state], [msg, chatbot, state])
+    msg.submit(librarian_agent, [msg, chatbot, state, sort_option], [chatbot, state])
+    msg.submit(lambda: "", None, msg)
 
-# Launch without passing theme to Blocks, handle it here if needed
 demo.launch()