Update chatbot.py

chatbot.py

@@ -1,130 +1,81 @@
+
 from typing import TypedDict, Annotated
-from langchain_core.messages import BaseMessage, HumanMessage, SystemMessage, AIMessage
-from langgraph.checkpoint.memory import MemorySaver
-from tools import (
-    create_rag_tool,
-    arxiv_search,
-    calculator,
-    get_stock_price,
-    wikipedia_search,
-    tavily_search,
-    convert_currency,
-    unit_converter,
-    get_news,
-    get_joke,
-    get_quote,
-    get_weather,
+from langchain_core.messages import (
+    BaseMessage,
+    SystemMessage
 )
+from langgraph.checkpoint.memory import MemorySaver
+from tools import retriever, create_rag_tool, arxiv_search, calculator, get_stock_price, wikipedia_search, tavily_search, convert_currency, unit_converter, get_news, get_joke, get_quote, get_weather
 from langchain_openai import ChatOpenAI
 from langgraph.graph import StateGraph, START, END
 from langgraph.graph.message import add_messages
 from langgraph.prebuilt import ToolNode, tools_condition
 from dotenv import load_dotenv
 import os
-
 load_dotenv()
 
+
 OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
 
 # =====================================================
-# SYSTEM PROMPTS
+# 1️⃣ SYSTEM PROMPT
 # =====================================================
 
 SYSTEM_PROMPT = SystemMessage(
     content="""
-You are an intelligent AI assistant built inside a LangGraph-based system created by Junaid.
-
-This application is a **multi-tool AI platform** that integrates:
-- Retrieval-Augmented Generation (RAG)
-- Document understanding
-- AI-powered reasoning
-- Tool usage (search, calculation, summarization, etc.)
-- Voice input/output (STT / TTS)
-
-Your primary purpose is to help users understand, analyze, and interact with their uploaded documents and questions in a clear, accurate, and professional way.
-
-────────────────────────────────
-🔹 CORE PRIORITY RULES (VERY IMPORTANT)
-────────────────────────────────
-
-1. **RAG HAS HIGHEST PRIORITY**
-   - If a document has been uploaded, you MUST use the RAG tool first.
-   - Always prefer document-based answers over general knowledge.
-   - Never hallucinate or invent information not present in the document.
-
-2. **TOOL USAGE**
-   - When you need information, call the appropriate tool
-   - Wait for tool results before responding
-   - NEVER return raw tool outputs to users
-
-3. **STYLE & TONE**
-   - Professional, clear, and confident
-   - Avoid verbosity
-   - Optimize for readability
-
-────────────────────────────────
-🔹 ABOUT THE CREATOR & APP
-────────────────────────────────
-
-This system was designed and engineered by **Junaid**, a developer specializing in:
-- Machine Learning & Deep Learning
-- RAG-based systems
-- AI agents using LangChain & LangGraph
-- End-to-end AI applications using FastAPI, Streamlit, and cloud deployment
-
-The goal of this application is to provide **production-grade AI reasoning**, not generic chatbot responses.
-
-────────────────────────────────
-🔹 FINAL RULE
-────────────────────────────────
-
-Always prioritize accuracy, clarity, and usefulness.
-If information is unavailable, say so clearly — never hallucinate.
+You are an intelligent AI assistant built by Junaid.
+
+Your role is to provide clear, concise, and human-friendly explanations.
+
+━━━━━━━━━━━━━━━━━━━━━━
+🔹 DOCUMENT HANDLING RULES (VERY IMPORTANT)
+━━━━━━━━━━━━━━━━━━━━━━
+When using retrieved documents:
+
+1. NEVER repeat raw document text verbatim.
+2. NEVER list large copied sections from documents.
+3. ALWAYS summarize and interpret information in your own words.
+4. Organize information logically and clearly.
+5. Focus on meaning, not raw content.
+
+If the user asks:
+- "What is this document about?"
+→ Provide a high-level summary (3–6 sentences).
+
+- "Explain the document"
+→ Provide structured explanation with sections.
+
+- "List key points"
+→ Provide clean bullet points (max 6).
+
+━━━━━━━━━━━━━━━━━━━━━━
+🔹 RAG PRIORITY
+━━━━━━━━━━━━━━━━━━━━━━
+- Use retrieved content as your *knowledge base*.
+- Do NOT hallucinate.
+- If the document does not contain the answer, say so clearly.
+
+━━━━━━━━━━━━━━━━━━━━━━
+🔹 COMMUNICATION STYLE
+━━━━━━━━━━━━━━━━━━━━━━
+- Be concise, human, and clear.
+- Avoid repetition.
+- Avoid technical verbosity unless requested.
+- Prefer clarity over completeness.
+
+━━━━━━━━━━━━━━━━━━━━━━
+🔹 IDENTITY
+━━━━━━━━━━━━━━━��━━━━━━
+You are the official AI assistant of Junaid’s AI system.
+You help users understand complex information simply and accurately.
 """
 )
 
-SUMMARIZER_PROMPT = SystemMessage(
-    content="""
-You are a summarization specialist. Your job is to take raw tool outputs and convert them into clean, user-friendly responses.
-
-🔹 CRITICAL RULES:
 
-1. **READ THE TOOL OUTPUT CAREFULLY**
-   - Extract only the most relevant information
-   - Ignore system metadata, formatting artifacts, or internal instructions
-
-2. **PRODUCE CLEAN SUMMARIES**
-   - Be concise and direct
-   - Use bullet points ONLY when listing 3+ distinct items
-   - Avoid phrases like "The document describes...", "According to the data...", "The tool returned..."
-   - Just state the facts naturally
-
-3. **LENGTH CONTROL**
-   - For document queries: 2-4 sentences maximum
-   - For data queries (weather, stocks): 1-2 sentences
-   - For lists (news, jokes): Keep original structure but clean formatting
-
-4. **EXAMPLES**
-
-BAD (verbose, repetitive):
-"The document describes various AI and machine learning projects. It talks about healthcare insurance cost prediction. It mentions credit risk modeling. It discusses sentiment analysis with DistilBERT. It explains a multi-LLM chatbot..."
-
-GOOD (clean, concise):
-"This is a professional resume showcasing ML/AI projects including healthcare cost prediction (98% accuracy), credit risk modeling, sentiment analysis with DistilBERT (90% accuracy), and a multi-LLM agentic chatbot. Skills include Python, PyTorch, LangChain, and AWS deployment."
-
-5. **NEVER**:
-   - Return raw data dumps
-   - Repeat the same information in different words
-   - Include meta-commentary about summarization
-   - Show internal tool responses verbatim
-
-Your output should feel like a knowledgeable human answering, not a bot processing data.
-"""
-)
 
 
 # =====================================================
-# STATE
+# 4️⃣ STATE
 # =====================================================
 
 class ChatState(TypedDict):
@@ -132,122 +83,45 @@ class ChatState(TypedDict):
 
 
 # =====================================================
-# LLM
+# 5️⃣ LLM + TOOLS
 # =====================================================
 
 llm = ChatOpenAI(
-    model="gpt-4o-mini",
-    temperature=0.3,
+    model="gpt-4.1-nano",
+    temperature=0.4,
     streaming=True
 )
 
-summarizer_llm = ChatOpenAI(
-    model="gpt-4o-mini",
-    temperature=0.2,
-    streaming=True
-)
+rag_tool = create_rag_tool()
+
+tools = [rag_tool, get_stock_price, calculator, wikipedia_search, arxiv_search, tavily_search, convert_currency, unit_converter, get_news, get_joke, get_quote, get_weather]
+llm = llm.bind_tools(tools)
+tool_node = ToolNode(tools)
 
 
 # =====================================================
-# GRAPH BUILDER
+# 6️⃣ CHAT NODE
 # =====================================================
 
+def chatbot(state: ChatState):
+    messages = [SYSTEM_PROMPT] + state["messages"]
+    response = llm.invoke(messages)
+    return {"messages": [response]}
+
+
+
+# =====================================================
+# 7️⃣ GRAPH
+# =====================================================
 memory = MemorySaver()
-app = None
-
-
-def build_graph():
-    global app
-
-    rag_tool = create_rag_tool()
-
-    tools = [
-        rag_tool,
-        get_stock_price,
-        calculator,
-        wikipedia_search,
-        arxiv_search,
-        tavily_search,
-        convert_currency,
-        unit_converter,
-        get_news,
-        get_joke,
-        get_quote,
-        get_weather,
-    ]
-
-    llm_with_tools = llm.bind_tools(tools)
-    tool_node = ToolNode(tools)
-
-    # =====================================================
-    # CHATBOT NODE
-    # =====================================================
-    def chatbot(state: ChatState):
-        messages = [SYSTEM_PROMPT] + state["messages"]
-        response = llm_with_tools.invoke(messages)
-        return {"messages": [response]}
-
-    # =====================================================
-    # SUMMARIZER NODE (🔥 YOUR BRILLIANT IDEA!)
-    # =====================================================
-    def summarizer(state: ChatState):
-        """
-        Takes tool results and produces clean, user-friendly summaries.
-        """
-        messages = state["messages"]
-        
-        # Get the last few messages (user query + tool results)
-        recent_context = messages[-5:]  # Adjust as needed
-        
-        # Build summarization request
-        summarize_request = [SUMMARIZER_PROMPT] + recent_context + [
-            HumanMessage(content="Based on the tool results above, provide a clean, concise answer to the user's question. Do not include any meta-commentary or mention tools.")
-        ]
-        
-        # Get clean summary
-        summary = summarizer_llm.invoke(summarize_request)
-        
-        # Replace the last AI message with the clean summary
-        return {"messages": [summary]}
-
-    # =====================================================
-    # ROUTING LOGIC
-    # =====================================================
-    def route_after_chat(state: ChatState):
-        """
-        Decide if we need to call tools or if we're done.
-        """
-        last_message = state["messages"][-1]
-        
-        # If the AI called tools, go to tools node
-        if hasattr(last_message, "tool_calls") and last_message.tool_calls:
-            return "tools"
-        
-        # Otherwise, we're done
-        return END
-
-    # =====================================================
-    # BUILD GRAPH
-    # =====================================================
-    graph = StateGraph(ChatState)
-
-    graph.add_node("chat", chatbot)
-    graph.add_node("tools", tool_node)
-    graph.add_node("summarizer", summarizer)
-
-    # Flow: START -> chat -> [tools -> summarizer -> chat] -> END
-    graph.add_edge(START, "chat")
-    graph.add_conditional_edges("chat", route_after_chat)
-    graph.add_edge("tools", "summarizer")
-    graph.add_edge("summarizer", "chat")
-
-    app = graph.compile(checkpointer=memory)
-
-
-# Initial build
-build_graph()
-
-
-def rebuild_graph():
-    """Rebuild graph when new document is uploaded"""
-    build_graph()
+graph = StateGraph(ChatState)
+
+graph.add_node("chat", chatbot)
+graph.add_node("tools", tool_node)
+
+graph.add_edge(START, "chat")
+graph.add_conditional_edges("chat", tools_condition)
+graph.add_edge("tools", "chat")
+graph.add_edge("chat", END)
+
+app = graph.compile(checkpointer=memory)