Upload 2 files

app.py

@@ -0,0 +1,83 @@
+import streamlit as st
+from deep_research_system import run_deep_research_system
+import traceback
+import streamlit.components.v1 as components
+ 
+if __name__ == "__main__":
+    # Define API base URL (Update for Hugging Face deployment)
+    API_URL = "http://localhost:8000/"
+
+    # Set page configuration
+    st.set_page_config(page_title="Deep Research AI", layout="centered")
+
+    # Title and instructions
+    st.title("Deep Research AI Agentic System")
+    st.write("Enter a question below to get the latest insights from web research. Click 'Reset' to start over.")
+
+    # Initialize session state
+    if "show_reset_button" not in st.session_state:
+        st.session_state.show_reset_button = False
+    if "question" not in st.session_state:
+        st.session_state.question = ""
+    if "reset_triggered" not in st.session_state:
+        st.session_state.reset_triggered = False
+
+    # JavaScript to clear the input field
+    clear_input_js = """
+    <script>
+        const input = document.querySelector('input[aria-label="Your Question"]');
+        if (input) {
+            input.value = '';
+        }
+    </script>
+    """
+
+    # Use a form to manage the question input and submission
+    with st.form(key="question_form"):
+        st.session_state.question = st.text_input(
+            "Your Question",
+            placeholder="e.g., What are the latest advancements in quantum computing?",
+            value=st.session_state.question,
+            key="question_input"
+        )
+        submit_button = st.form_submit_button("Get Answer")
+
+    # Process the form submission
+    if submit_button:
+        if st.session_state.question:
+            st.write(f"Research Agent: Searching for '{st.session_state.question}'...")
+            try:
+                with st.spinner("Gathering research data..."):
+                    answer = run_deep_research_system(st.session_state.question)
+                st.write("Research Agent: Found 5 relevant sources.")
+                st.write("Answer Drafter Agent: Drafted the final answer.")
+                st.write("**Final Answer:**")
+                st.write(answer)
+                st.session_state.show_reset_button = True
+            except Exception as e:
+                st.error(f"An error occurred: {str(e)}\n{traceback.format_exc()}")
+                st.session_state.show_reset_button = True
+        else:
+            st.warning("Please enter a question!")
+        # Reset the trigger flag after submission
+        st.session_state.reset_triggered = False
+
+    # Function to clear the input state
+    def clear_input():
+        st.session_state.show_reset_button = False
+        st.session_state.question = ""
+        st.session_state.pop("question_input", None)
+        st.session_state.pop("question_form", None)
+        st.session_state.reset_triggered = True
+
+    # Show Reset button only if show_reset_button is True
+    if st.session_state.show_reset_button:
+        if st.button("Reset"):
+            # Clear the input and reset state
+            clear_input()
+            # Refresh the webpage
+            st.rerun()
+
+    # Execute JavaScript to clear the input field if reset was triggered
+    if st.session_state.reset_triggered:
+        components.html(clear_input_js, height=0)

deep_research_system.py

@@ -0,0 +1,126 @@
+import os
+from dotenv import load_dotenv
+from tavily import TavilyClient
+from langchain_openai import ChatOpenAI
+from typing import Dict, TypedDict, List
+from langgraph.graph import StateGraph, END
+from langchain_core.prompts import ChatPromptTemplate
+
+# Load environment variables
+load_dotenv()
+
+OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
+TAVILY_API_KEY = os.getenv("TAVILY_API_KEY")
+
+# Check if API keys are set
+try:
+    if not OPENAI_API_KEY or not TAVILY_API_KEY:
+        raise ValueError("API keys for OpenAI and Tavily must be set in environment variables.")
+except ValueError as e:
+    print(f"Error: {e}")
+    raise
+
+# Initialize the LLM and Tavily client
+try:
+    # Initialize the LLM (using OpenAI as an example, replace with your preferred model)
+    llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
+    # Initialize Tavily client for web search
+    tavily_client = TavilyClient(api_key=os.environ["TAVILY_API_KEY"])
+except Exception as e:
+    print(f"Error initializing LLM or Tavily client: {e}")
+    raise
+
+# Define the shared state for the agents
+class ResearchState(TypedDict):
+    query: str
+    research_data: List[Dict]
+    final_answer: str
+
+# Research Agent: Crawls the web and gathers information
+def research_agent(state: ResearchState) -> ResearchState:
+    query = state["query"]
+    print(f"Research Agent: Searching for '{query}'...")
+
+    # Use Tavily to search the web
+    search_results = tavily_client.search(query, max_results=5)
+
+    # Extract relevant information from search results
+    research_data = []
+    for result in search_results["results"]:
+        research_data.append({
+            "title": result["title"],
+            "url": result["url"],
+            "content": result["content"][:500]  # Limit content length for brevity
+        })
+
+    print(f"Research Agent: Found {len(research_data)} relevant sources.")
+    return {"research_data": research_data}
+
+# Answer Drafter Agent: Processes research data and drafts a response
+def answer_drafter_agent(state: ResearchState) -> ResearchState:
+    research_data = state["research_data"]
+    query = state["query"]
+
+    # Create a prompt for the answer drafter
+    prompt = ChatPromptTemplate.from_template(
+        """
+        You are an expert at drafting concise and accurate answers. Based on the following research data, provide a clear and informative response to the query: "{query}".
+
+        Research Data:
+        {research_data}
+
+        Provide a well-structured answer in 3-5 sentences, citing the sources where relevant.
+        """
+    )
+
+    # Format the research data for the prompt
+    research_text = "\n".join([f"- {item['title']}: {item['content']} (Source: {item['url']})" for item in research_data])
+    chain = prompt | llm
+
+    # Generate the final answer
+    response = chain.invoke({"query": query, "research_data": research_text})
+    final_answer = response.content
+
+    print("Answer Drafter Agent: Drafted the final answer.")
+    return {"final_answer": final_answer}
+
+# Define the LangGraph workflow
+def create_workflow():
+    workflow = StateGraph(ResearchState)
+
+    # Add nodes for each agent
+    workflow.add_node("research_agent", research_agent)
+    workflow.add_node("answer_drafter_agent", answer_drafter_agent)
+
+    # Define the flow: Research Agent -> Answer Drafter Agent -> End
+    workflow.add_edge("research_agent", "answer_drafter_agent")
+    workflow.add_edge("answer_drafter_agent", END)
+
+    # Set the entry point
+    workflow.set_entry_point("research_agent")
+
+    return workflow.compile()
+
+# Main function to run the system
+def run_deep_research_system(query: str) -> str:
+    # Initialize the workflow
+    app = create_workflow()
+
+    # Initial state
+    initial_state = {
+        "query": query,
+        "research_data": [],
+        "final_answer": ""
+    }
+
+    # Run the workflow
+    final_state = app.invoke(initial_state)
+
+    return final_state["final_answer"]
+
+# Example usage
+if __name__ == "__main__":
+    query = "What are the latest advancements in quantum computing?"
+    answer = run_deep_research_system(query)
+    print("\nFinal Answer:")
+    print(answer)