Update app.py

app.py

@@ -5,8 +5,8 @@ import pandas as pd
 from typing import TypedDict, Annotated, Sequence
 import operator
 from langchain_core.messages import BaseMessage, HumanMessage
-from langchain_community.tools import DuckDuckGoSearchRun
-from langchain_huggingface import HuggingFaceEndpoint, ChatHuggingFace
+from langchain_community.tools.tavily_search import TavilySearchResults
+from langchain_openai import ChatOpenAI
 from langgraph.graph import StateGraph, END
 from langgraph.prebuilt import ToolNode, tools_condition
 
@@ -16,35 +16,31 @@ from langgraph.prebuilt import ToolNode, tools_condition
 class AgentState(TypedDict):
     messages: Annotated[Sequence[BaseMessage], operator.add]
 
-# This function builds our final, robust agent using LangGraph.
+# This function builds our final, robust agent using LangGraph and OpenAI
 def create_langgraph_agent():
-    print("Initializing LangGraph Agent...")
+    print("Initializing LangGraph Agent with OpenAI...")
     
-    # 1. Set up the LLM Endpoint connection
-    llm_endpoint = HuggingFaceEndpoint(
-        repo_id="deepseek-ai/deepseek-coder-6.7b-instruct",
-        task="conversational",
-        max_new_tokens=1024,
-        do_sample=False,
-    )
+    # 1. Set up the LLM (The "Brain") using OpenAI's GPT-3.5 Turbo
+    # The API key is automatically found from the OPENAI_API_KEY secret
+    llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
     
-    # 2. Wrap the endpoint in the ChatHuggingFace class to make it a chat model
-    # and bind the tools to it.
-    tools = [DuckDuckGoSearchRun()]
-    chat_model = ChatHuggingFace(llm=llm_endpoint)
-    chat_model_with_tools = chat_model.bind_tools(tools)
+    # We bind tools to the LLM. It's a modern model that supports this natively.
+    tools = [TavilySearchResults(max_results=3)]
+    llm_with_tools = llm.bind_tools(tools)
     print("LLM and tools initialized.")
 
-    # 3. Define the agent's logic (the "agent" node)
+    # 2. Define the Graph Nodes
+    # The 'agent' node calls the LLM
     def agent_node(state):
         print("Calling agent node...")
-        response = chat_model_with_tools.invoke(state["messages"])
+        response = llm_with_tools.invoke(state["messages"])
         return {"messages": [response]}
     
+    # The 'tool' node executes the tools
     tool_node = ToolNode(tools)
     print("Graph nodes defined.")
 
-    # 4. Define the Graph
+    # 3. Define the Graph
     graph = StateGraph(AgentState)
     graph.add_node("agent", agent_node)
     graph.add_node("tools", tool_node)
@@ -53,7 +49,7 @@ def create_langgraph_agent():
     graph.add_conditional_edges("agent", tools_condition)
     graph.add_edge("tools", "agent")
 
-    # 5. Compile the graph into a runnable app
+    # 4. Compile the graph into a runnable app
     app = graph.compile()
     print("LangGraph agent compiled and ready.")
     return app
@@ -78,6 +74,13 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         return "Please Login to Hugging Face with the button.", None
     username = f"{profile.username}"
     
+    # Tavily API key check
+    if not os.getenv("TAVILY_API_KEY"):
+         return "Tavily API key not found. Please set the TAVILY_API_KEY secret in your Space settings.", None
+    # OpenAI API key check
+    if not os.getenv("OPENAI_API_KEY"):
+         return "OpenAI API key not found. Please set the OPENAI_API_KEY secret in your Space settings.", None
+
     try:
         agent_executor = create_langgraph_agent()
     except Exception as e:
@@ -119,7 +122,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
 
 # Gradio Interface
 with gr.Blocks() as demo:
-    gr.Markdown("# Agent Evaluation Runner (DeepSeek + LangGraph)")
+    gr.Markdown("# Agent Evaluation Runner (OpenAI + LangGraph)")
     gr.LoginButton()
     run_button = gr.Button("Run Evaluation & Submit All Answers")
     status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)