separated out agent file

agent.py

@@ -0,0 +1,169 @@
+from textwrap import dedent
+from typing import TypedDict, List, Dict, Any, Optional, Annotated
+import os
+
+# from langchain_openai import ChatOpenAI
+# from langchain_huggingface.llms import HuggingFaceEndpoint
+from langgraph.graph import StateGraph, START, END
+from langgraph.prebuilt import ToolNode, tools_condition
+from langchain_core.messages import AnyMessage, SystemMessage, HumanMessage
+from langgraph.graph.message import add_messages
+from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
+
+from langfuse.langchain import CallbackHandler
+from langfuse import get_client
+
+from tools import fetch_website, get_wiki_full, youtube_transcript, python_repl_tool, duckduckgo_search_results
+
+os.environ["LANGFUSE_PUBLIC_KEY"] = os.getenv("LANGFUSE_PUBLIC_KEY", "pk-lf-***")  # Public key is safe to expose in client-side code
+os.environ["LANGFUSE_SECRET_KEY"] = os.getenv("LANGFUSE_SECRET_KEY", "sk-lf-***") 
+os.environ["LANGFUSE_BASE_URL"] = os.getenv("LANGFUSE_BASE_URL", "https://us.cloud.langfuse.com") # 🇺🇸 US region
+
+langfuse = get_client()
+# Verify connection
+if langfuse.auth_check():
+    print("Langfuse client is authenticated and ready!")
+else:
+    print("Authentication failed. Please check your credentials and host.")
+langfuse_handler = CallbackHandler()
+
+# Initialize the Hugging Face model
+hf_model_name = "openai/gpt-oss-120b" # "Qwen/Qwen2.5-72B-Instruct"
+hf_model_provider = "nscale" # "hf-inference"
+
+llm = HuggingFaceEndpoint(
+    repo_id=hf_model_name, 
+    provider=hf_model_provider,
+    max_new_tokens=8192,
+    do_sample=False,
+    # temperature=0.,
+)
+
+chat_model = ChatHuggingFace(llm=llm)
+
+# Equip llm with tools
+tools_list = [
+    fetch_website,
+    get_wiki_full,
+    youtube_transcript,
+    python_repl_tool,
+    duckduckgo_search_results
+]
+
+llm_with_tools = chat_model.bind_tools(
+    tools_list
+)
+
+# Define Agent Workflow
+
+class AgentState(TypedDict):
+    messages: Annotated[list[AnyMessage], add_messages]
+
+
+def assistant(state: AgentState):
+    # System message
+    textual_description_of_tool = dedent(
+        """
+        duckduckgo_search_results(query: str) -> list[dict]:
+            Perform a web search using DuckDuckGo and return the results.
+            Args:
+                query: The search query string.
+            Returns:
+                A list of search results, where each result is a dictionary that includes the snippet, title, and link.
+
+        fetch_website(url: str) -> str:
+            Fetch the content of a website.
+            Args:
+                url: The URL of the website to fetch.
+            Returns:
+                The title and content of the website.
+
+        get_wiki_full(query: str) -> str:
+            Scrape the content of a Wikipedia page based on the user query.
+            Args:
+                query: The user query to search for on Wikipedia.
+            Returns:
+                A single string containing the content of the Wikipedia page.
+
+        youtube_transcript(url: str) -> str:
+            Fetch the transcript of a youtube video.
+            Args:
+                url: input youtube url.
+            Returns:
+                A single string containing the transcript of the youtube videos.
+
+        python_repl_tool(code: str) -> str:
+            Execute Python code and return the output.
+            Args:
+                code: A string of Python code to execute.
+            Returns:
+                The output of the executed code or any error messages.
+        """
+    )
+
+    sys_msg = SystemMessage(
+        content=dedent(
+            f"""
+            You are a helpful assistant at answering user questions. \
+            Your final answer will be between <answer> and </answer> tags. \
+            You can access provided tools:\n{textual_description_of_tool}\n"""
+        )
+    )
+
+    return {
+        "messages": [llm_with_tools.invoke([sys_msg] + state["messages"])],
+    }
+
+# Build the StateGraph for the agent
+# The graph
+builder = StateGraph(AgentState)
+
+# Define nodes: these do the work
+builder.add_node("assistant", assistant)
+builder.add_node("tools", ToolNode(tools_list))
+
+# Define edges: these determine how the control flow moves
+builder.add_edge(START, "assistant")
+builder.add_conditional_edges(
+    "assistant",
+    # If the latest message requires a tool, route to tools
+    # Otherwise, provide a direct response
+    tools_condition,
+)
+builder.add_edge("tools", "assistant")
+agent_graph = builder.compile()
+
+def extract_answer(text):
+    match = re.search(r'<answer>(.*?)</answer>', text, re.DOTALL)
+    if match:
+        return match.group(1).strip()
+    return 'None'
+
+class BasicAgent:
+    def __init__(self):
+        print("BasicAgent initialized.")
+    async def __call__(self, question: str) -> str:
+        print(f"Agent received question (first 100 chars): {question[:100]}...")
+        # fixed_answer = "This is a default answer."
+        # print(f"Agent returning fixed answer: {fixed_answer}")
+        # Create agent with all the tools
+        
+        # Example query agent might receive
+        # fixed_answer = await agent.run(question)
+        messages = [
+            HumanMessage(
+                # content="Who is Barack Obama?"
+                # content="Divide 6790 by 5"
+                content=question # + '/nothink'
+            )
+        ]
+        response = await agent_graph.ainvoke(
+            {"messages": messages}, 
+            config={
+                "recursion_limit": 10,
+                "callbacks": [langfuse_handler],
+            }
+        )
+        response_text = response['messages'][-1].content
+        # return response_text.split('</think>')[-1]
+        return extract_answer(response_text)

app.py

@@ -6,7 +6,7 @@ import requests
 import inspect
 import pandas as pd
 import re
-from textwrap import dedent
+
 
 # (Keep Constants as is)
 # --- Constants ---
@@ -15,146 +15,9 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 
-from typing import TypedDict, List, Dict, Any, Optional, Annotated
-from langgraph.graph import StateGraph, START, END
-# from langchain_openai import ChatOpenAI
-# from langchain_huggingface.llms import HuggingFaceEndpoint
-from langchain_core.messages import AnyMessage, SystemMessage, HumanMessage
-from langgraph.graph.message import add_messages
-from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
-from langgraph.prebuilt import ToolNode, tools_condition
-
-from tools import fetch_website, get_wiki_full, youtube_transcript, python_repl_tool, duckduckgo_search_results
-
-# Initialize the Hugging Face model
-hf_model_name = "openai/gpt-oss-120b" # "Qwen/Qwen2.5-72B-Instruct"
-hf_model_provider = "nscale" # "hf-inference"
-
-llm = HuggingFaceEndpoint(
-    repo_id=hf_model_name, 
-    provider=hf_model_provider,
-    max_new_tokens=8192,
-    do_sample=False,
-    # temperature=0.,
-)
-
-chat_model = ChatHuggingFace(llm=llm)
-
-# Equip llm with tools
-tools_list = [
-    fetch_website,
-    get_wiki_full,
-    youtube_transcript,
-    python_repl_tool,
-    duckduckgo_search_results
-]
-
-llm_with_tools = chat_model.bind_tools(
-    tools_list
-)
-
-# Define Agent Workflow
-
-class AgentState(TypedDict):
-    messages: Annotated[list[AnyMessage], add_messages]
-
-
-def assistant(state: AgentState):
-    # System message
-    textual_description_of_tool = dedent(
-        """
-        duckduckgo_search_results(query: str) -> list[dict]:
-            Perform a web search using DuckDuckGo and return the results.
-            Args:
-                query: The search query string.
-            Returns:
-                A list of search results, where each result is a dictionary that includes the snippet, title, and link.
-
-        fetch_website(url: str) -> str:
-            Fetch the content of a website.
-            Args:
-                url: The URL of the website to fetch.
-            Returns:
-                The title and content of the website.
-
-        get_wiki_full(query: str) -> str:
-            Scrape the content of a Wikipedia page based on the user query.
-            Args:
-                query: The user query to search for on Wikipedia.
-            Returns:
-                A single string containing the content of the Wikipedia page.
-
-        youtube_transcript(url: str) -> str:
-            Fetch the transcript of a youtube video.
-            Args:
-                url: input youtube url.
-            Returns:
-                A single string containing the transcript of the youtube videos.
-
-        python_repl_tool(code: str) -> str:
-            Execute Python code and return the output.
-            Args:
-                code: A string of Python code to execute.
-            Returns:
-                The output of the executed code or any error messages.
-        """
-    )
-
-    sys_msg = SystemMessage(
-        content=f"You are a helpful assistant at answering user questions. Your final answer will be between <answer> and </answer> tags. You can access provided tools:\n{textual_description_of_tool}\n"
-    )
-
-    return {
-        "messages": [llm_with_tools.invoke([sys_msg] + state["messages"])],
-    }
-
-# Build the StateGraph for the agent
-# The graph
-builder = StateGraph(AgentState)
-
-# Define nodes: these do the work
-builder.add_node("assistant", assistant)
-builder.add_node("tools", ToolNode(tools_list))
-
-# Define edges: these determine how the control flow moves
-builder.add_edge(START, "assistant")
-builder.add_conditional_edges(
-    "assistant",
-    # If the latest message requires a tool, route to tools
-    # Otherwise, provide a direct response
-    tools_condition,
-)
-builder.add_edge("tools", "assistant")
-agent_graph = builder.compile()
-
-def extract_answer(text):
-    match = re.search(r'<answer>(.*?)</answer>', text, re.DOTALL)
-    if match:
-        return match.group(1).strip()
-    return 'None'
+# offloaded to agent.py for better modularity and readability
+from agent import BasicAgent
 
-class BasicAgent:
-    def __init__(self):
-        print("BasicAgent initialized.")
-    async def __call__(self, question: str) -> str:
-        print(f"Agent received question (first 100 chars): {question[:100]}...")
-        # fixed_answer = "This is a default answer."
-        # print(f"Agent returning fixed answer: {fixed_answer}")
-        # Create agent with all the tools
-        
-        # Example query agent might receive
-        # fixed_answer = await agent.run(question)
-        messages = [
-            HumanMessage(
-                # content="Who is Barack Obama?"
-                # content="Divide 6790 by 5"
-                content=question # + '/nothink'
-            )
-        ]
-        response = await agent_graph.ainvoke({"messages": messages}, config={"recursion_limit": 10})
-        response_text = response['messages'][-1].content
-        # return response_text.split('</think>')[-1]
-        return extract_answer(response_text)
 
 async def run_and_submit_all( profile: gr.OAuthProfile | None):
     """