Upload 5 files

README.md

@@ -12,4 +12,28 @@ hf_oauth: true
 hf_oauth_expiration_minutes: 480
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+## Agent
+
+`agent.py` defines `GaiaAgent`, a LangGraph agent (explicit `StateGraph` with an
+`assistant` node and a `ToolNode`, routed by `tools_condition`) used to answer the
+GAIA benchmark questions for [Unit 4's hands-on assignment](https://huggingface.co/learn/agents-course/unit4/hands-on).
+Structure and prompt are adapted from the [fisherman611/gaia-agent](https://huggingface.co/spaces/fisherman611/gaia-agent)
+Space, trimmed to tools that need no extra paid API keys and no arbitrary code
+execution (that Space's Supabase RAG retriever, Tavily/Groq providers, and
+multi-language code interpreter were left out for that reason).
+
+It calls an LLM through the Hugging Face Inference API and has access to:
+web search (DuckDuckGo), Wikipedia search, arXiv search, arithmetic tools
+(add/subtract/multiply/divide/modulus/power/square_root), a tool to download a
+GAIA task's attached file, a generic URL downloader, and CSV/Excel analysis
+tools. The system prompt (`system_prompt.txt`) asks the model to end its reply
+with `FINAL ANSWER: ...`, which `agent.py` parses out before submission — per
+the assignment's rule that submitted answers must contain only the answer
+itself.
+
+To run it (locally or as a Space), set the `HF_TOKEN` secret/env var to a
+Hugging Face access token with Inference API permission. Optionally set
+`HF_AGENT_MODEL` to override the default model
+(`Qwen/Qwen2.5-Coder-32B-Instruct`).

agent.py

@@ -0,0 +1,337 @@
+"""LangGraph agent that answers GAIA benchmark questions (Hugging Face Agents Course, Unit 4).
+
+The agent is a LangGraph ReAct-style graph: an LLM served through the Hugging Face
+Inference API, bound to a toolset (web/Wikipedia/arXiv search, arithmetic, GAIA
+task-file download, and CSV/Excel analysis). Structure and prompt are adapted from
+https://huggingface.co/spaces/fisherman611/gaia-agent, trimmed to tools that need no
+extra paid API keys and no arbitrary code execution. See
+https://huggingface.co/learn/agents-course/unit4/hands-on for the assignment.
+"""
+
+import os
+import re
+import tempfile
+import uuid
+from pathlib import Path
+from typing import Optional
+from urllib.parse import urlparse
+
+import pandas as pd
+import requests
+from langchain_community.document_loaders import ArxivLoader, WikipediaLoader
+from langchain_community.tools import DuckDuckGoSearchRun
+from langchain_core.messages import HumanMessage, SystemMessage
+from langchain_core.tools import tool
+from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
+from langgraph.graph import START, MessagesState, StateGraph
+from langgraph.prebuilt import ToolNode, tools_condition
+
+DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+
+# Any HF Inference API model that supports tool calling works here. Override via the
+# HF_AGENT_MODEL env var without touching code.
+HF_MODEL_REPO_ID = os.getenv("HF_AGENT_MODEL", "Qwen/Qwen2.5-Coder-32B-Instruct")
+
+with open(Path(__file__).parent / "system_prompt.txt", "r", encoding="utf-8") as f:
+    SYSTEM_PROMPT = f.read()
+
+_FINAL_ANSWER_RE = re.compile(r"final answer\s*:\s*", re.IGNORECASE)
+
+
+### =============== SEARCH TOOLS =============== ###
+
+
+@tool
+def web_search(query: str) -> str:
+    """Search the web (DuckDuckGo) for a query and return a few results.
+
+    Args:
+        query: The search query.
+    """
+    return DuckDuckGoSearchRun().invoke(query)
+
+
+@tool
+def wiki_search(query: str) -> str:
+    """Search Wikipedia for a query and return up to 2 results.
+
+    Args:
+        query: The search query.
+    """
+    docs = WikipediaLoader(query=query, load_max_docs=2).load()
+    return "\n\n---\n\n".join(
+        f'<Document source="{d.metadata.get("source", "")}"/>\n{d.page_content}\n</Document>'
+        for d in docs
+    )
+
+
+@tool
+def arxiv_search(query: str) -> str:
+    """Search arXiv for a query and return up to 2 results (abstracts truncated).
+
+    Args:
+        query: The search query.
+    """
+    docs = ArxivLoader(query=query, load_max_docs=2).load()
+    return "\n\n---\n\n".join(
+        f'<Document source="{d.metadata.get("source", "")}"/>\n{d.page_content[:1000]}\n</Document>'
+        for d in docs
+    )
+
+
+### =============== MATH TOOLS =============== ###
+
+
+@tool
+def add(a: float, b: float) -> float:
+    """Add two numbers.
+
+    Args:
+        a: the first number
+        b: the second number
+    """
+    return a + b
+
+
+@tool
+def subtract(a: float, b: float) -> float:
+    """Subtract two numbers.
+
+    Args:
+        a: the first number
+        b: the second number
+    """
+    return a - b
+
+
+@tool
+def multiply(a: float, b: float) -> float:
+    """Multiply two numbers.
+
+    Args:
+        a: the first number
+        b: the second number
+    """
+    return a * b
+
+
+@tool
+def divide(a: float, b: float) -> float:
+    """Divide two numbers.
+
+    Args:
+        a: the numerator
+        b: the denominator
+    """
+    if b == 0:
+        raise ValueError("Cannot divide by zero.")
+    return a / b
+
+
+@tool
+def modulus(a: int, b: int) -> int:
+    """Get the remainder of a divided by b.
+
+    Args:
+        a: the first number
+        b: the second number
+    """
+    return a % b
+
+
+@tool
+def power(a: float, b: float) -> float:
+    """Raise a to the power of b.
+
+    Args:
+        a: the base
+        b: the exponent
+    """
+    return a**b
+
+
+@tool
+def square_root(a: float) -> float:
+    """Get the square root of a non-negative number.
+
+    Args:
+        a: the number to get the square root of
+    """
+    if a < 0:
+        raise ValueError("Cannot take the square root of a negative number.")
+    return a**0.5
+
+
+### =============== FILE TOOLS =============== ###
+
+
+@tool
+def download_task_file(task_id: str) -> str:
+    """Download the file attached to a GAIA task (if any) and return its text content.
+
+    Only useful when the question references an attached file. Pass the task's task_id.
+    Returns decoded text (truncated to 4000 characters) for text-like files, or a short
+    description (content type and size) for files that can't be decoded as text.
+    """
+    try:
+        response = requests.get(f"{DEFAULT_API_URL}/files/{task_id}", timeout=30)
+        response.raise_for_status()
+    except requests.exceptions.RequestException as e:
+        return f"Error downloading file for task {task_id}: {e}"
+
+    try:
+        return response.content.decode("utf-8")[:4000]
+    except UnicodeDecodeError:
+        content_type = response.headers.get("content-type", "unknown")
+        return (
+            f"File for task {task_id} is binary (content-type: {content_type}, "
+            f"{len(response.content)} bytes) and cannot be read as text."
+        )
+
+
+@tool
+def download_file_from_url(url: str, filename: Optional[str] = None) -> str:
+    """Download a file from a URL to a temporary path, for later analysis.
+
+    Args:
+        url: the URL of the file to download.
+        filename: optional filename to save as; a random one is used if omitted.
+    """
+    try:
+        if not filename:
+            filename = os.path.basename(urlparse(url).path) or f"downloaded_{uuid.uuid4().hex[:8]}"
+        filepath = os.path.join(tempfile.gettempdir(), filename)
+        response = requests.get(url, stream=True, timeout=30)
+        response.raise_for_status()
+        with open(filepath, "wb") as f:
+            for chunk in response.iter_content(chunk_size=8192):
+                f.write(chunk)
+        return f"File downloaded to {filepath}."
+    except Exception as e:
+        return f"Error downloading file: {e}"
+
+
+@tool
+def analyze_csv_file(file_path: str) -> str:
+    """Load a CSV file and return its shape, columns, and summary statistics.
+
+    Args:
+        file_path: path to the CSV file (e.g. from download_file_from_url).
+    """
+    try:
+        df = pd.read_csv(file_path)
+        return (
+            f"{len(df)} rows, {len(df.columns)} columns.\n"
+            f"Columns: {', '.join(df.columns)}\n\n"
+            f"Summary statistics:\n{df.describe(include='all')}"
+        )
+    except Exception as e:
+        return f"Error analyzing CSV file: {e}"
+
+
+@tool
+def analyze_excel_file(file_path: str) -> str:
+    """Load an Excel file and return its shape, columns, and summary statistics.
+
+    Args:
+        file_path: path to the .xlsx/.xls file (e.g. from download_file_from_url).
+    """
+    try:
+        df = pd.read_excel(file_path)
+        return (
+            f"{len(df)} rows, {len(df.columns)} columns.\n"
+            f"Columns: {', '.join(df.columns)}\n\n"
+            f"Summary statistics:\n{df.describe(include='all')}"
+        )
+    except Exception as e:
+        return f"Error analyzing Excel file: {e}"
+
+
+def _build_tools():
+    return [
+        web_search,
+        wiki_search,
+        arxiv_search,
+        add,
+        subtract,
+        multiply,
+        divide,
+        modulus,
+        power,
+        square_root,
+        download_task_file,
+        download_file_from_url,
+        analyze_csv_file,
+        analyze_excel_file,
+    ]
+
+
+def _build_llm():
+    endpoint = HuggingFaceEndpoint(
+        repo_id=HF_MODEL_REPO_ID,
+        huggingfacehub_api_token=os.getenv("HF_TOKEN"),
+        temperature=0,
+        max_new_tokens=1024,
+    )
+    return ChatHuggingFace(llm=endpoint)
+
+
+def build_graph():
+    """Build the compiled LangGraph agent graph."""
+    tools = _build_tools()
+    llm_with_tools = _build_llm().bind_tools(tools)
+
+    def assistant(state: MessagesState):
+        return {"messages": [llm_with_tools.invoke(state["messages"])]}
+
+    builder = StateGraph(MessagesState)
+    builder.add_node("assistant", assistant)
+    builder.add_node("tools", ToolNode(tools))
+    builder.add_edge(START, "assistant")
+    builder.add_conditional_edges("assistant", tools_condition)
+    builder.add_edge("tools", "assistant")
+
+    return builder.compile()
+
+
+def _extract_final_answer(text: str) -> str:
+    match = _FINAL_ANSWER_RE.search(text)
+    answer = text[match.end():] if match else text
+    answer = answer.strip()
+    if len(answer) >= 2 and answer[0] == answer[-1] and answer[0] in "\"'":
+        answer = answer[1:-1].strip()
+    return answer
+
+
+class GaiaAgent:
+    """A LangGraph ReAct agent (HF Inference API LLM + tools) for GAIA questions."""
+
+    def __init__(self):
+        self._graph = build_graph()
+        print("GaiaAgent initialized.")
+
+    def __call__(self, question: str, task_id: Optional[str] = None) -> str:
+        print(f"Agent received question (first 80 chars): {question[:80]}...")
+
+        user_content = question
+        if task_id:
+            user_content += (
+                f"\n\n(task_id: {task_id} - use download_task_file if a file is attached)"
+            )
+
+        try:
+            result = self._graph.invoke(
+                {
+                    "messages": [
+                        SystemMessage(content=SYSTEM_PROMPT),
+                        HumanMessage(content=user_content),
+                    ]
+                }
+            )
+            answer = _extract_final_answer(result["messages"][-1].content)
+        except Exception as e:
+            print(f"Agent error: {e}")
+            answer = f"AGENT ERROR: {e}"
+
+        print(f"Agent returning answer (first 80 chars): {answer[:80]}...")
+        return answer

app.py

@@ -4,24 +4,11 @@ import requests
 import inspect
 import pandas as pd
 
-# (Keep Constants as is)
-# --- Constants ---
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# --- Basic Agent Definition ---
-# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
-class BasicAgent:
-    def __init__(self):
-        print("BasicAgent initialized.")
-    def __call__(self, question: str) -> str:
-        print(f"Agent received question (first 50 chars): {question[:50]}...")
-        fixed_answer = "This is a default answer."
-        print(f"Agent returning fixed answer: {fixed_answer}")
-        return fixed_answer
+from agent import GaiaAgent, DEFAULT_API_URL
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
-    Fetches all questions, runs the BasicAgent on them, submits all answers,
+    Fetches all questions, runs the GaiaAgent on them, submits all answers,
     and displays the results.
     """
     # --- Determine HF Space Runtime URL and Repo URL ---
@@ -40,7 +27,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
 
     # 1. Instantiate Agent ( modify this part to create your agent)
     try:
-        agent = BasicAgent()
+        agent = GaiaAgent()
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
@@ -80,7 +67,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = agent(question_text)
+            submitted_answer = agent(question_text, task_id=task_id)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
         except Exception as e:

requirements.txt

@@ -1,2 +1,12 @@
 gradio
-requests
+itsdangerous
+requests
+pandas
+langgraph
+langchain-core
+langchain-huggingface
+langchain-community
+duckduckgo-search
+wikipedia
+arxiv
+openpyxl

system_prompt.txt

@@ -0,0 +1,5 @@
+You are a helpful assistant tasked with answering questions using a set of tools.
+Now, I will ask you a question. Report your thoughts, and finish your answer with the following template:
+FINAL ANSWER: [YOUR FINAL ANSWER].
+YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, apply the rules above for each element (number or string), and ensure there is exactly one space after each comma.
+Your answer should only start with "FINAL ANSWER: ", then follows with the answer.