Download app.py from agents-course/Final_Assignment_Template: direct link, hf CLI and curl.
- Browser
- Download file 14.1 kB
-
https://huggingface.co/spaces/agents-course/Final_Assignment_Template/resolve/refs%2Fpr%2F348/app.py
- Command line
-
hf download hf://spaces/agents-course/Final_Assignment_Template@refs/pr/348/app.py
-
curl -L -o app.py https://huggingface.co/spaces/agents-course/Final_Assignment_Template/resolve/refs%2Fpr%2F348/app.py
14.1 kB
| import os | |
| import gradio as gr | |
| import requests | |
| import pandas as pd | |
| import re | |
| import json | |
| import math | |
| import unicodedata | |
| from datetime import datetime | |
| # --- LangGraph + LangChain imports --- | |
| from langgraph.prebuilt import create_react_agent | |
| from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint | |
| from langchain_core.tools import tool | |
| from langchain_community.tools import DuckDuckGoSearchRun | |
| from langchain_community.utilities import WikipediaAPIWrapper | |
| from langchain_core.messages import SystemMessage | |
| # --- Constants --- | |
| DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| # TOOLS | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| def web_search(query: str) -> str: | |
| """Search the web using DuckDuckGo. Use for current events, facts, and general knowledge.""" | |
| try: | |
| search = DuckDuckGoSearchRun() | |
| return search.run(query) | |
| except Exception as e: | |
| return f"Search error: {e}" | |
| def wikipedia_search(query: str) -> str: | |
| """Search Wikipedia for encyclopedic knowledge, historical facts, biographies, science.""" | |
| try: | |
| wiki = WikipediaAPIWrapper(top_k_results=2, doc_content_chars_max=3000) | |
| return wiki.run(query) | |
| except Exception as e: | |
| return f"Wikipedia error: {e}" | |
| def python_repl(code: str) -> str: | |
| """ | |
| Execute Python code for math calculations, data processing, logic. | |
| Always print() the final result. | |
| Example: print(2 + 2) | |
| """ | |
| import io, sys, math, json, re, unicodedata, datetime | |
| old_stdout = sys.stdout | |
| sys.stdout = io.StringIO() | |
| try: | |
| exec(code, { | |
| "math": math, "json": json, "re": re, | |
| "unicodedata": unicodedata, "datetime": datetime, | |
| "__builtins__": __builtins__ | |
| }) | |
| output = sys.stdout.getvalue() | |
| return output.strip() if output.strip() else "Code executed (no output). Use print() to see results." | |
| except Exception as e: | |
| return f"Code error: {e}" | |
| finally: | |
| sys.stdout = old_stdout | |
| def read_file_from_url(url: str) -> str: | |
| """ | |
| Download and read a file from a URL (txt, csv, json, py, etc.). | |
| Returns the file content as text. | |
| """ | |
| try: | |
| response = requests.get(url, timeout=15) | |
| response.raise_for_status() | |
| content_type = response.headers.get("Content-Type", "") | |
| if "text" in content_type or "json" in content_type: | |
| return response.text[:5000] | |
| else: | |
| return f"Binary file ({content_type}), cannot read as text." | |
| except Exception as e: | |
| return f"Error reading file: {e}" | |
| def get_task_file(task_id: str) -> str: | |
| """ | |
| Fetch the file associated with a GAIA task by its task_id. | |
| Returns file content or description. | |
| """ | |
| try: | |
| api_url = "https://agents-course-unit4-scoring.hf.space" | |
| url = f"{api_url}/files/{task_id}" | |
| response = requests.get(url, timeout=15) | |
| if response.status_code == 200: | |
| content_type = response.headers.get("Content-Type", "") | |
| if "text" in content_type or "json" in content_type: | |
| return response.text[:5000] | |
| elif "image" in content_type: | |
| return f"[Image file attached to task {task_id} - content-type: {content_type}]" | |
| elif "audio" in content_type: | |
| return f"[Audio file attached to task {task_id} - content-type: {content_type}]" | |
| else: | |
| return f"[File attached: {content_type}]" | |
| else: | |
| return f"No file found for task {task_id}" | |
| except Exception as e: | |
| return f"Error fetching task file: {e}" | |
| def calculator(expression: str) -> str: | |
| """ | |
| Evaluate a simple math expression safely. | |
| Examples: '2 + 2', '100 * 1.07 ** 5', 'math.sqrt(144)' | |
| """ | |
| try: | |
| result = eval(expression, {"math": math, "__builtins__": {}}) | |
| return str(result) | |
| except Exception as e: | |
| return f"Calculation error: {e}. Try python_repl for complex code." | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| # SYSTEM PROMPT | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| SYSTEM_PROMPT = """You are a precise, expert AI assistant solving GAIA benchmark questions. | |
| GAIA questions require careful reasoning and often multiple steps. Follow these rules: | |
| ## Answer Format (CRITICAL) | |
| - Your FINAL answer must be the **bare minimum**: a number, a word, a name, a date, a short phrase. | |
| - NO explanations, NO punctuation at the end, NO "The answer is...", NO sentences. | |
| - Examples of correct final answers: `42`, `Marie Curie`, `Paris`, `1969`, `blue`, `$14.50` | |
| - For lists, separate items with commas: `item1, item2, item3` | |
| ## Strategy | |
| 1. **Read carefully** β identify exactly what is being asked. | |
| 2. **Use tools** β search the web, Wikipedia, or run code to verify facts. | |
| 3. **Verify numbers** β always double-check calculations with the calculator or python_repl. | |
| 4. **Check for files** β if the question mentions an attachment or file, use get_task_file. | |
| 5. **Be specific** β GAIA answers are exact; approximate answers are wrong. | |
| ## Tool Usage | |
| - Use `web_search` for recent events, facts, and general knowledge. | |
| - Use `wikipedia_search` for biographies, history, science. | |
| - Use `python_repl` for calculations, data manipulation, logic puzzles. | |
| - Use `calculator` for quick arithmetic. | |
| - Use `get_task_file` when a question refers to an attached file or document. | |
| ## Final Answer | |
| Always end your response with: | |
| FINAL ANSWER: <your answer here> | |
| """ | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| # AGENT | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| class BasicAgent: | |
| def __init__(self): | |
| print("Initializing LangGraph ReAct Agent with Llama 3.3 70B...") | |
| hf_token = os.getenv("HF_TOKEN") | |
| llm_endpoint = HuggingFaceEndpoint( | |
| repo_id="meta-llama/Llama-3.3-70B-Instruct", | |
| huggingfacehub_api_token=hf_token, | |
| task="text-generation", | |
| max_new_tokens=1024, | |
| temperature=0.1, | |
| do_sample=False, | |
| ) | |
| llm = ChatHuggingFace(llm=llm_endpoint) | |
| tools = [ | |
| web_search, | |
| wikipedia_search, | |
| python_repl, | |
| calculator, | |
| read_file_from_url, | |
| get_task_file, | |
| ] | |
| self.agent = create_react_agent( | |
| model=llm, | |
| tools=tools, | |
| state_modifier=SYSTEM_PROMPT, | |
| ) | |
| print("Agent ready.") | |
| def __call__(self, question: str) -> str: | |
| print(f"\n[AGENT] Question: {question[:100]}...") | |
| try: | |
| result = self.agent.invoke({ | |
| "messages": [("user", question)] | |
| }) | |
| # Extract last AI message | |
| last_message = result["messages"][-1].content | |
| print(f"[AGENT] Raw output: {last_message[:200]}...") | |
| # Extract FINAL ANSWER if present | |
| answer = self._extract_final_answer(last_message) | |
| print(f"[AGENT] Final answer: {answer}") | |
| return answer | |
| except Exception as e: | |
| print(f"[AGENT] Error: {e}") | |
| return f"Error: {e}" | |
| def _extract_final_answer(self, text: str) -> str: | |
| """Extract the FINAL ANSWER from agent output.""" | |
| # Try to find "FINAL ANSWER: ..." pattern | |
| patterns = [ | |
| r"FINAL ANSWER:\s*(.+?)(?:\n|$)", | |
| r"Final Answer:\s*(.+?)(?:\n|$)", | |
| r"final answer:\s*(.+?)(?:\n|$)", | |
| ] | |
| for pattern in patterns: | |
| match = re.search(pattern, text, re.IGNORECASE) | |
| if match: | |
| return match.group(1).strip() | |
| # Fallback: return last non-empty line | |
| lines = [l.strip() for l in text.strip().split("\n") if l.strip()] | |
| return lines[-1] if lines else text.strip() | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| # GRADIO RUNNER | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| def run_and_submit_all(profile: gr.OAuthProfile | None): | |
| space_id = os.getenv("SPACE_ID") | |
| if profile: | |
| username = f"{profile.username}" | |
| print(f"User logged in: {username}") | |
| else: | |
| print("User not logged in.") | |
| return "Please Login to Hugging Face with the button.", None | |
| api_url = DEFAULT_API_URL | |
| questions_url = f"{api_url}/questions" | |
| submit_url = f"{api_url}/submit" | |
| # 1. Init Agent | |
| try: | |
| agent = BasicAgent() | |
| except Exception as e: | |
| print(f"Error instantiating agent: {e}") | |
| return f"Error initializing agent: {e}", None | |
| agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" | |
| print(f"Agent code: {agent_code}") | |
| # 2. Fetch Questions | |
| print(f"Fetching questions from: {questions_url}") | |
| try: | |
| response = requests.get(questions_url, timeout=15) | |
| response.raise_for_status() | |
| questions_data = response.json() | |
| if not questions_data: | |
| return "Fetched questions list is empty or invalid format.", None | |
| print(f"Fetched {len(questions_data)} questions.") | |
| except Exception as e: | |
| return f"Error fetching questions: {e}", None | |
| # 3. Run Agent | |
| results_log = [] | |
| answers_payload = [] | |
| print(f"Running agent on {len(questions_data)} questions...") | |
| for item in questions_data: | |
| task_id = item.get("task_id") | |
| question_text = item.get("question") | |
| if not task_id or question_text is None: | |
| print(f"Skipping item with missing task_id or question: {item}") | |
| continue | |
| try: | |
| submitted_answer = agent(question_text) | |
| answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer}) | |
| results_log.append({ | |
| "Task ID": task_id, | |
| "Question": question_text[:100], | |
| "Submitted Answer": submitted_answer | |
| }) | |
| except Exception as e: | |
| print(f"Error on task {task_id}: {e}") | |
| results_log.append({ | |
| "Task ID": task_id, | |
| "Question": question_text[:100], | |
| "Submitted Answer": f"AGENT ERROR: {e}" | |
| }) | |
| if not answers_payload: | |
| return "Agent did not produce any answers to submit.", pd.DataFrame(results_log) | |
| # 4. Submit | |
| submission_data = { | |
| "username": username.strip(), | |
| "agent_code": agent_code, | |
| "answers": answers_payload | |
| } | |
| print(f"Submitting {len(answers_payload)} answers...") | |
| try: | |
| response = requests.post(submit_url, json=submission_data, timeout=120) | |
| response.raise_for_status() | |
| result_data = response.json() | |
| final_status = ( | |
| f"Submission Successful!\n" | |
| f"User: {result_data.get('username')}\n" | |
| f"Overall Score: {result_data.get('score', 'N/A')}% " | |
| f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n" | |
| f"Message: {result_data.get('message', 'No message received.')}" | |
| ) | |
| print("Submission successful.") | |
| return final_status, pd.DataFrame(results_log) | |
| except requests.exceptions.HTTPError as e: | |
| error_detail = f"Server responded with status {e.response.status_code}." | |
| try: | |
| error_json = e.response.json() | |
| error_detail += f" Detail: {error_json.get('detail', e.response.text)}" | |
| except Exception: | |
| error_detail += f" Response: {e.response.text[:500]}" | |
| return f"Submission Failed: {error_detail}", pd.DataFrame(results_log) | |
| except Exception as e: | |
| return f"Submission Failed: {e}", pd.DataFrame(results_log) | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| # GRADIO UI | |
| # βββββββββββββββββββββββββββββββββββββββββββββ | |
| with gr.Blocks() as demo: | |
| gr.Markdown("# π€ GAIA Agent β LangGraph + Llama 3.3 70B") | |
| gr.Markdown(""" | |
| **Stack:** LangGraph ReAct Β· Llama 3.3 70B (HF Inference) Β· DuckDuckGo Β· Wikipedia Β· Python REPL | |
| **Instructions:** | |
| 1. Log in with your HuggingFace account below. | |
| 2. Make sure `HF_TOKEN` is set as a Space secret (with access to Llama 3.3 70B). | |
| 3. Click **Run Evaluation & Submit All Answers**. | |
| > β οΈ The run can take several minutes β the agent reasons through each question step by step. | |
| """) | |
| gr.LoginButton() | |
| run_button = gr.Button("βΆοΈ Run Evaluation & Submit All Answers", variant="primary") | |
| status_output = gr.Textbox(label="Run Status / Submission Result", lines=6, interactive=False) | |
| results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True) | |
| run_button.click( | |
| fn=run_and_submit_all, | |
| outputs=[status_output, results_table] | |
| ) | |
| if __name__ == "__main__": | |
| print("\n" + "-" * 30 + " App Starting " + "-" * 30) | |
| space_host = os.getenv("SPACE_HOST") | |
| space_id = os.getenv("SPACE_ID") | |
| if space_host: | |
| print(f"β SPACE_HOST: {space_host}") | |
| if space_id: | |
| print(f"β SPACE_ID: {space_id}") | |
| print(f" Repo: https://huggingface.co/spaces/{space_id}/tree/main") | |
| print("-" * 60 + "\n") | |
| demo.launch(debug=True, share=False) |