Test_Magus / app.py
SergeyO7's picture
Update app.py
00bb7a7 verified
import os
import gradio as gr
import requests
import inspect
import pandas as pd
import aiohttp
import asyncio
import json
from agent import MagAgent
from token_bucket import Limiter, MemoryStorage
import aiofiles
from typing import Optional
import time
# --- Constants ---
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
# Rate limiting configuration
MAX_MODEL_CALLS_PER_MINUTE = 14 # Conservative buffer below 15 RPM
TOKEN_BUCKET_CAPACITY = MAX_MODEL_CALLS_PER_MINUTE
TOKEN_BUCKET_REFILL_RATE = MAX_MODEL_CALLS_PER_MINUTE / 60.0 # Tokens per second
# Initialize global token bucket with MemoryStorage
storage = MemoryStorage()
token_bucket = Limiter(rate=TOKEN_BUCKET_REFILL_RATE, capacity=TOKEN_BUCKET_CAPACITY, storage=storage)
async def check_n_load_attach(session: aiohttp.ClientSession, task_id: str, question: str, api_url: str = "https://agents-course-unit4-scoring.hf.space") -> Optional[str]:
file_url = f"{api_url}/files/{task_id}"
try:
async with session.get(file_url
, timeout=15) as response:
if response.status == 200:
content_type = str(response.headers.get("Content-Type", "")).lower()
content = await response.read() # Read the file content
# Determine extension based on content_type, content, or question
extension = await determine_extension(content_type, content, question)
if extension:
filename = f"{task_id}{extension}"
local_file_path = os.path.join("downloads", filename)
os.makedirs("downloads", exist_ok=True)
async with aiofiles.open(local_file_path, "wb") as file:
await file.write(content)
print(f"File downloaded successfully: {local_file_path}")
return local_file_path
else:
print(f"Unsupported content type: {content_type} for task {task_id}")
return None
else:
print(f"Failed to download file for task {task_id}: HTTP {response.status}")
return None
except aiohttp.ClientError as e:
print(f"Error downloading attachment for task {task_id}: {str(e)}")
return None
async def determine_extension(content_type: str, content: bytes, question
: str) -> Optional[str]:
# Check if the question mentions Excel
if "excel" in question.lower():
# Check for XLS signature
if content.startswith(b'\xD0\xCF\x11\xE0\xA1\xB1\x1A\xE1'):
return ".xls"
# Check for XLSX signature (ZIP archive)
elif content.startswith(b'\x50\x4B\x03\x04'):
return ".xlsx"
else:
return ".xlsx" # Default to XLSX if unsure
# Standard MIME type checks
if "image/png" in content_type:
return ".png"
elif "jpeg" in content_type or "jpg" in content_type:
return ".jpg"
elif "spreadsheetml.sheet" in content_type:
return ".xlsx"
elif "vnd.ms-excel" in content_type:
return ".xls"
elif "audio/mpeg" in content_type:
return ".mp3"
elif "application/pdf" in content_type:
return ".pdf"
elif "text/x-python" in content_type:
return ".py"
else:
return None
async def fetch_questions(session: aiohttp.ClientSession, questions_url: str) -> list:
"""Fetch questions asynchronously."""
try:
async with session.get(questions_url, timeout=15) as response:
response.raise_for_status()
questions_data = await response.json()
if not questions_data:
print("Fetched questions list is empty.")
return []
print(f"Fetched {len(questions_data)} questions.")
return questions_data
except aiohttp.ClientError as e:
print(f"Error fetching questions: {e}")
return None
except Exception as e:
print(f"An unexpected error occurred fetching questions: {e}")
return None
async def submit_answers(session: aiohttp.ClientSession, submit_url: str, submission_data: dict) -> dict:
"""Submit answers asynchronously."""
try:
async with session.post(submit_url, json=submission_data, timeout=60) as response:
response.raise_for_status()
return await response.json()
except aiohttp.ClientResponseError as e:
print(f"Submission Failed: Server responded with status {e.status}. Detail: {e.message}")
return None
except aiohttp.ClientError as e:
print(f"Submission Failed: Network error - {e}")
return None
except Exception as e:
print(f"An unexpected error occurred during submission: {e}")
return None
async def process_question(agent, question_text: str, task_id: str, file_path: Optional[str], results_log: list):
"""Process a single question with global rate limiting and retry logic."""
submitted_answer = None
max_retries = 4
retry_delay = 20 # Initial retry delay in seconds
atimeout = 300
for attempt in range(max_retries +1):
try:
while not token_bucket.consume(1):
print(f"Rate limit reached for task {task_id}. Waiting to retry...")
await asyncio.sleep(retry_delay)
print(f"Processing task {task_id} (attempt {attempt + 1})...")
submitted_answer = await asyncio.wait_for(
agent(question_text, file_path),
timeout=atimeout # Increased timeout for audio processing
)
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
print(f"Completed task {task_id} with answer: {submitted_answer[:50]}...")
################### Addl sleep
# await asyncio.sleep(retry_delay)
return {"task_id": task_id, "submitted_answer": submitted_answer}
except aiohttp.ClientResponseError as e:
if e.status == 429:
print(f"Rate limit hit for task {task_id}. Retrying after {retry_delay}s...")
retry_delay *= 2 # Exponential backoff
await asyncio.sleep(retry_delay)
continue
else:
submitted_answer = f"AGENT ERROR: {e}"
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
print(f"Failed task {task_id}: {submitted_answer}")
return None
except asyncio.TimeoutError:
submitted_answer = f"AGENT ERROR: Timeout after {atimeout} seconds"
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
print(f"Failed task {task_id}: {submitted_answer}")
return None
except Exception as e:
submitted_answer = f"AGENT ERROR: {e}"
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
print(f"Failed task {task_id}: {submitted_answer}")
return None
async def run_and_submit_all(profile: gr.OAuthProfile | None):
"""
Fetches all questions asynchronously, runs the MagAgent on them, submits all answers,
and displays the results.
"""
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"
try:
agent = MagAgent(rate_limiter=token_bucket)
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(agent_code)
async with aiohttp.ClientSession() as session:
questions_data = await fetch_questions(session, questions_url)
if questions_data is None:
return "Error fetching questions.", None
if not questions_data:
return "Fetched questions list is empty or invalid format.", None
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
#select question we want to focus
if "sosa11" in question_text.lower() or "bird" in question_text.lower() or "yankee" in question_text.lower():
file_path = await check_n_load_attach(session, task_id, question_text.lower())
result = await process_question(agent, question_text, task_id, file_path, results_log)
if result:
answers_payload.append(result)
else:
print("Skipping unrelated question.")
if not answers_payload:
print("Agent did not produce any answers to submit.")
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
print(status_update)
result_data = await submit_answers(session, submit_url, submission_data)
if result_data is None:
status_message = "Submission Failed."
print(status_message)
results_df = pd.DataFrame(results_log)
return status_message, results_df
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.")
results_df = pd.DataFrame(results_log)
return final_status, results_df
# --- Build Gradio Interface using Blocks ---
with gr.Blocks() as demo:
gr.Markdown("# Magus Agent Evaluation Runner")
gr.Markdown(
"""
**Instructions:**
1. Log in to your Hugging Face account using the button below.
2. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, and submit answers.
---
**Notes:**
The agent uses asynchronous operations for efficiency. Answers are processed and submitted asynchronously.
"""
)
gr.LoginButton()
run_button = gr.Button("Run Evaluation & Submit All Answers")
status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, 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_startup = os.getenv("SPACE_HOST")
space_id_startup = os.getenv("SPACE_ID")
if space_host_startup:
print(f"✅ SPACE_HOST found: {space_host_startup}")
print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
else:
print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
if space_id_startup:
print(f"✅ SPACE_ID found: {space_id_startup}")
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
else:
print("ℹ️ SPACE_HOST environment variable not found (running locally?). Repo URL cannot be determined.")
print("-"*(60 + len(" App Starting ")) + "\n")
print("Launching Gradio Interface for Mag Agent Evaluation...")
demo.launch(debug=True, share=False)