Sonu313131's picture
Update app.py
99d61ed verified
raw
history blame
10.6 kB
import os
import gradio as gr
import requests
import inspect
import pandas as pd
import asyncio
from smolagents import ToolCallingAgent, InferenceClientModel, OpenAIServerModel
from smolagents import DuckDuckGoSearchTool, Tool, CodeAgent
from huggingface_hub import login
#h
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
#login(token=os.environ["HUGGINGFACEHUB_API_TOKEN"])
openai_key = os.environ.get("OPENAI_API_KEY")
search_tool = DuckDuckGoSearchTool()
##Tool 1
import re
from youtube_transcript_api import YouTubeTranscriptApi
from smolagents import Tool
from smolagents import Tool
from youtube_transcript_api import YouTubeTranscriptApi
from urllib.parse import urlparse, parse_qs
class YouTubeTranscriptTool(Tool):
name = "youtube_transcript"
description = "Fetches the full transcript of a YouTube video from its URL."
inputs = {
"url": {
"type": "string",
"description": "The full YouTube video URL"
}
}
output_type = "string"
def forward(self, url: str) -> str:
try:
# Extract video ID from URL
parsed = urlparse(url)
video_id = parse_qs(parsed.query).get("v", [None])[0]
if not video_id:
return "Error: Invalid YouTube URL or missing video ID."
# Fetch the transcript
transcript_list = YouTubeTranscriptApi.get_transcript(video_id)
transcript_text = " ".join(entry["text"] for entry in transcript_list)
return transcript_text[:5000] # Optional: truncate to 5000 chars
except Exception as e:
return f"Error retrieving transcript: {str(e)}"
##Tool 2
import wikipedia
from smolagents import Tool
from smolagents.models import InferenceClientModel
class WikipediaQATool(Tool):
name = "wikipedia_qa"
description = (
"Searches Wikipedia for a topic, reads its content, and answers the input question "
"based on the content of the Wikipedia page."
)
inputs = {
"question": {
"type": "string",
"description": "The question that should be answered using Wikipedia."
},
"topic": {
"type": "string",
"description": "The topic to search for on Wikipedia."
}
}
output_type = "string"
def __init__(self, model=None):
super().__init__()
self.model = model or InferenceClientModel(
model="mistralai/Magistral-Small-2506", provider="featherless-ai"
)
def forward(self, question: str, topic: str) -> str:
try:
page = wikipedia.page(topic)
content = page.content[:2000] # Limit for context
# Build QA prompt
prompt = (
f"You are a Wikipedia expert. Based only on the following content from the Wikipedia page on '{topic}', "
f"answer the question briefly and factually.\n\n"
f"=== Wikipedia Content ===\n{content}\n\n"
f"=== Question ===\n{question}\n\n"
f"Answer in a single line. Avoid any extra explanation.\n"
f"FINAL ANSWER:"
)
response = self.model(prompt)
return response.strip()
except wikipedia.DisambiguationError as e:
return f"Disambiguation error: multiple results found: {', '.join(e.options[:5])}"
except wikipedia.PageError:
return "Wikipedia page not found."
except Exception as e:
return f"Error while retrieving Wikipedia content: {str(e)}"
wiki_tool = WikipediaQATool()
#excel_tool = ExcelAnalysisTool()
yt_tool = YouTubeTranscriptTool()
async def run_and_submit_all(profile: gr.OAuthProfile | None):
log_output = ""
try:
agent = ToolCallingAgent(
tools=[search_tool, yt_tool],
model=OpenAIServerModel(model_id="gpt-4o-mini",
api_key=os.environ["OPENAI_API_KEY"],
temperature=0.0),
max_steps=4,
verbosity_level=2
)
except Exception as e:
yield f"Error initializing agent: {e}", None, log_output
return
space_id = os.getenv("SPACE_ID")
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
questions_url = f"{DEFAULT_API_URL}/questions"
try:
response = requests.get(questions_url, timeout=15)
response.raise_for_status()
questions_data = response.json()
selected_indices = [0, 2, 4, 6, 10, 12, 14, 15, 16, 17, 19] # Replace with the indices you want
questions_data = [questions_data[i] for i in selected_indices if i < len(questions_data)]
if not questions_data:
yield "Fetched questions list is empty or invalid format.", None, log_output
return
except Exception as e:
yield f"Error fetching questions: {e}", None, log_output
return
results_log = []
answers_payload = []
loop = asyncio.get_event_loop()
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:
continue
log_output += f"πŸ” Solving Task ID: {task_id}...\n"
yield None, None, log_output
try:
system_prompt = (
"""You must only reply with a single line:
FINAL ANSWER: [your answer]
Never include reasoning, markdown, Task Outcome, Explanation, or examples.
NEVER use numbered points or extra formatting.
If your answer is a string, write it in lowercase, no articles, no quotes.
If your answer is a number, use digits only. If the answer is "no one" or "none", write exactly that.
DO NOT provide any explanation or context. Just the line: FINAL ANSWER: ...
If the answer is "st. petersberg" answer as "saint petersburg" (without abbreviations)
If the answer is "three" answer as "3".
"""
)
full_prompt = system_prompt + f"Question: {question_text.strip()}"
agent_result = await loop.run_in_executor(None, agent, full_prompt)
# Extract final answer cleanly
if isinstance(agent_result, dict) and "final_answer" in agent_result:
final_answer = str(agent_result["final_answer"]).strip()
elif isinstance(agent_result, str):
response_text = agent_result.strip()
# Remove known boilerplate
if "Here is the final answer from your managed agent" in response_text:
response_text = response_text.split(":", 1)[-1].strip()
if "FINAL ANSWER:" in response_text:
_, final_answer = response_text.rsplit("FINAL ANSWER:", 1)
final_answer = final_answer.strip()
else:
final_answer = response_text
else:
final_answer = str(agent_result).strip()
answers_payload.append({
"task_id": task_id,
"submitted_answer": final_answer
})
results_log.append({
"Task ID": task_id,
"Question": question_text,
"Submitted Answer": final_answer
})
log_output += f"βœ… Done: {task_id} β€” Answer: {final_answer[:60]}\n"
yield None, None, log_output
except Exception as e:
print(f"Error running agent on task {task_id}: {e}")
results_log.append({
"Task ID": task_id,
"Question": question_text,
"Submitted Answer": f"AGENT ERROR: {e}"
})
log_output += f"⛔️ Error: {task_id} β€” {e}\n"
yield None, None, log_output
if not answers_payload:
yield "Agent did not produce any answers to submit.", pd.DataFrame(results_log), log_output
return
username = profile.username if profile else "unknown"
submit_url = f"{DEFAULT_API_URL}/submit"
submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
try:
response = requests.post(submit_url, json=submission_data, timeout=60)
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.')}"
)
results_df = pd.DataFrame(results_log)
yield final_status, results_df, log_output
except Exception as e:
status_message = f"Submission Failed: {e}"
results_df = pd.DataFrame(results_log)
yield status_message, results_df, log_output
with gr.Blocks() as demo:
gr.Markdown("# Basic Agent Evaluation Runner")
gr.Markdown("""
**Instructions:**
1. Clone this space and define your agent logic.
2. Log in to your Hugging Face account.
3. Click 'Run Evaluation & Submit All Answers'.
---
**Note:**
The run may take time. Async is now used to improve responsiveness.
""")
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)
progress_log = gr.Textbox(label="Progress Log", lines=10, interactive=False)
run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table, progress_log])
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"βœ… SPACE_HOST: https://{space_host_startup}.hf.space")
if space_id_startup:
print(f"βœ… SPACE_ID: https://huggingface.co/spaces/{space_id_startup}")
print("Launching Gradio Interface...")
demo.launch(debug=True, share=False)