MF180's picture
Update app.py
7d2add2 verified
Raw History Blame Contribute Delete
19.8 kB
import os
import io
import time
import contextlib
import gradio as gr
import requests
import inspect
import pandas as pd
import spaces
from duckduckgo_search import DDGS
# (Keep Constants as is)
# --- Constants ---
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
GROQ_API_URL = "https://api.groq.com/openai/v1/chat/completions"
GROQ_TRANSCRIBE_URL = "https://api.groq.com/openai/v1/audio/transcriptions"
GROQ_MODEL = "llama-3.3-70b-versatile"
# --- Dummy GPU function ---
# ZeroGPU requires at least one @spaces.GPU-decorated function to exist at
# startup. Our agent never actually needs a GPU (it just makes API calls),
# but the long evaluation loop must NOT be decorated with @spaces.GPU,
# because ZeroGPU enforces a short max duration per GPU call and the full
# 20-question loop takes much longer than that, causing it to be aborted.
# This tiny no-op function satisfies the startup check without limiting
# the runtime of the real evaluation loop below.
@spaces.GPU(duration=10)
def _zerogpu_warmup():
return "ok"
# --- Basic Agent Definition ---
class BasicAgent:
def __init__(self):
print("BasicAgent initialized.")
self.api_url = DEFAULT_API_URL
self.groq_api_key = os.environ.get("GROQ_API_KEY")
# ---------- Utility: reversed text detection ----------
def is_reversed_text(self, text: str) -> bool:
reversed_text = text[::-1]
common_words = [" the ", " is ", " you ", " and ", " of ", " a ", " to ", " write ", " answer ", " sentence "]
hits = sum(1 for w in common_words if w in reversed_text.lower())
return hits >= 2
# ---------- Utility: web search ----------
def web_search(self, query: str, max_results: int = 5) -> str:
try:
with DDGS() as ddgs:
results = list(ddgs.text(query, max_results=max_results))
if not results:
return "No search results found."
formatted = []
for r in results:
title = r.get("title", "")
body = r.get("body", "")
href = r.get("href", "")
formatted.append(f"Title: {title}\nSnippet: {body}\nURL: {href}")
return "\n\n".join(formatted)
except Exception as e:
print(f"Search error: {e}")
return f"Search failed: {e}"
# ---------- Utility: LLM call via Groq, with retry ----------
def call_llm(self, messages, max_tokens=512, retries=3):
if not self.groq_api_key:
raise RuntimeError("GROQ_API_KEY is not set in this Space's secrets.")
headers = {
"Authorization": f"Bearer {self.groq_api_key}",
"Content-Type": "application/json",
}
payload = {
"model": GROQ_MODEL,
"messages": messages,
"max_tokens": max_tokens,
"temperature": 0,
}
last_error = None
for attempt in range(retries):
try:
resp = requests.post(GROQ_API_URL, headers=headers, json=payload, timeout=60)
if resp.status_code != 200:
print(f"Groq API returned status {resp.status_code}. Body: {resp.text[:1000]}")
resp.raise_for_status()
data = resp.json()
return data["choices"][0]["message"]["content"].strip()
except Exception as e:
last_error = e
wait = 5 * (attempt + 1)
print(f"LLM call failed (attempt {attempt+1}/{retries}): {e}. Retrying in {wait}s...")
time.sleep(wait)
raise last_error
# ---------- Utility: download attached file ----------
def fetch_file(self, task_id: str):
"""Downloads the file attached to a task, if any. Returns (bytes, None) or (None, error).
NOTE: as of this writing, the official course scoring Space's /files/{task_id}
endpoint is broken (returns 404 for every task) due to a known upstream issue,
so this will currently fail for all file-based questions. Left in place so it
will start working automatically once Hugging Face fixes the endpoint.
"""
url = f"{self.api_url}/files/{task_id}"
try:
resp = requests.get(url, timeout=30)
resp.raise_for_status()
return resp.content, None
except Exception as e:
print(f"File fetch error for task {task_id}: {e}")
return None, str(e)
# ---------- Tool: transcribe audio via Groq Whisper ----------
def transcribe_audio(self, audio_bytes: bytes, file_name: str) -> str:
if not self.groq_api_key:
return "Transcription unavailable: GROQ_API_KEY not set."
try:
headers = {"Authorization": f"Bearer {self.groq_api_key}"}
files = {"file": (file_name, audio_bytes)}
data = {"model": "whisper-large-v3"}
resp = requests.post(GROQ_TRANSCRIBE_URL, headers=headers, files=files, data=data, timeout=60)
resp.raise_for_status()
return resp.json().get("text", "").strip()
except Exception as e:
print(f"Transcription error: {e}")
return f"Transcription failed: {e}"
# ---------- Tool: read spreadsheet ----------
def read_spreadsheet(self, file_bytes: bytes, file_name: str) -> str:
try:
if file_name.lower().endswith(".csv"):
df = pd.read_csv(io.BytesIO(file_bytes))
else:
df = pd.read_excel(io.BytesIO(file_bytes))
return df.to_string(max_rows=200)
except Exception as e:
print(f"Spreadsheet read error: {e}")
return f"Spreadsheet read failed: {e}"
# ---------- Tool: execute attached Python code ----------
def run_python_code(self, code_bytes: bytes) -> str:
code_str = code_bytes.decode("utf-8", errors="replace")
output_buffer = io.StringIO()
try:
with contextlib.redirect_stdout(output_buffer):
exec_globals = {}
exec(code_str, exec_globals)
output = output_buffer.getvalue().strip()
if not output:
output = "Code ran but produced no printed output."
return output
except Exception as e:
print(f"Code execution error: {e}")
return f"Code execution failed: {e}"
# ---------- Main entry point ----------
def __call__(self, question: str, task_id: str = None, file_name: str = None) -> str:
print(f"Agent received question (first 50 chars): {question[:50]}...")
# Step 0: Handle reversed-text questions directly
if self.is_reversed_text(question):
print("Detected reversed text question.")
decoded = question[::-1]
answer_prompt = (
"The following is a normal English instruction. Follow it exactly and "
"return ONLY the final answer, nothing else, no explanation:\n\n"
f"{decoded}"
)
try:
answer = self.call_llm(
messages=[
{"role": "system", "content": "You follow instructions precisely and answer with only the final answer."},
{"role": "user", "content": answer_prompt},
],
max_tokens=100,
)
except Exception as e:
print(f"Error solving reversed text question: {e}")
answer = "Unknown"
print(f"Agent returning answer: {answer}")
return answer
# Step 1: If there's an attached file, fetch and process it
file_context = ""
if file_name:
print(f"Question has attached file: {file_name}")
file_bytes, err = self.fetch_file(task_id)
if file_bytes:
lower_name = file_name.lower()
if lower_name.endswith((".mp3", ".wav", ".m4a")):
transcript = self.transcribe_audio(file_bytes, file_name)
file_context = f"Transcript of attached audio file '{file_name}':\n{transcript}"
elif lower_name.endswith((".xlsx", ".xls", ".csv")):
table_text = self.read_spreadsheet(file_bytes, file_name)
file_context = f"Contents of attached spreadsheet '{file_name}':\n{table_text}"
elif lower_name.endswith(".py"):
code_output = self.run_python_code(file_bytes)
print(f"Agent returning code execution result: {code_output}")
return code_output
elif lower_name.endswith((".txt", ".json", ".md")):
try:
file_context = f"Contents of attached file '{file_name}':\n{file_bytes.decode('utf-8', errors='replace')}"
except Exception as e:
file_context = f"Could not decode attached file '{file_name}': {e}"
else:
file_context = f"An attached file '{file_name}' was provided but its type is not supported by this agent."
else:
# Known upstream issue: /files/{task_id} currently 404s for every
# task. Tell the LLM plainly rather than pretending we have context.
file_context = f"An attachment '{file_name}' was referenced but could not be downloaded (server error: {err})."
# Step 2: Web search for additional context
search_results = self.web_search(question)
# Step 3: Ask the LLM to answer using all available context
system_prompt = (
"You are a general AI assistant answering benchmark questions. "
"You may be given web search results and/or attached file contents as context below. "
"Use them if relevant; if the context does not help, rely on your own knowledge. "
"Report your final answer as a concise string: a number, a short phrase, "
"or a comma separated list, with no explanation, no extra words, "
"no units unless explicitly asked for, and no leading phrases like 'The answer is'. "
"Answer with ONLY the final answer, nothing else. "
"If you genuinely cannot determine the answer from the given context or your own knowledge, "
"answer with your single best guess rather than refusing."
)
user_prompt_parts = [f"Question: {question}"]
if file_context:
user_prompt_parts.append(f"Attached file context:\n{file_context}")
user_prompt_parts.append(f"Web search results:\n{search_results}")
user_prompt_parts.append("Based on the above, provide only the final answer.")
user_prompt = "\n\n".join(user_prompt_parts)
try:
answer = self.call_llm(
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": user_prompt},
],
max_tokens=512,
)
except Exception as e:
print(f"Error calling LLM after retries: {e}")
answer = "Unknown"
print(f"Agent returning answer: {answer}")
return answer
# NOTE: no @spaces.GPU here on purpose. The full evaluation loop can take
# several minutes (20 questions x search + LLM + retries), which exceeds
# ZeroGPU's allowed duration for a single GPU call and causes it to be
# aborted. Running this as plain CPU code avoids that entirely.
def run_and_submit_all( profile: gr.OAuthProfile | None):
"""
Fetches all questions, runs the BasicAgent on them, submits all answers,
and displays the results.
"""
# --- Determine HF Space Runtime URL and Repo URL ---
space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
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. Instantiate Agent ( modify this part to create your agent)
try:
agent = BasicAgent()
except Exception as e:
print(f"Error instantiating agent: {e}")
return f"Error initializing agent: {e}", None
# In the case of an app running as a hugging Face space, this link points toward your codebase ( usefull for others so please keep it public)
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
print(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:
print("Fetched questions list is empty.")
return "Fetched questions list is empty or invalid format.", None
print(f"Fetched {len(questions_data)} questions.")
except requests.exceptions.RequestException as e:
print(f"Error fetching questions: {e}")
return f"Error fetching questions: {e}", None
except requests.exceptions.JSONDecodeError as e:
print(f"Error decoding JSON response from questions endpoint: {e}")
print(f"Response text: {response.text[:500]}")
return f"Error decoding server response for questions: {e}", None
except Exception as e:
print(f"An unexpected error occurred fetching questions: {e}")
return f"An unexpected error occurred fetching questions: {e}", None
# 3. Run your 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")
file_name = item.get("file_name") or None
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, task_id=task_id, file_name=file_name)
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:
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}"})
# Small delay between questions to help avoid rate limits
time.sleep(2)
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)
# 4. Prepare Submission
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)
# 5. Submit
print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
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.')}"
)
print("Submission successful.")
results_df = pd.DataFrame(results_log)
return final_status, results_df
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 requests.exceptions.JSONDecodeError:
error_detail += f" Response: {e.response.text[:500]}"
status_message = f"Submission Failed: {error_detail}"
print(status_message)
results_df = pd.DataFrame(results_log)
return status_message, results_df
except requests.exceptions.Timeout:
status_message = "Submission Failed: The request timed out."
print(status_message)
results_df = pd.DataFrame(results_log)
return status_message, results_df
except requests.exceptions.RequestException as e:
status_message = f"Submission Failed: Network error - {e}"
print(status_message)
results_df = pd.DataFrame(results_log)
return status_message, results_df
except Exception as e:
status_message = f"An unexpected error occurred during submission: {e}"
print(status_message)
results_df = pd.DataFrame(results_log)
return status_message, results_df
# --- Build Gradio Interface using Blocks ---
with gr.Blocks() as demo:
gr.Markdown("# Basic Agent Evaluation Runner")
gr.Markdown(
"""
**Instructions:**
1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
2. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
---
**Disclaimers:**
Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).
This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.
"""
)
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_ID environment variable not found (running locally?). Repo URL cannot be determined.")
print("-"*(60 + len(" App Starting ")) + "\n")
print("Launching Gradio Interface for Basic Agent Evaluation...")
demo.launch(debug=True, share=False)