Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,78 +2,76 @@ import os
|
|
| 2 |
import gradio as gr
|
| 3 |
import requests
|
| 4 |
import pandas as pd
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
# --------------------------
|
| 10 |
-
# SMART RULE-BASED AGENT
|
| 11 |
-
# --------------------------
|
| 12 |
-
def agent_answer(question: str) -> str:
|
| 13 |
-
q = question.lower()
|
| 14 |
-
|
| 15 |
-
# 1️⃣ Mercedes Sosa
|
| 16 |
-
if "mercedes sosa" in q and "studio albums" in q:
|
| 17 |
-
return "4"
|
| 18 |
-
|
| 19 |
-
# 2️⃣ 1928 Olympics – least athletes
|
| 20 |
-
if "1928 summer olympics" in q and "least number of athletes" in q:
|
| 21 |
-
return "AFG"
|
| 22 |
-
|
| 23 |
-
# 3️⃣ Opposite of left
|
| 24 |
-
if "opposite of left" in q:
|
| 25 |
-
return "right"
|
| 26 |
-
|
| 27 |
-
# 4️⃣ Malko Competition
|
| 28 |
-
if "malko competition" in q and "first name" in q:
|
| 29 |
-
return "Erik"
|
| 30 |
-
|
| 31 |
-
# 5️⃣ Bird species in video
|
| 32 |
-
if "bird species" in q:
|
| 33 |
-
return "4"
|
| 34 |
-
|
| 35 |
-
# 6️⃣ Chess move fallback
|
| 36 |
-
if "chess" in q:
|
| 37 |
-
return "Qh5"
|
| 38 |
-
|
| 39 |
-
# 7️⃣ Excel sales question (safe numeric format)
|
| 40 |
-
if "excel file" in q and "total sales" in q:
|
| 41 |
-
return "1234.56"
|
| 42 |
-
|
| 43 |
-
# 8️⃣ Pitcher question (safe format)
|
| 44 |
-
if "pitcher" in q and "taishō tamai" in q:
|
| 45 |
-
return "Suzuki, Tanaka"
|
| 46 |
|
| 47 |
-
#
|
| 48 |
-
|
|
|
|
| 49 |
|
|
|
|
| 50 |
|
| 51 |
-
#
|
| 52 |
-
#
|
| 53 |
-
#
|
| 54 |
-
def
|
| 55 |
if not profile:
|
| 56 |
-
return "❌ Please login to Hugging Face
|
| 57 |
|
| 58 |
username = profile.username.strip()
|
| 59 |
-
space_id = os.getenv("SPACE_ID"
|
| 60 |
-
agent_code = f"https://huggingface.co/spaces/{space_id}"
|
| 61 |
|
| 62 |
# Fetch questions
|
| 63 |
-
|
| 64 |
-
questions = requests.get(f"{API_URL}/questions", timeout=20).json()
|
| 65 |
-
except Exception as e:
|
| 66 |
-
return f"❌ Error fetching questions: {e}", None
|
| 67 |
|
| 68 |
answers = []
|
| 69 |
logs = []
|
| 70 |
|
| 71 |
for q in questions:
|
| 72 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 73 |
answers.append({
|
| 74 |
"task_id": q["task_id"],
|
| 75 |
"submitted_answer": ans
|
| 76 |
})
|
|
|
|
| 77 |
logs.append({
|
| 78 |
"Task ID": q["task_id"],
|
| 79 |
"Question": q["question"],
|
|
@@ -86,46 +84,29 @@ def run_and_submit(profile: gr.OAuthProfile | None):
|
|
| 86 |
"answers": answers
|
| 87 |
}
|
| 88 |
|
| 89 |
-
|
| 90 |
-
response = requests.post(
|
| 91 |
-
f"{API_URL}/submit",
|
| 92 |
-
json=payload,
|
| 93 |
-
timeout=60
|
| 94 |
-
)
|
| 95 |
-
response.raise_for_status()
|
| 96 |
-
result = response.json()
|
| 97 |
-
except Exception as e:
|
| 98 |
-
return f"❌ Submission failed: {e}", pd.DataFrame(logs)
|
| 99 |
|
| 100 |
status = (
|
| 101 |
f"✅ Submission Successful!\n"
|
| 102 |
-
f"User: {
|
| 103 |
-
f"Score: {
|
| 104 |
-
f"Correct: {
|
| 105 |
-
f"Message: {
|
| 106 |
)
|
| 107 |
|
| 108 |
return status, pd.DataFrame(logs)
|
| 109 |
|
| 110 |
-
|
| 111 |
-
#
|
| 112 |
-
#
|
| 113 |
-
# --------------------------
|
| 114 |
with gr.Blocks() as demo:
|
| 115 |
-
gr.Markdown("# 🤖 GAIA Level-1 Agent (
|
| 116 |
-
gr.Markdown("Login → Run → Submit")
|
| 117 |
-
|
| 118 |
gr.LoginButton()
|
| 119 |
-
submit_btn = gr.Button("Run Evaluation & Submit")
|
| 120 |
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
submit_btn.click(
|
| 125 |
-
fn=run_and_submit,
|
| 126 |
-
outputs=[status_box, log_table]
|
| 127 |
-
)
|
| 128 |
|
|
|
|
| 129 |
|
| 130 |
-
|
| 131 |
-
demo.launch()
|
|
|
|
| 2 |
import gradio as gr
|
| 3 |
import requests
|
| 4 |
import pandas as pd
|
| 5 |
+
from openai import OpenAI
|
| 6 |
+
|
| 7 |
+
# =============================
|
| 8 |
+
# CONSTANTS (DO NOT CHANGE)
|
| 9 |
+
# =============================
|
| 10 |
+
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 11 |
+
|
| 12 |
+
SYSTEM_PROMPT = """
|
| 13 |
+
You are a general AI assistant.
|
| 14 |
+
Answer the question and finish your response with:
|
| 15 |
+
|
| 16 |
+
FINAL ANSWER: <answer>
|
| 17 |
+
|
| 18 |
+
Rules:
|
| 19 |
+
- If number: no commas, no units unless specified
|
| 20 |
+
- If string: no articles, no abbreviations
|
| 21 |
+
- If list: comma-separated, minimal words
|
| 22 |
+
"""
|
| 23 |
+
|
| 24 |
+
# =============================
|
| 25 |
+
# MODEL
|
| 26 |
+
# =============================
|
| 27 |
+
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
|
| 28 |
+
|
| 29 |
+
def llm_answer(question: str) -> str:
|
| 30 |
+
response = client.chat.completions.create(
|
| 31 |
+
model="gpt-4o-mini",
|
| 32 |
+
messages=[
|
| 33 |
+
{"role": "system", "content": SYSTEM_PROMPT},
|
| 34 |
+
{"role": "user", "content": question}
|
| 35 |
+
],
|
| 36 |
+
temperature=0
|
| 37 |
+
)
|
| 38 |
|
| 39 |
+
text = response.choices[0].message.content
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
|
| 41 |
+
# Extract FINAL ANSWER only
|
| 42 |
+
if "FINAL ANSWER:" in text:
|
| 43 |
+
return text.split("FINAL ANSWER:")[-1].strip()
|
| 44 |
|
| 45 |
+
return text.strip()
|
| 46 |
|
| 47 |
+
# =============================
|
| 48 |
+
# GAIA PIPELINE
|
| 49 |
+
# =============================
|
| 50 |
+
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 51 |
if not profile:
|
| 52 |
+
return "❌ Please login to Hugging Face", None
|
| 53 |
|
| 54 |
username = profile.username.strip()
|
| 55 |
+
space_id = os.getenv("SPACE_ID")
|
| 56 |
+
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
| 57 |
|
| 58 |
# Fetch questions
|
| 59 |
+
questions = requests.get(f"{DEFAULT_API_URL}/questions").json()
|
|
|
|
|
|
|
|
|
|
| 60 |
|
| 61 |
answers = []
|
| 62 |
logs = []
|
| 63 |
|
| 64 |
for q in questions:
|
| 65 |
+
try:
|
| 66 |
+
ans = llm_answer(q["question"])
|
| 67 |
+
except Exception as e:
|
| 68 |
+
ans = "I don't know"
|
| 69 |
+
|
| 70 |
answers.append({
|
| 71 |
"task_id": q["task_id"],
|
| 72 |
"submitted_answer": ans
|
| 73 |
})
|
| 74 |
+
|
| 75 |
logs.append({
|
| 76 |
"Task ID": q["task_id"],
|
| 77 |
"Question": q["question"],
|
|
|
|
| 84 |
"answers": answers
|
| 85 |
}
|
| 86 |
|
| 87 |
+
res = requests.post(f"{DEFAULT_API_URL}/submit", json=payload).json()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 88 |
|
| 89 |
status = (
|
| 90 |
f"✅ Submission Successful!\n"
|
| 91 |
+
f"User: {res.get('username')}\n"
|
| 92 |
+
f"Score: {res.get('score')}%\n"
|
| 93 |
+
f"Correct: {res.get('correct_count')}/{res.get('total_attempted')}\n"
|
| 94 |
+
f"Message: {res.get('message')}"
|
| 95 |
)
|
| 96 |
|
| 97 |
return status, pd.DataFrame(logs)
|
| 98 |
|
| 99 |
+
# =============================
|
| 100 |
+
# UI
|
| 101 |
+
# =============================
|
|
|
|
| 102 |
with gr.Blocks() as demo:
|
| 103 |
+
gr.Markdown("# 🤖 GAIA Level-1 Agent (Unit-4)")
|
|
|
|
|
|
|
| 104 |
gr.LoginButton()
|
|
|
|
| 105 |
|
| 106 |
+
run_btn = gr.Button("Run Evaluation & Submit")
|
| 107 |
+
status = gr.Textbox(label="Submission Result", lines=5)
|
| 108 |
+
table = gr.Dataframe(label="Questions & Answers")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 109 |
|
| 110 |
+
run_btn.click(run_and_submit_all, outputs=[status, table])
|
| 111 |
|
| 112 |
+
demo.launch()
|
|
|