Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -2,182 +2,106 @@ import os
|
|
2 |
import gradio as gr
|
3 |
import requests
|
4 |
import pandas as pd
|
5 |
-
from
|
6 |
|
7 |
-
#
|
8 |
-
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
9 |
|
10 |
-
|
11 |
-
class BasicAgent:
|
12 |
-
def __init__(self):
|
13 |
-
print("BasicAgent initialized.")
|
14 |
-
def __call__(self, question: str) -> str:
|
15 |
-
print(f"Agent received question (first 50 chars): {question[:50]}...")
|
16 |
-
fixed_answer = "This is a default answer."
|
17 |
-
print(f"Agent returning fixed answer: {fixed_answer}")
|
18 |
-
return fixed_answer
|
19 |
|
|
|
20 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
21 |
"""
|
22 |
-
|
23 |
-
and displays the results.
|
24 |
"""
|
25 |
-
space_id = os.getenv("SPACE_ID")
|
26 |
|
27 |
if profile:
|
28 |
username = f"{profile.username}"
|
29 |
-
print(f"
|
30 |
else:
|
31 |
-
print("
|
32 |
-
return "
|
33 |
|
34 |
api_url = DEFAULT_API_URL
|
35 |
questions_url = f"{api_url}/questions"
|
|
|
36 |
submit_url = f"{api_url}/submit"
|
37 |
|
38 |
-
#
|
39 |
try:
|
40 |
-
|
|
|
41 |
except Exception as e:
|
42 |
-
print(f"Error
|
43 |
-
return f"Error
|
44 |
-
|
45 |
-
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
46 |
-
print(agent_code)
|
47 |
|
48 |
-
# 2.
|
49 |
-
print(f"
|
50 |
try:
|
51 |
response = requests.get(questions_url, timeout=15)
|
52 |
response.raise_for_status()
|
53 |
questions_data = response.json()
|
54 |
if not questions_data:
|
55 |
-
print("
|
56 |
-
return "
|
57 |
-
print(f"
|
58 |
-
except
|
59 |
-
print(f"Error
|
60 |
-
return f"Error
|
61 |
|
62 |
-
# 3.
|
63 |
results_log = []
|
64 |
answers_payload = []
|
65 |
-
print(f"
|
66 |
for item in questions_data:
|
67 |
task_id = item.get("task_id")
|
68 |
-
question_text = item.get("question")
|
|
|
|
|
|
|
69 |
if not task_id or question_text is None:
|
70 |
-
print(f"
|
71 |
continue
|
72 |
try:
|
73 |
-
submitted_answer = agent(question_text) #
|
74 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
75 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
76 |
except Exception as e:
|
77 |
-
print(f"Error
|
78 |
-
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"
|
79 |
|
80 |
-
|
81 |
-
|
82 |
-
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
83 |
-
|
84 |
-
# 4. Prepare Submission
|
85 |
-
submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
|
86 |
-
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
87 |
-
print(status_update)
|
88 |
-
|
89 |
-
# 5. Submit
|
90 |
-
print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
|
91 |
try:
|
92 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
93 |
response.raise_for_status()
|
94 |
result_data = response.json()
|
95 |
-
final_status = (
|
96 |
-
|
97 |
-
|
98 |
-
f"Overall Score: {result_data.get('score', 'N/A')}% "
|
99 |
-
f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
|
100 |
-
f"Message: {result_data.get('message', 'No message received.')}"
|
101 |
-
)
|
102 |
-
print("Submission successful.")
|
103 |
-
results_df = pd.DataFrame(results_log)
|
104 |
-
return final_status, results_df
|
105 |
-
except requests.exceptions.HTTPError as e:
|
106 |
-
error_detail = f"Server responded with status {e.response.status_code}."
|
107 |
-
try:
|
108 |
-
error_json = e.response.json()
|
109 |
-
error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
|
110 |
-
except requests.exceptions.JSONDecodeError:
|
111 |
-
error_detail += f" Response: {e.response.text[:500]}"
|
112 |
-
status_message = f"Submission Failed: {error_detail}"
|
113 |
-
print(status_message)
|
114 |
-
results_df = pd.DataFrame(results_log)
|
115 |
-
return status_message, results_df
|
116 |
-
except requests.exceptions.Timeout:
|
117 |
-
status_message = "Submission Failed: The request timed out."
|
118 |
-
print(status_message)
|
119 |
-
results_df = pd.DataFrame(results_log)
|
120 |
-
return status_message, results_df
|
121 |
except requests.exceptions.RequestException as e:
|
122 |
-
|
123 |
-
|
124 |
-
results_df = pd.DataFrame(results_log)
|
125 |
-
return status_message, results_df
|
126 |
-
except Exception as e:
|
127 |
-
status_message = f"An unexpected error occurred during submission: {e}"
|
128 |
-
print(status_message)
|
129 |
-
results_df = pd.DataFrame(results_log)
|
130 |
-
return status_message, results_df
|
131 |
-
|
132 |
|
133 |
-
|
134 |
-
# --- Build Gradio Interface using Blocks ---
|
135 |
with gr.Blocks() as demo:
|
136 |
-
gr.Markdown("#
|
137 |
-
gr.Markdown(
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
|
143 |
-
---
|
144 |
-
**Disclaimers:**
|
145 |
-
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).
|
146 |
-
This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution.
|
147 |
-
"""
|
148 |
-
)
|
149 |
-
|
150 |
-
gr.LoginButton()
|
151 |
|
152 |
-
|
|
|
|
|
153 |
|
154 |
-
|
155 |
-
|
|
|
|
|
156 |
|
157 |
-
run_button.click(
|
158 |
-
fn=run_and_submit_all,
|
159 |
-
outputs=[status_output, results_table]
|
160 |
-
)
|
161 |
|
162 |
if __name__ == "__main__":
|
163 |
-
|
164 |
-
space_host_startup = os.getenv("SPACE_HOST")
|
165 |
-
space_id_startup = os.getenv("SPACE_ID")
|
166 |
-
|
167 |
-
if space_host_startup:
|
168 |
-
print(f"✅ SPACE_HOST found: {space_host_startup}")
|
169 |
-
print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
|
170 |
-
else:
|
171 |
-
print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
|
172 |
-
|
173 |
-
if space_id_startup:
|
174 |
-
print(f"✅ SPACE_ID found: {space_id_startup}")
|
175 |
-
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
|
176 |
-
print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
|
177 |
-
else:
|
178 |
-
print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
|
179 |
-
|
180 |
-
print("-"*(60 + len(" App Starting ")) + "\n")
|
181 |
-
|
182 |
-
print("Launching Gradio Interface for DuckDuckGo Agent Evaluation...")
|
183 |
-
demo.launch(debug=True, share=False)
|
|
|
2 |
import gradio as gr
|
3 |
import requests
|
4 |
import pandas as pd
|
5 |
+
from dotenv import load_dotenv
|
6 |
|
7 |
+
load_dotenv() # Cargar variables de entorno
|
|
|
8 |
|
9 |
+
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
+
# Función para ejecutar y enviar todas las respuestas
|
12 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
13 |
"""
|
14 |
+
Recoge todas las preguntas, ejecuta el agente sobre ellas, envía las respuestas y muestra los resultados.
|
|
|
15 |
"""
|
16 |
+
space_id = os.getenv("SPACE_ID") # ID del espacio para enlaces al código
|
17 |
|
18 |
if profile:
|
19 |
username = f"{profile.username}"
|
20 |
+
print(f"Usuario logueado: {username}")
|
21 |
else:
|
22 |
+
print("Usuario no logueado.")
|
23 |
+
return "Por favor, inicia sesión en Hugging Face.", None
|
24 |
|
25 |
api_url = DEFAULT_API_URL
|
26 |
questions_url = f"{api_url}/questions"
|
27 |
+
attachments_url = f"{api_url}/files/"
|
28 |
submit_url = f"{api_url}/submit"
|
29 |
|
30 |
+
# Crear agente (modificado)
|
31 |
try:
|
32 |
+
print("Iniciando agente...")
|
33 |
+
agent = agent.BasicAgent() # Usar el agente principal
|
34 |
except Exception as e:
|
35 |
+
print(f"Error al iniciar el agente: {e}")
|
36 |
+
return f"Error al iniciar el agente: {e}", None
|
|
|
|
|
|
|
37 |
|
38 |
+
# 2. Recoger las preguntas
|
39 |
+
print(f"Recogiendo preguntas desde: {questions_url}")
|
40 |
try:
|
41 |
response = requests.get(questions_url, timeout=15)
|
42 |
response.raise_for_status()
|
43 |
questions_data = response.json()
|
44 |
if not questions_data:
|
45 |
+
print("La lista de preguntas está vacía.")
|
46 |
+
return "La lista de preguntas está vacía.", None
|
47 |
+
print(f"Se recogieron {len(questions_data)} preguntas.")
|
48 |
+
except Exception as e:
|
49 |
+
print(f"Error recogiendo preguntas: {e}")
|
50 |
+
return f"Error recogiendo preguntas: {e}", None
|
51 |
|
52 |
+
# 3. Ejecutar el agente
|
53 |
results_log = []
|
54 |
answers_payload = []
|
55 |
+
print(f"Ejecutando el agente sobre {len(questions_data)} preguntas...")
|
56 |
for item in questions_data:
|
57 |
task_id = item.get("task_id")
|
58 |
+
question_text = item.get("question", "")
|
59 |
+
attachment_b64 = item.get("attachment_b64", "")
|
60 |
+
if attachment_b64:
|
61 |
+
question_text = f"{question_text}\n\n[ATTACHMENT:]\n{attachment_b64}"
|
62 |
if not task_id or question_text is None:
|
63 |
+
print(f"Saltando tarea con ID o pregunta faltante: {item}")
|
64 |
continue
|
65 |
try:
|
66 |
+
submitted_answer = agent.forward(question_text) # Respuesta del agente
|
67 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
68 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
69 |
except Exception as e:
|
70 |
+
print(f"Error ejecutando agente en tarea {task_id}: {e}")
|
71 |
+
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"ERROR: {e}"})
|
72 |
|
73 |
+
# 4. Enviar las respuestas
|
74 |
+
submission_data = {"username": username.strip(), "agent_code": "agent_code_placeholder", "answers": answers_payload}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
try:
|
76 |
response = requests.post(submit_url, json=submission_data, timeout=60)
|
77 |
response.raise_for_status()
|
78 |
result_data = response.json()
|
79 |
+
final_status = f"¡Envío exitoso!\nUsuario: {result_data.get('username')}\nPuntaje final: {result_data.get('score', 'N/A')}%"
|
80 |
+
print("Envío exitoso.")
|
81 |
+
return final_status, pd.DataFrame(results_log)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
82 |
except requests.exceptions.RequestException as e:
|
83 |
+
print(f"Error al enviar respuestas: {e}")
|
84 |
+
return f"Error al enviar respuestas: {e}", pd.DataFrame(results_log)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
|
86 |
+
# Interfaz Gradio
|
|
|
87 |
with gr.Blocks() as demo:
|
88 |
+
gr.Markdown("# Evaluación Básica del Agente")
|
89 |
+
gr.Markdown("""
|
90 |
+
**Instrucciones:**
|
91 |
+
1. Modifica este espacio con tu lógica de agente y las herramientas necesarias.
|
92 |
+
2. Inicia sesión en Hugging Face con el botón abajo.
|
93 |
+
3. Haz clic en 'Ejecutar Evaluación y Enviar Todas las Respuestas' para obtener resultados.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
94 |
|
95 |
+
**Aviso:**
|
96 |
+
Puede tomar tiempo procesar las respuestas, así que ten paciencia.
|
97 |
+
""")
|
98 |
|
99 |
+
gr.LoginButton()
|
100 |
+
run_button = gr.Button("Ejecutar Evaluación y Enviar Todas las Respuestas")
|
101 |
+
status_output = gr.Textbox(label="Resultado de Ejecución / Envío", lines=5, interactive=False)
|
102 |
+
results_table = gr.DataFrame(label="Preguntas y Respuestas del Agente", wrap=True)
|
103 |
|
104 |
+
run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
|
|
|
|
|
|
|
105 |
|
106 |
if __name__ == "__main__":
|
107 |
+
demo.launch(debug=True, share=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|