Update app.py
Browse files
app.py
CHANGED
|
@@ -8,12 +8,9 @@ import re
|
|
| 8 |
from langchain_core.messages import HumanMessage
|
| 9 |
from langgraph_agent import build_graph
|
| 10 |
|
| 11 |
-
# --- Constants ---
|
| 12 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 13 |
|
| 14 |
-
# --- Basic Agent Definition ---
|
| 15 |
class BasicAgent:
|
| 16 |
-
"""A langgraph agent."""
|
| 17 |
def __init__(self):
|
| 18 |
print("BasicAgent initialized.")
|
| 19 |
self.graph = build_graph()
|
|
@@ -36,24 +33,21 @@ class BasicAgent:
|
|
| 36 |
|
| 37 |
answer = answer.strip()
|
| 38 |
|
| 39 |
-
# Extract final answer after "FINAL ANSWER:"
|
| 40 |
match = re.search(r'FINAL ANSWER:\s*(.*)', answer, re.IGNORECASE | re.DOTALL)
|
| 41 |
if match:
|
| 42 |
final_answer = match.group(1).strip()
|
| 43 |
-
# Remove surrounding quotes if present
|
| 44 |
if (final_answer.startswith('"') and final_answer.endswith('"')) or \
|
| 45 |
(final_answer.startswith("'") and final_answer.endswith("'")):
|
| 46 |
final_answer = final_answer[1:-1].strip()
|
| 47 |
answer = final_answer
|
| 48 |
else:
|
| 49 |
-
print("Warning: 'FINAL ANSWER:' not found
|
| 50 |
|
| 51 |
if not answer:
|
| 52 |
answer = "I am unable to determine the information using the available tools."
|
| 53 |
|
| 54 |
return answer
|
| 55 |
|
| 56 |
-
|
| 57 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 58 |
space_id = os.getenv("SPACE_ID")
|
| 59 |
|
|
@@ -102,11 +96,22 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 102 |
continue
|
| 103 |
try:
|
| 104 |
model_answer = agent(question_text, task_id=task_id)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 105 |
print(f"Answer for task {task_id}: '{model_answer}'")
|
| 106 |
answers_payload.append({
|
| 107 |
"task_id": task_id,
|
| 108 |
"model_answer": model_answer,
|
| 109 |
-
|
| 110 |
})
|
| 111 |
results_log.append({
|
| 112 |
"Task ID": task_id,
|
|
@@ -118,7 +123,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 118 |
answers_payload.append({
|
| 119 |
"task_id": task_id,
|
| 120 |
"model_answer": f"AGENT ERROR: {e}",
|
| 121 |
-
|
| 122 |
})
|
| 123 |
results_log.append({
|
| 124 |
"Task ID": task_id,
|
|
@@ -130,7 +135,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 130 |
print("Agent did not produce any answers to submit.")
|
| 131 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
| 132 |
|
| 133 |
-
# Serialize
|
| 134 |
json_lines_str = "\n".join(json.dumps(ans, ensure_ascii=False) for ans in answers_payload)
|
| 135 |
file_like = io.BytesIO(json_lines_str.encode("utf-8"))
|
| 136 |
file_like.name = "submission.jsonl"
|
|
@@ -163,35 +168,13 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 163 |
print("Submission successful.")
|
| 164 |
results_df = pd.DataFrame(results_log)
|
| 165 |
return final_status, results_df
|
| 166 |
-
except requests.exceptions.HTTPError as e:
|
| 167 |
-
error_detail = f"Server responded with status {e.response.status_code}."
|
| 168 |
-
try:
|
| 169 |
-
error_json = e.response.json()
|
| 170 |
-
error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
|
| 171 |
-
except requests.exceptions.JSONDecodeError:
|
| 172 |
-
error_detail += f" Response: {e.response.text[:500]}"
|
| 173 |
-
status_message = f"Submission Failed: {error_detail}"
|
| 174 |
-
print(status_message)
|
| 175 |
-
results_df = pd.DataFrame(results_log)
|
| 176 |
-
return status_message, results_df
|
| 177 |
-
except requests.exceptions.Timeout:
|
| 178 |
-
status_message = "Submission Failed: The request timed out."
|
| 179 |
-
print(status_message)
|
| 180 |
-
results_df = pd.DataFrame(results_log)
|
| 181 |
-
return status_message, results_df
|
| 182 |
-
except requests.exceptions.RequestException as e:
|
| 183 |
-
status_message = f"Submission Failed: Network error - {e}"
|
| 184 |
-
print(status_message)
|
| 185 |
-
results_df = pd.DataFrame(results_log)
|
| 186 |
-
return status_message, results_df
|
| 187 |
except Exception as e:
|
| 188 |
-
status_message = f"
|
| 189 |
print(status_message)
|
| 190 |
results_df = pd.DataFrame(results_log)
|
| 191 |
return status_message, results_df
|
| 192 |
|
| 193 |
|
| 194 |
-
# --- Build Gradio Interface using Blocks ---
|
| 195 |
with gr.Blocks() as demo:
|
| 196 |
gr.Markdown("# Basic Agent Evaluation Runner")
|
| 197 |
gr.Markdown(
|
|
@@ -226,6 +209,7 @@ with gr.Blocks() as demo:
|
|
| 226 |
outputs=[status_output, results_table]
|
| 227 |
)
|
| 228 |
|
|
|
|
| 229 |
if __name__ == "__main__":
|
| 230 |
print("\n" + "-"*30 + " App Starting " + "-"*30)
|
| 231 |
space_host_startup = os.getenv("SPACE_HOST")
|
|
|
|
| 8 |
from langchain_core.messages import HumanMessage
|
| 9 |
from langgraph_agent import build_graph
|
| 10 |
|
|
|
|
| 11 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 12 |
|
|
|
|
| 13 |
class BasicAgent:
|
|
|
|
| 14 |
def __init__(self):
|
| 15 |
print("BasicAgent initialized.")
|
| 16 |
self.graph = build_graph()
|
|
|
|
| 33 |
|
| 34 |
answer = answer.strip()
|
| 35 |
|
|
|
|
| 36 |
match = re.search(r'FINAL ANSWER:\s*(.*)', answer, re.IGNORECASE | re.DOTALL)
|
| 37 |
if match:
|
| 38 |
final_answer = match.group(1).strip()
|
|
|
|
| 39 |
if (final_answer.startswith('"') and final_answer.endswith('"')) or \
|
| 40 |
(final_answer.startswith("'") and final_answer.endswith("'")):
|
| 41 |
final_answer = final_answer[1:-1].strip()
|
| 42 |
answer = final_answer
|
| 43 |
else:
|
| 44 |
+
print("Warning: 'FINAL ANSWER:' not found; submitting full answer.")
|
| 45 |
|
| 46 |
if not answer:
|
| 47 |
answer = "I am unable to determine the information using the available tools."
|
| 48 |
|
| 49 |
return answer
|
| 50 |
|
|
|
|
| 51 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 52 |
space_id = os.getenv("SPACE_ID")
|
| 53 |
|
|
|
|
| 96 |
continue
|
| 97 |
try:
|
| 98 |
model_answer = agent(question_text, task_id=task_id)
|
| 99 |
+
model_answer = model_answer.strip()
|
| 100 |
+
# Clean trailing quotes/brackets if any
|
| 101 |
+
if model_answer.endswith("]'") or model_answer.endswith(']"'):
|
| 102 |
+
model_answer = model_answer[:-2].strip()
|
| 103 |
+
if (model_answer.startswith('"') and model_answer.endswith('"')) or \
|
| 104 |
+
(model_answer.startswith("'") and model_answer.endswith("'")):
|
| 105 |
+
model_answer = model_answer[1:-1].strip()
|
| 106 |
+
|
| 107 |
+
# Placeholder reasoning trace; replace with actual reasoning if available
|
| 108 |
+
reasoning_trace = f"Reasoning steps for task {task_id}."
|
| 109 |
+
|
| 110 |
print(f"Answer for task {task_id}: '{model_answer}'")
|
| 111 |
answers_payload.append({
|
| 112 |
"task_id": task_id,
|
| 113 |
"model_answer": model_answer,
|
| 114 |
+
"reasoning_trace": reasoning_trace
|
| 115 |
})
|
| 116 |
results_log.append({
|
| 117 |
"Task ID": task_id,
|
|
|
|
| 123 |
answers_payload.append({
|
| 124 |
"task_id": task_id,
|
| 125 |
"model_answer": f"AGENT ERROR: {e}",
|
| 126 |
+
"reasoning_trace": ""
|
| 127 |
})
|
| 128 |
results_log.append({
|
| 129 |
"Task ID": task_id,
|
|
|
|
| 135 |
print("Agent did not produce any answers to submit.")
|
| 136 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
| 137 |
|
| 138 |
+
# Serialize to JSON Lines string
|
| 139 |
json_lines_str = "\n".join(json.dumps(ans, ensure_ascii=False) for ans in answers_payload)
|
| 140 |
file_like = io.BytesIO(json_lines_str.encode("utf-8"))
|
| 141 |
file_like.name = "submission.jsonl"
|
|
|
|
| 168 |
print("Submission successful.")
|
| 169 |
results_df = pd.DataFrame(results_log)
|
| 170 |
return final_status, results_df
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 171 |
except Exception as e:
|
| 172 |
+
status_message = f"Submission Failed: {e}"
|
| 173 |
print(status_message)
|
| 174 |
results_df = pd.DataFrame(results_log)
|
| 175 |
return status_message, results_df
|
| 176 |
|
| 177 |
|
|
|
|
| 178 |
with gr.Blocks() as demo:
|
| 179 |
gr.Markdown("# Basic Agent Evaluation Runner")
|
| 180 |
gr.Markdown(
|
|
|
|
| 209 |
outputs=[status_output, results_table]
|
| 210 |
)
|
| 211 |
|
| 212 |
+
|
| 213 |
if __name__ == "__main__":
|
| 214 |
print("\n" + "-"*30 + " App Starting " + "-"*30)
|
| 215 |
space_host_startup = os.getenv("SPACE_HOST")
|