Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
|
|
| 1 |
import os
|
| 2 |
import gradio as gr
|
| 3 |
import requests
|
|
@@ -12,6 +13,7 @@ load_dotenv()
|
|
| 12 |
# --- Constants ---
|
| 13 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 14 |
|
|
|
|
| 15 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 16 |
"""
|
| 17 |
Fetches all questions, runs the LangGraph Agent on them, submits all answers,
|
|
@@ -19,6 +21,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 19 |
"""
|
| 20 |
# --- Determine HF Space Runtime URL and Repo URL ---
|
| 21 |
space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
|
|
|
|
| 22 |
if profile:
|
| 23 |
username = f"{profile.username}"
|
| 24 |
print(f"User logged in: {username}")
|
|
@@ -38,8 +41,8 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 38 |
except Exception as e:
|
| 39 |
print(f"Error instantiating agent graph: {e}")
|
| 40 |
return f"Error initializing agent graph: {e}", None
|
| 41 |
-
|
| 42 |
-
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else "Agent code link unavailable (SPACE_ID not set)"
|
| 43 |
print(f"Agent code link: {agent_code}")
|
| 44 |
|
| 45 |
# 2. Fetch Questions
|
|
@@ -68,33 +71,56 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 68 |
answers_payload = []
|
| 69 |
print(f"Running agent on {len(questions_data)} questions...")
|
| 70 |
|
|
|
|
|
|
|
| 71 |
for item in questions_data:
|
| 72 |
task_id = item.get("task_id")
|
| 73 |
question_text = item.get("question")
|
|
|
|
| 74 |
if not task_id or question_text is None:
|
| 75 |
print(f"Skipping item with missing task_id or question: {item}")
|
| 76 |
continue
|
| 77 |
|
|
|
|
| 78 |
print(f"--- Starting processing Task ID: {task_id}, Question: {question_text[:100]}...")
|
|
|
|
| 79 |
try:
|
|
|
|
| 80 |
result_state = agent_graph.invoke({"messages": [HumanMessage(content=question_text)]})
|
| 81 |
-
|
|
|
|
|
|
|
|
|
|
| 82 |
if result_state and "messages" in result_state and result_state["messages"]:
|
| 83 |
last_message = result_state["messages"][-1]
|
|
|
|
| 84 |
if hasattr(last_message, 'content') and last_message.content:
|
| 85 |
submitted_answer = last_message.content
|
|
|
|
|
|
|
|
|
|
|
|
|
| 86 |
if not isinstance(submitted_answer, str):
|
| 87 |
submitted_answer = str(submitted_answer)
|
|
|
|
| 88 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 89 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
|
|
|
|
|
|
| 90 |
print(f"--- Finished processing Task ID: {task_id}")
|
|
|
|
| 91 |
print(f"--- Extracted answer for Task ID: {task_id}: {submitted_answer[:100]}...")
|
|
|
|
|
|
|
| 92 |
except Exception as e:
|
| 93 |
print(f"Error running agent graph on task {task_id}: {e}")
|
| 94 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
|
|
|
|
|
|
|
| 95 |
|
| 96 |
if not answers_payload:
|
| 97 |
print("Agent did not produce any answers to submit.")
|
|
|
|
| 98 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
| 99 |
|
| 100 |
# 4. Prepare Submission
|
|
@@ -118,6 +144,22 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 118 |
print("Submission successful.")
|
| 119 |
results_df = pd.DataFrame(results_log)
|
| 120 |
return final_status, results_df
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 121 |
except requests.exceptions.RequestException as e:
|
| 122 |
status_message = f"Submission Failed: Network error - {e}"
|
| 123 |
print(status_message)
|
|
@@ -129,27 +171,58 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
| 129 |
results_df = pd.DataFrame(results_log)
|
| 130 |
return status_message, results_df
|
| 131 |
|
|
|
|
| 132 |
# --- Build Gradio Interface using Blocks ---
|
| 133 |
with gr.Blocks() as demo:
|
| 134 |
-
gr.Markdown("# LangGraph Agent Evaluation Runner")
|
| 135 |
gr.Markdown(
|
| 136 |
"""
|
| 137 |
**Instructions:**
|
| 138 |
|
| 139 |
-
1.
|
| 140 |
-
2.
|
| 141 |
-
3.
|
| 142 |
-
4.
|
| 143 |
|
| 144 |
-
|
|
|
|
|
|
|
|
|
|
| 145 |
"""
|
| 146 |
)
|
|
|
|
| 147 |
gr.LoginButton()
|
|
|
|
| 148 |
run_button = gr.Button("Run Evaluation & Submit All Answers")
|
|
|
|
| 149 |
status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
|
| 150 |
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
| 151 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 152 |
|
| 153 |
if __name__ == "__main__":
|
| 154 |
-
print("
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 155 |
demo.launch(debug=True, share=False, auth=None)
|
|
|
|
| 1 |
+
|
| 2 |
import os
|
| 3 |
import gradio as gr
|
| 4 |
import requests
|
|
|
|
| 13 |
# --- Constants ---
|
| 14 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
| 15 |
|
| 16 |
+
|
| 17 |
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
| 18 |
"""
|
| 19 |
Fetches all questions, runs the LangGraph Agent on them, submits all answers,
|
|
|
|
| 21 |
"""
|
| 22 |
# --- Determine HF Space Runtime URL and Repo URL ---
|
| 23 |
space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
|
| 24 |
+
|
| 25 |
if profile:
|
| 26 |
username = f"{profile.username}"
|
| 27 |
print(f"User logged in: {username}")
|
|
|
|
| 41 |
except Exception as e:
|
| 42 |
print(f"Error instantiating agent graph: {e}")
|
| 43 |
return f"Error initializing agent graph: {e}", None
|
| 44 |
+
# 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)
|
| 45 |
+
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else "Agent code link unavailable (SPACE_ID not set)" # Added a check for SPACE_ID
|
| 46 |
print(f"Agent code link: {agent_code}")
|
| 47 |
|
| 48 |
# 2. Fetch Questions
|
|
|
|
| 71 |
answers_payload = []
|
| 72 |
print(f"Running agent on {len(questions_data)} questions...")
|
| 73 |
|
| 74 |
+
# Removed the problematic print statement from here
|
| 75 |
+
|
| 76 |
for item in questions_data:
|
| 77 |
task_id = item.get("task_id")
|
| 78 |
question_text = item.get("question")
|
| 79 |
+
|
| 80 |
if not task_id or question_text is None:
|
| 81 |
print(f"Skipping item with missing task_id or question: {item}")
|
| 82 |
continue
|
| 83 |
|
| 84 |
+
# Moved the print statement inside the loop, after task_id and question_text are assigned
|
| 85 |
print(f"--- Starting processing Task ID: {task_id}, Question: {question_text[:100]}...")
|
| 86 |
+
|
| 87 |
try:
|
| 88 |
+
# Invoke the LangGraph agent
|
| 89 |
result_state = agent_graph.invoke({"messages": [HumanMessage(content=question_text)]})
|
| 90 |
+
|
| 91 |
+
# Extract the final answer from the last message
|
| 92 |
+
submitted_answer = "Error: Agent did not provide a response." # Default in case extraction fails
|
| 93 |
+
|
| 94 |
if result_state and "messages" in result_state and result_state["messages"]:
|
| 95 |
last_message = result_state["messages"][-1]
|
| 96 |
+
# The final content is typically in the content attribute of the last message
|
| 97 |
if hasattr(last_message, 'content') and last_message.content:
|
| 98 |
submitted_answer = last_message.content
|
| 99 |
+
# else: Handle cases where the last message might be a tool message etc.,
|
| 100 |
+
# for simplicity, we just use the default error message if content is missing.
|
| 101 |
+
|
| 102 |
+
# Ensure submitted_answer is a string
|
| 103 |
if not isinstance(submitted_answer, str):
|
| 104 |
submitted_answer = str(submitted_answer)
|
| 105 |
+
|
| 106 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
| 107 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
| 108 |
+
|
| 109 |
+
# Moved this print statement inside the loop as well
|
| 110 |
print(f"--- Finished processing Task ID: {task_id}")
|
| 111 |
+
# Moved this print statement inside the loop as well
|
| 112 |
print(f"--- Extracted answer for Task ID: {task_id}: {submitted_answer[:100]}...")
|
| 113 |
+
|
| 114 |
+
|
| 115 |
except Exception as e:
|
| 116 |
print(f"Error running agent graph on task {task_id}: {e}")
|
| 117 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
|
| 118 |
+
# Note: If an error occurs, the 'Finished' and 'Extracted answer' prints for this specific task won't happen,
|
| 119 |
+
# which is reasonable behavior.
|
| 120 |
|
| 121 |
if not answers_payload:
|
| 122 |
print("Agent did not produce any answers to submit.")
|
| 123 |
+
# Even if no answers, show the log of errors
|
| 124 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
| 125 |
|
| 126 |
# 4. Prepare Submission
|
|
|
|
| 144 |
print("Submission successful.")
|
| 145 |
results_df = pd.DataFrame(results_log)
|
| 146 |
return final_status, results_df
|
| 147 |
+
except requests.exceptions.HTTPError as e:
|
| 148 |
+
error_detail = f"Server responded with status {e.response.status_code}."
|
| 149 |
+
try:
|
| 150 |
+
error_json = e.response.json()
|
| 151 |
+
error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
|
| 152 |
+
except requests.exceptions.JSONDecodeError:
|
| 153 |
+
error_detail += f" Response: {e.response.text[:500]}"
|
| 154 |
+
status_message = f"Submission Failed: {error_detail}"
|
| 155 |
+
print(status_message)
|
| 156 |
+
results_df = pd.DataFrame(results_log)
|
| 157 |
+
return status_message, results_df
|
| 158 |
+
except requests.exceptions.Timeout:
|
| 159 |
+
status_message = "Submission Failed: The request timed out."
|
| 160 |
+
print(status_message)
|
| 161 |
+
results_df = pd.DataFrame(results_log)
|
| 162 |
+
return status_message, results_df
|
| 163 |
except requests.exceptions.RequestException as e:
|
| 164 |
status_message = f"Submission Failed: Network error - {e}"
|
| 165 |
print(status_message)
|
|
|
|
| 171 |
results_df = pd.DataFrame(results_log)
|
| 172 |
return status_message, results_df
|
| 173 |
|
| 174 |
+
|
| 175 |
# --- Build Gradio Interface using Blocks ---
|
| 176 |
with gr.Blocks() as demo:
|
| 177 |
+
gr.Markdown("# LangGraph Agent Evaluation Runner") # Updated title
|
| 178 |
gr.Markdown(
|
| 179 |
"""
|
| 180 |
**Instructions:**
|
| 181 |
|
| 182 |
+
1. Please clone this space, then modify the code in `agent.py` and `app.py` to define your agent's logic, the tools, the necessary packages, etc ...
|
| 183 |
+
2. **Make sure you have your `DEEPSEEK_API_KEY` set as a Space Secret.**
|
| 184 |
+
3. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
|
| 185 |
+
4. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
|
| 186 |
|
| 187 |
+
---
|
| 188 |
+
**Disclaimers:**
|
| 189 |
+
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).
|
| 190 |
+
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.
|
| 191 |
"""
|
| 192 |
)
|
| 193 |
+
|
| 194 |
gr.LoginButton()
|
| 195 |
+
|
| 196 |
run_button = gr.Button("Run Evaluation & Submit All Answers")
|
| 197 |
+
|
| 198 |
status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
|
| 199 |
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
| 200 |
+
|
| 201 |
+
run_button.click(
|
| 202 |
+
fn=run_and_submit_all,
|
| 203 |
+
outputs=[status_output, results_table]
|
| 204 |
+
)
|
| 205 |
|
| 206 |
if __name__ == "__main__":
|
| 207 |
+
print("\n" + "-" * 30 + " App Starting " + "-" * 30)
|
| 208 |
+
# Check for SPACE_HOST and SPACE_ID at startup for information
|
| 209 |
+
space_host_startup = os.getenv("SPACE_HOST")
|
| 210 |
+
space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
|
| 211 |
+
|
| 212 |
+
if space_host_startup:
|
| 213 |
+
print(f"✅ SPACE_HOST found: {space_host_startup}")
|
| 214 |
+
print(f" Runtime URL should be: https://{space_host_startup}.hf.space")
|
| 215 |
+
else:
|
| 216 |
+
print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
|
| 217 |
+
|
| 218 |
+
if space_id_startup: # Print repo URLs if SPACE_ID is found
|
| 219 |
+
print(f"✅ SPACE_ID found: {space_id_startup}")
|
| 220 |
+
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
|
| 221 |
+
print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
|
| 222 |
+
else:
|
| 223 |
+
print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
|
| 224 |
+
|
| 225 |
+
print("-" * (60 + len(" App Starting ")) + "\n")
|
| 226 |
+
|
| 227 |
+
print("Launching Gradio Interface for LangGraph Agent Evaluation...") # Updated message
|
| 228 |
demo.launch(debug=True, share=False, auth=None)
|