Spaces:
Sleeping
Sleeping
initial commit 7/20
Browse files- .gitignore +21 -0
- agent.py +208 -0
- app.py +50 -29
- app_dev.py +133 -0
- fetch.py +139 -0
- pyproject.toml +20 -0
- requirements.txt +0 -0
- uv.lock +0 -0
.gitignore
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
*.csv
|
2 |
+
*.xlsx
|
3 |
+
*.py
|
4 |
+
*.png
|
5 |
+
*.jpg
|
6 |
+
*.jpeg
|
7 |
+
*.gif
|
8 |
+
*.mp3
|
9 |
+
*.wav
|
10 |
+
*.mp4
|
11 |
+
*.avi
|
12 |
+
*.mov
|
13 |
+
__pycache__/
|
14 |
+
*.pyc
|
15 |
+
*.pyo
|
16 |
+
*.pyd
|
17 |
+
*.db
|
18 |
+
*.env
|
19 |
+
*.envrc
|
20 |
+
*.log
|
21 |
+
.vscode/
|
agent.py
ADDED
@@ -0,0 +1,208 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import base64
|
2 |
+
import logging
|
3 |
+
import os
|
4 |
+
from io import BytesIO
|
5 |
+
from typing import Any
|
6 |
+
|
7 |
+
from smolagents import (
|
8 |
+
CodeAgent,
|
9 |
+
DuckDuckGoSearchTool,
|
10 |
+
OpenAIServerModel,
|
11 |
+
SpeechToTextTool,
|
12 |
+
VisitWebpageTool,
|
13 |
+
WikipediaSearchTool,
|
14 |
+
tool,
|
15 |
+
)
|
16 |
+
|
17 |
+
system_prompt = """You are an AI Agent that is tasked to answer questions in a concise and accurate manner.
|
18 |
+
I will ask you a question and provide you with additional context if available.
|
19 |
+
Context can be in the form of Data(data), Code(code), Audio(audio), or Images(image_url).
|
20 |
+
Context is provided by specifying the content type followed by the content itself.
|
21 |
+
|
22 |
+
For example: code: print("Hello World") or Data: [1, 2, 3, 4, 5] or audio: [base64 encoded audio] or image_url: [base64 encoded image].
|
23 |
+
|
24 |
+
YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings.
|
25 |
+
DO NOT use formatting such as bold, italics, or code blocks in your final answer.
|
26 |
+
DO NOT use sources, references, or abbreviations in your final answer.
|
27 |
+
If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise.
|
28 |
+
If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise.
|
29 |
+
If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string.
|
30 |
+
If you are asked for a specific number format, follow the instructions carefully.
|
31 |
+
If you are asked for a number only answer with the number itself, without any additional text or formatting.
|
32 |
+
If you are asked for a string only answer with the string itself, without any additional text or formatting.
|
33 |
+
If you are asked for a list only answer with the list itself, without any additional text or formatting.
|
34 |
+
|
35 |
+
Finish your Answer with the following template:
|
36 |
+
FINAL ANSWER: [YOUR FINAL ANSWER]
|
37 |
+
|
38 |
+
For example, if the question is "What is the capital of France?", you should answer:
|
39 |
+
FINAL ANSWER: Paris
|
40 |
+
|
41 |
+
If the question is "What is 2 + 2?", you should answer:
|
42 |
+
FINAL ANSWER: 4
|
43 |
+
|
44 |
+
If the question is "What is 1 divided by 2, answer with 2 digits after the decimal point?", you should answer:
|
45 |
+
FINAL ANSWER: 0.50
|
46 |
+
|
47 |
+
Think step by step, and use the tools provided to gather information if necessary.
|
48 |
+
Report your thoughts, and finish your answer with the following template: FINAL ANSWER: [YOUR FINAL ANSWER].
|
49 |
+
"""
|
50 |
+
|
51 |
+
|
52 |
+
# def is_correct_format(answer: str, _) -> bool:
|
53 |
+
# """Check if the answer contains a final answer in the correct format.
|
54 |
+
|
55 |
+
# Args:
|
56 |
+
# answer: The answer to check.
|
57 |
+
|
58 |
+
# Returns:
|
59 |
+
# True if the answer contains a final answer, False otherwise.
|
60 |
+
# This ensures the final output is in the correct format.
|
61 |
+
# """
|
62 |
+
# return (
|
63 |
+
# "ANSWER:" in answer
|
64 |
+
# or "FINAL ANSWER:" in answer
|
65 |
+
# or "Answer:" in answer
|
66 |
+
# or "Final Answer:" in answer
|
67 |
+
# or "answer:" in answer
|
68 |
+
# or "final answer:" in answer
|
69 |
+
# or "answer:" in answer.lower()
|
70 |
+
# or "final answer:" in answer.lower()
|
71 |
+
# )
|
72 |
+
|
73 |
+
|
74 |
+
@tool
|
75 |
+
def wikipedia_suggested_page(query: str) -> str:
|
76 |
+
"""Search Wikipedia for suggested pages based on the query.
|
77 |
+
|
78 |
+
Args:
|
79 |
+
query: The search query. The query should be coarse and not provide too many details.
|
80 |
+
E.g. "Python programming" or "Artificial Intelligence".
|
81 |
+
|
82 |
+
Returns:
|
83 |
+
A list of suggested page titles. Pages are \n separated.
|
84 |
+
"""
|
85 |
+
from wikipedia import suggest
|
86 |
+
|
87 |
+
try:
|
88 |
+
return suggest(query)
|
89 |
+
except Exception as e:
|
90 |
+
logging.error(f"Error fetching Wikipedia suggestions for '{query}': {e}")
|
91 |
+
return f"Error fetching suggestions: {e}"
|
92 |
+
|
93 |
+
|
94 |
+
@tool
|
95 |
+
def wikipedia_page(title: str) -> str:
|
96 |
+
"""Search Wikipedia for a page based on the title.
|
97 |
+
|
98 |
+
Args:
|
99 |
+
title: The title of the Wikipedia page to search for.
|
100 |
+
|
101 |
+
Returns:
|
102 |
+
The content of the Wikipedia page.
|
103 |
+
"""
|
104 |
+
from wikipedia import page
|
105 |
+
|
106 |
+
try:
|
107 |
+
return page(title, auto_suggest=True).content
|
108 |
+
except Exception as e:
|
109 |
+
logging.error(f"Error fetching Wikipedia page for '{title}': {e}")
|
110 |
+
return f"Error fetching page: {e}"
|
111 |
+
|
112 |
+
|
113 |
+
class BasicAgent:
|
114 |
+
def __init__(self):
|
115 |
+
model = OpenAIServerModel(
|
116 |
+
model_id="gpt-4o-mini",
|
117 |
+
api_key=os.getenv("OPENAI_API_KEY"),
|
118 |
+
)
|
119 |
+
search = DuckDuckGoSearchTool(max_results=5)
|
120 |
+
# speech_to_text = SpeechToTextTool()
|
121 |
+
visitor = VisitWebpageTool(max_output_length=4000)
|
122 |
+
wiki_search = WikipediaSearchTool()
|
123 |
+
self.agent = CodeAgent(
|
124 |
+
max_steps=5,
|
125 |
+
verbosity_level=0,
|
126 |
+
tools=[
|
127 |
+
search,
|
128 |
+
# speech_to_text,
|
129 |
+
visitor,
|
130 |
+
wiki_search,
|
131 |
+
wikipedia_suggested_page,
|
132 |
+
wikipedia_page,
|
133 |
+
],
|
134 |
+
model=model,
|
135 |
+
instructions=system_prompt,
|
136 |
+
additional_authorized_imports=["pandas", "numpy"],
|
137 |
+
use_structured_outputs_internally=True,
|
138 |
+
add_base_tools=True,
|
139 |
+
# final_answer_checks=[is_correct_format],
|
140 |
+
)
|
141 |
+
logging.info(
|
142 |
+
f"System prompt set for BasicAgent: {self.agent.memory.system_prompt}"
|
143 |
+
)
|
144 |
+
|
145 |
+
def __call__(self, question: str, content, content_type) -> Any:
|
146 |
+
match content_type:
|
147 |
+
case "xlsx":
|
148 |
+
additional_args = {"data": content}
|
149 |
+
case "py":
|
150 |
+
additional_args = {"code": content}
|
151 |
+
case "audio":
|
152 |
+
additional_args = {"audio": content}
|
153 |
+
case "png":
|
154 |
+
buffer = BytesIO()
|
155 |
+
content.save(buffer, format="PNG")
|
156 |
+
buffer.seek(0)
|
157 |
+
image_content = (
|
158 |
+
"data:image/png;base64,"
|
159 |
+
+ base64.b64encode(buffer.getvalue()).decode("utf-8")
|
160 |
+
)
|
161 |
+
additional_args = {"image_url": image_content}
|
162 |
+
case _:
|
163 |
+
additional_args = None
|
164 |
+
response = self.agent.run(
|
165 |
+
question,
|
166 |
+
additional_args=additional_args,
|
167 |
+
images=[content] if content_type == "png" else None,
|
168 |
+
reset=True,
|
169 |
+
)
|
170 |
+
return response
|
171 |
+
|
172 |
+
@staticmethod
|
173 |
+
def formatting(answer: str) -> str:
|
174 |
+
"""Extract the final answer from the response."""
|
175 |
+
if "FINAL ANSWER:" in answer:
|
176 |
+
answer = answer.split("FINAL ANSWER:")[-1].strip()
|
177 |
+
if "ANSWER:" in answer:
|
178 |
+
answer = answer.split("ANSWER:")[-1].strip()
|
179 |
+
if "Answer:" in answer:
|
180 |
+
answer = answer.split("Answer:")[-1].strip()
|
181 |
+
if "Final Answer:" in answer:
|
182 |
+
answer = answer.split("Final Answer:")[-1].strip()
|
183 |
+
if "answer:" in answer.lower():
|
184 |
+
answer = answer.split("answer:")[-1].strip()
|
185 |
+
if "final answer:" in answer.lower():
|
186 |
+
answer = answer.split("final answer:")[-1].strip()
|
187 |
+
if "answer is:" in answer.lower():
|
188 |
+
answer = answer.split("answer is:")[-1].strip()
|
189 |
+
if "is:" in answer.lower():
|
190 |
+
answer = answer.split("is:")[-1].strip()
|
191 |
+
if "**" in answer:
|
192 |
+
answer = answer.split("**")[-1].strip().replace("**", "")
|
193 |
+
if "```" in answer:
|
194 |
+
answer = answer.split("```")[-1].strip().replace("```", "")
|
195 |
+
if "```python" in answer:
|
196 |
+
answer = answer.split("```python")[-1].strip().replace("```", "")
|
197 |
+
if "```json" in answer:
|
198 |
+
answer = answer.split("```json")[-1].strip().replace("```", "")
|
199 |
+
if "```yaml" in answer:
|
200 |
+
answer = answer.split("```yaml")[-1].strip().replace("```", "")
|
201 |
+
if "```txt" in answer:
|
202 |
+
answer = answer.split("```txt")[-1].strip().replace("```", "")
|
203 |
+
answer = answer.capitalize()
|
204 |
+
|
205 |
+
answer = answer.replace('"', '').strip()
|
206 |
+
answer = answer.replace("'", "").strip()
|
207 |
+
answer = answer.replace("[", "").replace("]", "").strip()
|
208 |
+
return answer.strip() # Fallback to return the whole answer if no specific format found
|
app.py
CHANGED
@@ -1,34 +1,38 @@
|
|
|
|
1 |
import os
|
|
|
2 |
import gradio as gr
|
3 |
-
import requests
|
4 |
-
import inspect
|
5 |
import pandas as pd
|
|
|
6 |
|
7 |
# (Keep Constants as is)
|
8 |
# --- Constants ---
|
9 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
10 |
|
|
|
11 |
# --- Basic Agent Definition ---
|
12 |
# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
|
13 |
class BasicAgent:
|
14 |
def __init__(self):
|
15 |
print("BasicAgent initialized.")
|
|
|
16 |
def __call__(self, question: str) -> str:
|
17 |
print(f"Agent received question (first 50 chars): {question[:50]}...")
|
18 |
fixed_answer = "This is a default answer."
|
19 |
print(f"Agent returning fixed answer: {fixed_answer}")
|
20 |
return fixed_answer
|
21 |
|
22 |
-
|
|
|
23 |
"""
|
24 |
Fetches all questions, runs the BasicAgent on them, submits all answers,
|
25 |
and displays the results.
|
26 |
"""
|
27 |
# --- Determine HF Space Runtime URL and Repo URL ---
|
28 |
-
space_id = os.getenv("SPACE_ID")
|
29 |
|
30 |
if profile:
|
31 |
-
username= f"{profile.username}"
|
32 |
print(f"User logged in: {username}")
|
33 |
else:
|
34 |
print("User not logged in.")
|
@@ -55,16 +59,12 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
55 |
response.raise_for_status()
|
56 |
questions_data = response.json()
|
57 |
if not questions_data:
|
58 |
-
|
59 |
-
|
60 |
print(f"Fetched {len(questions_data)} questions.")
|
61 |
except requests.exceptions.RequestException as e:
|
62 |
print(f"Error fetching questions: {e}")
|
63 |
return f"Error fetching questions: {e}", None
|
64 |
-
except requests.exceptions.JSONDecodeError as e:
|
65 |
-
print(f"Error decoding JSON response from questions endpoint: {e}")
|
66 |
-
print(f"Response text: {response.text[:500]}")
|
67 |
-
return f"Error decoding server response for questions: {e}", None
|
68 |
except Exception as e:
|
69 |
print(f"An unexpected error occurred fetching questions: {e}")
|
70 |
return f"An unexpected error occurred fetching questions: {e}", None
|
@@ -81,18 +81,36 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
81 |
continue
|
82 |
try:
|
83 |
submitted_answer = agent(question_text)
|
84 |
-
answers_payload.append(
|
85 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
except Exception as e:
|
87 |
-
|
88 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
|
90 |
if not answers_payload:
|
91 |
print("Agent did not produce any answers to submit.")
|
92 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
93 |
|
94 |
-
# 4. Prepare Submission
|
95 |
-
submission_data = {
|
|
|
|
|
|
|
|
|
96 |
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
97 |
print(status_update)
|
98 |
|
@@ -162,20 +180,19 @@ with gr.Blocks() as demo:
|
|
162 |
|
163 |
run_button = gr.Button("Run Evaluation & Submit All Answers")
|
164 |
|
165 |
-
status_output = gr.Textbox(
|
|
|
|
|
166 |
# Removed max_rows=10 from DataFrame constructor
|
167 |
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
168 |
|
169 |
-
run_button.click(
|
170 |
-
fn=run_and_submit_all,
|
171 |
-
outputs=[status_output, results_table]
|
172 |
-
)
|
173 |
|
174 |
if __name__ == "__main__":
|
175 |
-
print("\n" + "-"*30 + " App Starting " + "-"*30)
|
176 |
# Check for SPACE_HOST and SPACE_ID at startup for information
|
177 |
space_host_startup = os.getenv("SPACE_HOST")
|
178 |
-
space_id_startup = os.getenv("SPACE_ID")
|
179 |
|
180 |
if space_host_startup:
|
181 |
print(f"✅ SPACE_HOST found: {space_host_startup}")
|
@@ -183,14 +200,18 @@ if __name__ == "__main__":
|
|
183 |
else:
|
184 |
print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
|
185 |
|
186 |
-
if space_id_startup:
|
187 |
print(f"✅ SPACE_ID found: {space_id_startup}")
|
188 |
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
|
189 |
-
print(
|
|
|
|
|
190 |
else:
|
191 |
-
print(
|
|
|
|
|
192 |
|
193 |
-
print("-"*(60 + len(" App Starting ")) + "\n")
|
194 |
|
195 |
print("Launching Gradio Interface for Basic Agent Evaluation...")
|
196 |
-
demo.launch(debug=True, share=False)
|
|
|
1 |
+
import inspect
|
2 |
import os
|
3 |
+
|
4 |
import gradio as gr
|
|
|
|
|
5 |
import pandas as pd
|
6 |
+
import requests
|
7 |
|
8 |
# (Keep Constants as is)
|
9 |
# --- Constants ---
|
10 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
11 |
|
12 |
+
|
13 |
# --- Basic Agent Definition ---
|
14 |
# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
|
15 |
class BasicAgent:
|
16 |
def __init__(self):
|
17 |
print("BasicAgent initialized.")
|
18 |
+
|
19 |
def __call__(self, question: str) -> str:
|
20 |
print(f"Agent received question (first 50 chars): {question[:50]}...")
|
21 |
fixed_answer = "This is a default answer."
|
22 |
print(f"Agent returning fixed answer: {fixed_answer}")
|
23 |
return fixed_answer
|
24 |
|
25 |
+
|
26 |
+
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
27 |
"""
|
28 |
Fetches all questions, runs the BasicAgent on them, submits all answers,
|
29 |
and displays the results.
|
30 |
"""
|
31 |
# --- Determine HF Space Runtime URL and Repo URL ---
|
32 |
+
space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
|
33 |
|
34 |
if profile:
|
35 |
+
username = f"{profile.username}"
|
36 |
print(f"User logged in: {username}")
|
37 |
else:
|
38 |
print("User not logged in.")
|
|
|
59 |
response.raise_for_status()
|
60 |
questions_data = response.json()
|
61 |
if not questions_data:
|
62 |
+
print("Fetched questions list is empty.")
|
63 |
+
return "Fetched questions list is empty or invalid format.", None
|
64 |
print(f"Fetched {len(questions_data)} questions.")
|
65 |
except requests.exceptions.RequestException as e:
|
66 |
print(f"Error fetching questions: {e}")
|
67 |
return f"Error fetching questions: {e}", None
|
|
|
|
|
|
|
|
|
68 |
except Exception as e:
|
69 |
print(f"An unexpected error occurred fetching questions: {e}")
|
70 |
return f"An unexpected error occurred fetching questions: {e}", None
|
|
|
81 |
continue
|
82 |
try:
|
83 |
submitted_answer = agent(question_text)
|
84 |
+
answers_payload.append(
|
85 |
+
{"task_id": task_id, "submitted_answer": submitted_answer}
|
86 |
+
)
|
87 |
+
results_log.append(
|
88 |
+
{
|
89 |
+
"Task ID": task_id,
|
90 |
+
"Question": question_text,
|
91 |
+
"Submitted Answer": submitted_answer,
|
92 |
+
}
|
93 |
+
)
|
94 |
except Exception as e:
|
95 |
+
print(f"Error running agent on task {task_id}: {e}")
|
96 |
+
results_log.append(
|
97 |
+
{
|
98 |
+
"Task ID": task_id,
|
99 |
+
"Question": question_text,
|
100 |
+
"Submitted Answer": f"AGENT ERROR: {e}",
|
101 |
+
}
|
102 |
+
)
|
103 |
|
104 |
if not answers_payload:
|
105 |
print("Agent did not produce any answers to submit.")
|
106 |
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
107 |
|
108 |
+
# 4. Prepare Submission
|
109 |
+
submission_data = {
|
110 |
+
"username": username.strip(),
|
111 |
+
"agent_code": agent_code,
|
112 |
+
"answers": answers_payload,
|
113 |
+
}
|
114 |
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
115 |
print(status_update)
|
116 |
|
|
|
180 |
|
181 |
run_button = gr.Button("Run Evaluation & Submit All Answers")
|
182 |
|
183 |
+
status_output = gr.Textbox(
|
184 |
+
label="Run Status / Submission Result", lines=5, interactive=False
|
185 |
+
)
|
186 |
# Removed max_rows=10 from DataFrame constructor
|
187 |
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
188 |
|
189 |
+
run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
|
|
|
|
|
|
|
190 |
|
191 |
if __name__ == "__main__":
|
192 |
+
print("\n" + "-" * 30 + " App Starting " + "-" * 30)
|
193 |
# Check for SPACE_HOST and SPACE_ID at startup for information
|
194 |
space_host_startup = os.getenv("SPACE_HOST")
|
195 |
+
space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
|
196 |
|
197 |
if space_host_startup:
|
198 |
print(f"✅ SPACE_HOST found: {space_host_startup}")
|
|
|
200 |
else:
|
201 |
print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
|
202 |
|
203 |
+
if space_id_startup: # Print repo URLs if SPACE_ID is found
|
204 |
print(f"✅ SPACE_ID found: {space_id_startup}")
|
205 |
print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
|
206 |
+
print(
|
207 |
+
f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main"
|
208 |
+
)
|
209 |
else:
|
210 |
+
print(
|
211 |
+
"ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined."
|
212 |
+
)
|
213 |
|
214 |
+
print("-" * (60 + len(" App Starting ")) + "\n")
|
215 |
|
216 |
print("Launching Gradio Interface for Basic Agent Evaluation...")
|
217 |
+
demo.launch(debug=True, share=False)
|
app_dev.py
ADDED
@@ -0,0 +1,133 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
|
3 |
+
import gradio as gr
|
4 |
+
import pandas as pd
|
5 |
+
import requests
|
6 |
+
from langfuse import get_client
|
7 |
+
from openinference.instrumentation.smolagents import SmolagentsInstrumentor
|
8 |
+
|
9 |
+
from agent import BasicAgent
|
10 |
+
from fetch import DEFAULT_API_URL, fetch_questions, run_agent
|
11 |
+
|
12 |
+
submit_url = f"{DEFAULT_API_URL}/submit"
|
13 |
+
|
14 |
+
langfuse = get_client()
|
15 |
+
|
16 |
+
# Verify connection
|
17 |
+
if langfuse.auth_check():
|
18 |
+
print("Langfuse client is authenticated and ready!")
|
19 |
+
else:
|
20 |
+
print("Authentication failed. Please check your credentials and host.")
|
21 |
+
|
22 |
+
SmolagentsInstrumentor().instrument()
|
23 |
+
def run_and_submit_all(profile: gr.OAuthProfile | None):
|
24 |
+
"""
|
25 |
+
Fetches all questions, runs the BasicAgent on them, submits all answers,
|
26 |
+
and displays the results.
|
27 |
+
"""
|
28 |
+
# --- Determine HF Space Runtime URL and Repo URL ---
|
29 |
+
space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
|
30 |
+
|
31 |
+
if profile:
|
32 |
+
username = f"{profile.username}"
|
33 |
+
print(f"User logged in: {username}")
|
34 |
+
else:
|
35 |
+
print("User not logged in.")
|
36 |
+
return "Please Login to Hugging Face with the button.", None
|
37 |
+
# 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)
|
38 |
+
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
|
39 |
+
print(agent_code)
|
40 |
+
agent = BasicAgent()
|
41 |
+
questions_data = fetch_questions()
|
42 |
+
answers_payload, results_log = run_agent(agent, questions_data)
|
43 |
+
# 4. Prepare Submission
|
44 |
+
submission_data = {
|
45 |
+
"username": username.strip(),
|
46 |
+
"agent_code": agent_code,
|
47 |
+
"answers": answers_payload,
|
48 |
+
}
|
49 |
+
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
|
50 |
+
print(status_update)
|
51 |
+
|
52 |
+
# 5. Submit
|
53 |
+
print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
|
54 |
+
try:
|
55 |
+
response = requests.post(submit_url, json=submission_data, timeout=60)
|
56 |
+
response.raise_for_status()
|
57 |
+
result_data = response.json()
|
58 |
+
final_status = (
|
59 |
+
f"Submission Successful!\n"
|
60 |
+
f"User: {result_data.get('username')}\n"
|
61 |
+
f"Overall Score: {result_data.get('score', 'N/A')}% "
|
62 |
+
f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
|
63 |
+
f"Message: {result_data.get('message', 'No message received.')}"
|
64 |
+
)
|
65 |
+
print("Submission successful.")
|
66 |
+
results_df = pd.DataFrame(results_log)
|
67 |
+
return final_status, results_df
|
68 |
+
except requests.exceptions.HTTPError as e:
|
69 |
+
error_detail = f"Server responded with status {e.response.status_code}."
|
70 |
+
try:
|
71 |
+
error_json = e.response.json()
|
72 |
+
error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
|
73 |
+
except requests.exceptions.JSONDecodeError:
|
74 |
+
error_detail += f" Response: {e.response.text[:500]}"
|
75 |
+
status_message = f"Submission Failed: {error_detail}"
|
76 |
+
print(status_message)
|
77 |
+
results_df = pd.DataFrame(results_log)
|
78 |
+
return status_message, results_df
|
79 |
+
except requests.exceptions.Timeout:
|
80 |
+
status_message = "Submission Failed: The request timed out."
|
81 |
+
print(status_message)
|
82 |
+
results_df = pd.DataFrame(results_log)
|
83 |
+
return status_message, results_df
|
84 |
+
except requests.exceptions.RequestException as e:
|
85 |
+
status_message = f"Submission Failed: Network error - {e}"
|
86 |
+
print(status_message)
|
87 |
+
results_df = pd.DataFrame(results_log)
|
88 |
+
return status_message, results_df
|
89 |
+
except Exception as e:
|
90 |
+
status_message = f"An unexpected error occurred during submission: {e}"
|
91 |
+
print(status_message)
|
92 |
+
results_df = pd.DataFrame(results_log)
|
93 |
+
return status_message, results_df
|
94 |
+
|
95 |
+
|
96 |
+
# --- Build Gradio Interface using Blocks ---
|
97 |
+
with gr.Blocks() as demo:
|
98 |
+
gr.Markdown("# Basic Agent Evaluation Runner")
|
99 |
+
gr.Markdown(
|
100 |
+
"""
|
101 |
+
**Instructions:**
|
102 |
+
|
103 |
+
1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
|
104 |
+
2. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
|
105 |
+
3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
|
106 |
+
|
107 |
+
---
|
108 |
+
**Disclaimers:**
|
109 |
+
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).
|
110 |
+
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.
|
111 |
+
"""
|
112 |
+
)
|
113 |
+
|
114 |
+
gr.LoginButton()
|
115 |
+
|
116 |
+
run_button = gr.Button("Run Evaluation & Submit All Answers")
|
117 |
+
|
118 |
+
status_output = gr.Textbox(
|
119 |
+
label="Run Status / Submission Result", lines=5, interactive=False
|
120 |
+
)
|
121 |
+
# Removed max_rows=10 from DataFrame constructor
|
122 |
+
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
|
123 |
+
|
124 |
+
run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
|
125 |
+
|
126 |
+
|
127 |
+
if __name__ == "__main__":
|
128 |
+
print("\n" + "-" * 30 + " App Starting " + "-" * 30)
|
129 |
+
# Check for SPACE_HOST and SPACE_ID at startup for information
|
130 |
+
space_host_startup = os.getenv("SPACE_HOST")
|
131 |
+
space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
|
132 |
+
|
133 |
+
demo.launch(debug=True, share=False)
|
fetch.py
ADDED
@@ -0,0 +1,139 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import logging
|
2 |
+
from io import BytesIO
|
3 |
+
|
4 |
+
import pandas as pd
|
5 |
+
import requests
|
6 |
+
from PIL import Image
|
7 |
+
|
8 |
+
from agent import BasicAgent
|
9 |
+
|
10 |
+
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
11 |
+
|
12 |
+
questions_url = f"{DEFAULT_API_URL}/questions"
|
13 |
+
files_url = f"{DEFAULT_API_URL}/files"
|
14 |
+
|
15 |
+
|
16 |
+
def fetch_questions():
|
17 |
+
"""
|
18 |
+
Fetches questions from the API.
|
19 |
+
Returns a list of questions or an error message.
|
20 |
+
"""
|
21 |
+
try:
|
22 |
+
response = requests.get(questions_url, timeout=15)
|
23 |
+
response.raise_for_status()
|
24 |
+
questions_data = response.json()
|
25 |
+
if not questions_data:
|
26 |
+
logging.warning("Fetched questions list is empty.")
|
27 |
+
return None
|
28 |
+
logging.info(f"Fetched {len(questions_data)} questions.")
|
29 |
+
|
30 |
+
for question in questions_data:
|
31 |
+
content, content_type = _load_files(question)
|
32 |
+
if content is not None:
|
33 |
+
question["file_content"] = content
|
34 |
+
question["file_type"] = content_type
|
35 |
+
return questions_data
|
36 |
+
except requests.exceptions.RequestException as e:
|
37 |
+
logging.error(f"Error fetching questions: {e}")
|
38 |
+
return None
|
39 |
+
except Exception as e:
|
40 |
+
logging.error(f"An unexpected error occurred fetching questions: {e}")
|
41 |
+
return None
|
42 |
+
return questions_data
|
43 |
+
|
44 |
+
|
45 |
+
def _load_files(question_data: dict):
|
46 |
+
if file_name := question_data.get("file_name"):
|
47 |
+
extension = file_name.split(".")[-1]
|
48 |
+
if extension not in ["xlsx", "png", "py", "mp3", "wav"]:
|
49 |
+
logging.warning(
|
50 |
+
f"File {file_name} has an unsupported extension. Skipping file loading."
|
51 |
+
)
|
52 |
+
return None, None # Ensure a tuple is always returned
|
53 |
+
if task_id := question_data.get("task_id"):
|
54 |
+
try:
|
55 |
+
if extension == "mp3" or extension == "wav":
|
56 |
+
return f"{files_url}/{task_id}", "audio"
|
57 |
+
response = requests.get(f"{files_url}/{task_id}", timeout=15)
|
58 |
+
response.raise_for_status()
|
59 |
+
if response.status_code == 200:
|
60 |
+
# extensions: xlsx, png, py, else ignore
|
61 |
+
match extension:
|
62 |
+
case "xlsx":
|
63 |
+
if (
|
64 |
+
response.headers.get("Content-Type")
|
65 |
+
== "application/octet-stream"
|
66 |
+
):
|
67 |
+
logging.info(f"Processing Excel file: {file_name}")
|
68 |
+
return pd.read_excel(response.content).to_json(), "xlsx"
|
69 |
+
case "png":
|
70 |
+
if response.headers.get("Content-Type") == "image/png":
|
71 |
+
logging.info(f"Processing image file: {file_name}")
|
72 |
+
return Image.open(BytesIO(response.content)).convert("RGB"), "png"
|
73 |
+
case "py":
|
74 |
+
if response.headers.get("Content-Type", "").startswith(
|
75 |
+
"text/x-python"
|
76 |
+
):
|
77 |
+
logging.info(f"Processing Python file: {file_name}")
|
78 |
+
return response.content.decode(
|
79 |
+
"utf-8"
|
80 |
+
), "py" # Load Python file if needed
|
81 |
+
except requests.exceptions.RequestException as e:
|
82 |
+
logging.error(f"Error fetching file for task {task_id}: {e}")
|
83 |
+
raise e
|
84 |
+
except Exception as e:
|
85 |
+
logging.error(
|
86 |
+
f"An unexpected error occurred fetching file for task {task_id}: {e}"
|
87 |
+
)
|
88 |
+
raise e
|
89 |
+
return None, None
|
90 |
+
return None, None # Always return a tuple
|
91 |
+
|
92 |
+
|
93 |
+
def run_agent(agent, questions_data):
|
94 |
+
results_log = []
|
95 |
+
answers_payload = []
|
96 |
+
logging.info(f"Running agent on {len(questions_data)} questions...")
|
97 |
+
for item in questions_data:
|
98 |
+
payload, log_item = run_agent_on_question(agent, item)
|
99 |
+
if payload is not None:
|
100 |
+
answers_payload.append(payload)
|
101 |
+
if log_item is not None:
|
102 |
+
results_log.append(log_item)
|
103 |
+
if not answers_payload:
|
104 |
+
logging.info("Agent did not produce any answers to submit.")
|
105 |
+
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
|
106 |
+
return answers_payload, pd.DataFrame(results_log)
|
107 |
+
|
108 |
+
|
109 |
+
def run_agent_on_question(agent: BasicAgent, question):
|
110 |
+
"""
|
111 |
+
Runs the agent on a single question and returns the answer.
|
112 |
+
"""
|
113 |
+
task_id = question.get("task_id")
|
114 |
+
question_text = question.get("question")
|
115 |
+
content = question.get("file_content")
|
116 |
+
content_type = question.get("file_type")
|
117 |
+
if not task_id or question_text is None:
|
118 |
+
logging.warning(f"Skipping item with missing task_id or question: {question}")
|
119 |
+
return None, None
|
120 |
+
try:
|
121 |
+
submitted_answer = agent(question_text, content=content, content_type=content_type)
|
122 |
+
return (
|
123 |
+
{"task_id": task_id, "submitted_answer": submitted_answer},
|
124 |
+
{
|
125 |
+
"Task ID": task_id,
|
126 |
+
"Question": question_text,
|
127 |
+
"Submitted Answer": submitted_answer,
|
128 |
+
},
|
129 |
+
)
|
130 |
+
except Exception as e:
|
131 |
+
logging.error(f"Error running agent on task {task_id}: {e}")
|
132 |
+
return (
|
133 |
+
{
|
134 |
+
"Task ID": task_id,
|
135 |
+
"Question": question_text,
|
136 |
+
"Submitted Answer": f"AGENT ERROR: {e}",
|
137 |
+
},
|
138 |
+
None,
|
139 |
+
)
|
pyproject.toml
ADDED
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[project]
|
2 |
+
name = "final-assignment-template"
|
3 |
+
version = "0.1.0"
|
4 |
+
description = "Add your description here"
|
5 |
+
readme = "README.md"
|
6 |
+
requires-python = ">=3.12"
|
7 |
+
dependencies = [
|
8 |
+
"chess>=1.11.2",
|
9 |
+
"google-genai>=1.23.0",
|
10 |
+
"gradio[oauth]>=5.34.1",
|
11 |
+
"langfuse>=3.0.7",
|
12 |
+
"openpyxl>=3.1.5",
|
13 |
+
"pandas>=2.3.0",
|
14 |
+
"requests>=2.32.4",
|
15 |
+
"ruff>=0.12.0",
|
16 |
+
"smolagents[audio,litellm,openai,telemetry,toolkit,transformers]>=1.18.0",
|
17 |
+
"transformers>=4.53.0",
|
18 |
+
"wikipedia>=1.4.0",
|
19 |
+
"wikipedia-api>=0.8.1",
|
20 |
+
]
|
requirements.txt
CHANGED
The diff for this file is too large to render.
See raw diff
|
|
uv.lock
ADDED
The diff for this file is too large to render.
See raw diff
|
|