mnosouhi96 commited on
Commit
f43898e
·
1 Parent(s): a59c4db
Files changed (1) hide show
  1. app.py +139 -184
app.py CHANGED
@@ -1,211 +1,166 @@
1
- import re, io, subprocess, requests, pandas as pd, gradio as gr
2
- from smolagents import CodeAgent, InferenceClientModel, DuckDuckGoSearchTool, VisitWebpageTool, PythonInterpreterTool
 
 
3
 
4
- SPACE_ID = "marjanns/Final_Assignment_Template"
5
  DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
6
-
7
- def postprocess_exact(s):
8
- if s is None:
9
- return ""
10
- s = str(s).strip()
11
- if (s.startswith('"') and s.endswith('"')) or (s.startswith("'") and s.endswith("'")):
12
- s = s[1:-1].strip()
13
- s = re.sub(r"\s+", " ", s)
14
- s = re.sub(r"\.(\s*)$", "", s)
15
- return s
16
-
17
- def fetch_files(task_id):
18
- try:
19
- r = requests.get(f"{DEFAULT_API_URL}/files/{task_id}", timeout=30)
20
- r.raise_for_status()
21
- data = r.json()
22
- if isinstance(data, dict) and "files" in data:
23
- return data["files"]
24
- if isinstance(data, dict) and "file_url" in data:
25
- return [data]
26
- return []
27
- except Exception:
28
- return []
29
-
30
- def solve_reverse_sentence(q):
31
- if ".rewsna" in q:
32
- try:
33
- m = re.search(r'"(.*)"', q, re.S)
34
- src = m.group(1) if m else q
35
- rev = src[::-1]
36
- if "opposite of the word 'left'" in rev or 'opposite of the word "left"' in rev:
37
- return "right"
38
- except Exception:
39
- pass
40
- return None
41
-
42
- def solve_vegetables(q):
43
- if "I'm making a grocery list" in q and "alphabetize the list of vegetables" in q:
44
- m = re.search(r"list I have so far:\s*(.*?)\s*I need to make headings", q, re.I | re.S)
45
- if not m:
46
- return ""
47
- items = [x.strip().lower() for x in re.split(r",\s*", m.group(1))]
48
- botanical_fruits = {"tomato","zucchini","courgette","bell pepper","pepper","cucumber","eggplant","aubergine","green beans","beans","corn","maize","rice","plums","peanuts","acorns","whole allspice","allspice","coffee","whole bean coffee"}
49
- non_produce = {"milk","eggs","flour","oreos","whole allspice","whole bean coffee","peanuts","acorns","plums","rice"}
50
- veg = set()
51
- for it in items:
52
- if it in botanical_fruits or it in non_produce:
53
- continue
54
- if it in {"fresh basil","basil"}:
55
- veg.add("fresh basil")
56
- elif it in {"sweet potato","sweet potatoes"}:
57
- veg.add("sweet potatoes")
58
- elif it in {"broccoli","celery","lettuce"}:
59
- veg.add(it)
60
- return ", ".join(sorted(veg))
61
- return None
62
-
63
- def parse_md_table_block(q):
64
- m = re.search(r"\|\*\|.*?\n(\|[-|]+\n)?(.*?)\n\n", q, re.S | re.I)
65
- return m.group(0) if m else None
66
-
67
- def solve_non_commutative_subset(q):
68
- if "define * on the set S" in q and "not commutative" in q:
69
- block = parse_md_table_block(q)
70
- if not block:
71
- return ""
72
- lines = [ln.strip() for ln in block.strip().splitlines() if ln.strip().startswith("|")]
73
- header = [h.strip() for h in lines[0].strip("|").split("|")]
74
- elems = [e.strip() for e in header[1:]]
75
- tbl = {}
76
- for row in lines[2:]:
77
- cells = [c.strip() for c in row.strip("|").split("|")]
78
- r = cells[0]
79
- tbl[r] = {elems[i]: cells[i+1] for i in range(len(elems))}
80
- bad = set()
81
- for x in elems:
82
- for y in elems:
83
- if tbl[x][y] != tbl[y][x]:
84
- bad.add(x); bad.add(y)
85
- return ", ".join(sorted(bad))
86
- return None
87
 
88
  class BasicAgent:
89
  def __init__(self):
90
- self.model = InferenceClientModel(model_id="Qwen/Qwen2.5-7B-Instruct")
91
- self.agent = CodeAgent(
92
- model=self.model,
93
- tools=[DuckDuckGoSearchTool(), VisitWebpageTool(), PythonInterpreterTool()],
94
- add_base_tools=True,
95
- system_prompt="Answer GAIA L1 questions using tools. Output only the final answer string.",
96
- stream_outputs=False,
97
- )
98
 
99
- def solve_with_files(self, question, task_id):
100
- files = fetch_files(task_id)
101
- for f in files:
102
- url = f.get("file_url") or f.get("url") or ""
103
- name = (f.get("filename") or f.get("name") or "").lower()
104
- if not url:
105
- continue
106
- try:
107
- data = requests.get(url, timeout=60).content
108
- except Exception:
109
- continue
110
- if name.endswith((".xlsx",".xls")):
111
- try:
112
- df = pd.read_excel(io.BytesIO(data))
113
- if "Category" in df.columns:
114
- food = df[df["Category"].astype(str).str.lower().eq("food")]
115
- if "Sales" in food.columns:
116
- total = float(food["Sales"].sum())
117
- else:
118
- total = float(food.select_dtypes(include="number").sum().sum())
119
- return f"{total:.2f}"
120
- scols = df.select_dtypes(include="number")
121
- total = float(scols.sum().sum())
122
- return f"{total:.2f}"
123
- except Exception:
124
- pass
125
- if name.endswith(".py"):
126
- try:
127
- p = subprocess.run(["python","-"], input=data, capture_output=True, text=True, timeout=10)
128
- out = p.stdout.strip()
129
- if out:
130
- return postprocess_exact(out.splitlines()[-1])
131
- except Exception:
132
- pass
133
- if name.endswith((".mp3",".wav",".m4a",".flac")):
134
- return ""
135
- return None
136
-
137
- def __call__(self, question, task_id=None):
138
- s = solve_reverse_sentence(question)
139
- if s is not None:
140
- return s
141
- s = solve_vegetables(question)
142
- if s is not None:
143
- return s
144
- s = solve_non_commutative_subset(question)
145
- if s is not None:
146
- return s
147
- if task_id is not None:
148
- s = self.solve_with_files(question, task_id)
149
- if s is not None:
150
- return s
151
- out = self.agent.run("Return only the final answer string.\nQuestion: " + question)
152
- return postprocess_exact(out)
153
-
154
- def run_and_submit_all(evt=None, *, request: gr.Request):
155
- profile = getattr(request, "user", None)
156
- if not profile:
157
  return "Please Login to Hugging Face with the button.", None
158
- username = f"{profile.username}".strip()
 
 
 
 
159
  try:
160
- qresp = requests.get(f"{DEFAULT_API_URL}/questions", timeout=30)
161
- qresp.raise_for_status()
162
- questions = qresp.json()
163
- if not isinstance(questions, list) or not questions:
164
- return "Fetched questions list is empty or invalid.", None
165
  except Exception as e:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
166
  return f"Error fetching questions: {e}", None
167
- agent = BasicAgent()
168
- agent_code = f"https://huggingface.co/spaces/{SPACE_ID}/tree/main"
169
- results_log, answers_payload = [], []
170
- for item in questions:
171
- tid = item.get("task_id")
172
- q = item.get("question")
173
- if not tid or q is None:
 
 
 
 
 
 
 
 
 
174
  continue
175
  try:
176
- ans = agent(q, tid)
 
 
 
 
 
 
177
  except Exception as e:
178
- ans = f"AGENT ERROR: {e}"
179
- answers_payload.append({"task_id": tid, "submitted_answer": ans})
180
- results_log.append({"Task ID": tid, "Question": q, "Submitted Answer": ans})
 
 
 
 
181
  if not answers_payload:
182
- return "No answers produced.", pd.DataFrame(results_log)
183
- payload = {"username": username, "agent_code": agent_code, "answers": answers_payload}
 
 
 
 
 
 
184
  try:
185
- sresp = requests.post(f"{DEFAULT_API_URL}/submit", json=payload, timeout=120)
186
- sresp.raise_for_status()
187
- res = sresp.json()
188
- msg = f"Submission Successful!\nUser: {res.get('username', username)}\nOverall Score: {res.get('score','N/A')}% ({res.get('correct_count','?')}/{res.get('total_attempted','?')} correct)\nMessage: {res.get('message','') or ''}"
189
- return msg, pd.DataFrame(results_log)
 
 
 
 
 
 
 
 
190
  except requests.exceptions.HTTPError as e:
 
191
  try:
192
- detail = e.response.json().get("detail", e.response.text)
193
- except Exception:
194
- detail = e.response.text
195
- return f"Submission Failed: HTTP {e.response.status_code}. Detail: {detail[:500]}", pd.DataFrame(results_log)
 
 
 
 
196
  except requests.exceptions.Timeout:
197
- return "Submission Failed: The request timed out.", pd.DataFrame(results_log)
 
 
 
 
 
 
 
 
198
  except Exception as e:
199
- return f"Submission Failed: {e}", pd.DataFrame(results_log)
 
 
 
200
 
201
  with gr.Blocks() as demo:
202
  gr.Markdown("# Basic Agent Evaluation Runner")
203
- gr.Markdown("1. Ensure requirements are installed.\n2. Log in.\n3. Click the button to run and submit.\nScoring is EXACT MATCH: output only the final answer string.")
204
- gr.LoginButton()
 
 
 
 
 
 
 
 
 
 
 
205
  run_button = gr.Button("Run Evaluation & Submit All Answers")
206
- status_output = gr.Textbox(label="Run Status / Submission Result", lines=6, interactive=False)
207
  results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
208
- run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
 
209
 
210
  if __name__ == "__main__":
 
 
 
 
 
 
 
 
 
211
  demo.launch(debug=True, share=False)
 
1
+ import os
2
+ import gradio as gr
3
+ import requests
4
+ import pandas as pd
5
 
 
6
  DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
7
+ SPACE_ID = "marjanns/Final_Assignment_Template"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8
 
9
  class BasicAgent:
10
  def __init__(self):
11
+ print("BasicAgent initialized.")
12
+ def __call__(self, question: str) -> str:
13
+ print(f"Agent received question (first 50 chars): {question[:50]}...")
14
+ fixed_answer = "This is a default answer."
15
+ print(f"Agent returning fixed answer: {fixed_answer}")
16
+ return fixed_answer
 
 
17
 
18
+ def run_and_submit_all(profile: gr.OAuthProfile | None):
19
+ space_id = SPACE_ID
20
+ if profile:
21
+ username = f"{profile.username}"
22
+ print(f"User logged in: {username}")
23
+ else:
24
+ print("User not logged in.")
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25
  return "Please Login to Hugging Face with the button.", None
26
+
27
+ api_url = DEFAULT_API_URL
28
+ questions_url = f"{api_url}/questions"
29
+ submit_url = f"{api_url}/submit"
30
+
31
  try:
32
+ agent = BasicAgent()
 
 
 
 
33
  except Exception as e:
34
+ print(f"Error instantiating agent: {e}")
35
+ return f"Error initializing agent: {e}", None
36
+
37
+ agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
38
+ print(agent_code)
39
+
40
+ print(f"Fetching questions from: {questions_url}")
41
+ try:
42
+ response = requests.get(questions_url, timeout=15)
43
+ response.raise_for_status()
44
+ questions_data = response.json()
45
+ if not questions_data:
46
+ print("Fetched questions list is empty.")
47
+ return "Fetched questions list is empty or invalid format.", None
48
+ print(f"Fetched {len(questions_data)} questions.")
49
+ except requests.exceptions.RequestException as e:
50
+ print(f"Error fetching questions: {e}")
51
  return f"Error fetching questions: {e}", None
52
+ except requests.exceptions.JSONDecodeError as e:
53
+ print(f"Error decoding JSON response from questions endpoint: {e}")
54
+ print(f"Response text: {response.text[:500]}")
55
+ return f"Error decoding server response for questions: {e}", None
56
+ except Exception as e:
57
+ print(f"An unexpected error occurred fetching questions: {e}")
58
+ return f"An unexpected error occurred fetching questions: {e}", None
59
+
60
+ results_log = []
61
+ answers_payload = []
62
+ print(f"Running agent on {len(questions_data)} questions...")
63
+ for item in questions_data:
64
+ task_id = item.get("task_id")
65
+ question_text = item.get("question")
66
+ if not task_id or question_text is None:
67
+ print(f"Skipping item with missing task_id or question: {item}")
68
  continue
69
  try:
70
+ submitted_answer = agent(question_text)
71
+ answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
72
+ results_log.append({
73
+ "Task ID": task_id,
74
+ "Question": question_text,
75
+ "Submitted Answer": submitted_answer
76
+ })
77
  except Exception as e:
78
+ print(f"Error running agent on task {task_id}: {e}")
79
+ results_log.append({
80
+ "Task ID": task_id,
81
+ "Question": question_text,
82
+ "Submitted Answer": f"AGENT ERROR: {e}"
83
+ })
84
+
85
  if not answers_payload:
86
+ print("Agent did not produce any answers to submit.")
87
+ return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
88
+
89
+ submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
90
+ status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
91
+ print(status_update)
92
+
93
+ print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
94
  try:
95
+ response = requests.post(submit_url, json=submission_data, timeout=60)
96
+ response.raise_for_status()
97
+ result_data = response.json()
98
+ final_status = (
99
+ f"Submission Successful!\n"
100
+ f"User: {result_data.get('username')}\n"
101
+ f"Overall Score: {result_data.get('score', 'N/A')}% "
102
+ f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
103
+ f"Message: {result_data.get('message', 'No message received.')}"
104
+ )
105
+ print("Submission successful.")
106
+ results_df = pd.DataFrame(results_log)
107
+ return final_status, results_df
108
  except requests.exceptions.HTTPError as e:
109
+ error_detail = f"Server responded with status {e.response.status_code}."
110
  try:
111
+ error_json = e.response.json()
112
+ error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
113
+ except requests.exceptions.JSONDecodeError:
114
+ error_detail += f" Response: {e.response.text[:500]}"
115
+ status_message = f"Submission Failed: {error_detail}"
116
+ print(status_message)
117
+ results_df = pd.DataFrame(results_log)
118
+ return status_message, results_df
119
  except requests.exceptions.Timeout:
120
+ status_message = "Submission Failed: The request timed out."
121
+ print(status_message)
122
+ results_df = pd.DataFrame(results_log)
123
+ return status_message, results_df
124
+ except requests.exceptions.RequestException as e:
125
+ status_message = f"Submission Failed: Network error - {e}"
126
+ print(status_message)
127
+ results_df = pd.DataFrame(results_log)
128
+ return status_message, results_df
129
  except Exception as e:
130
+ status_message = f"An unexpected error occurred during submission: {e}"
131
+ print(status_message)
132
+ results_df = pd.DataFrame(results_log)
133
+ return status_message, results_df
134
 
135
  with gr.Blocks() as demo:
136
  gr.Markdown("# Basic Agent Evaluation Runner")
137
+ gr.Markdown(
138
+ """
139
+ **Instructions:**
140
+ 1) Clone this space, then modify the code to define your agent's logic.
141
+ 2) Log in to your Hugging Face account using the button below.
142
+ 3) Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
143
+ """
144
+ )
145
+
146
+ login = gr.LoginButton()
147
+ user_state = gr.State()
148
+ login.click(lambda p: p, inputs=login, outputs=user_state)
149
+
150
  run_button = gr.Button("Run Evaluation & Submit All Answers")
151
+ status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
152
  results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
153
+
154
+ run_button.click(fn=run_and_submit_all, inputs=[user_state], outputs=[status_output, results_table])
155
 
156
  if __name__ == "__main__":
157
+ print("\n" + "-"*30 + " App Starting " + "-"*30)
158
+ space_host_startup = os.getenv("SPACE_HOST")
159
+ print(f"Repo URL: https://huggingface.co/spaces/{SPACE_ID}")
160
+ print(f"Repo Tree URL: https://huggingface.co/spaces/{SPACE_ID}/tree/main")
161
+ if space_host_startup:
162
+ print(f"Runtime URL should be: https://{space_host_startup}.hf.space")
163
+ else:
164
+ print("SPACE_HOST not found (running locally?).")
165
+ print("-"*(60 + len(" App Starting ")) + "\n")
166
  demo.launch(debug=True, share=False)