dougtrajano commited on
Commit
e9ec5c5
·
verified ·
1 Parent(s): 4b43d1a

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +263 -41
app.py CHANGED
@@ -1,34 +1,217 @@
 
 
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
- def run_and_submit_all( profile: gr.OAuthProfile | None):
 
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") # Get the SPACE_ID for sending link to the code
29
 
30
  if profile:
31
- username= f"{profile.username}"
32
  print(f"User logged in: {username}")
33
  else:
34
  print("User not logged in.")
@@ -38,10 +221,9 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
38
  questions_url = f"{api_url}/questions"
39
  submit_url = f"{api_url}/submit"
40
 
41
- # 1. Instantiate Agent ( modify this part to create your agent)
42
  try:
43
  agent = BasicAgent()
44
- except Exception as e:
45
  print(f"Error instantiating agent: {e}")
46
  return f"Error initializing agent: {e}", None
47
  # 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)
@@ -55,17 +237,17 @@ 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
- print("Fetched questions list is empty.")
59
- return "Fetched questions list is empty or invalid format.", None
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
71
 
@@ -76,23 +258,62 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
76
  for item in questions_data:
77
  task_id = item.get("task_id")
78
  question_text = item.get("question")
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
79
  if not task_id or question_text is None:
80
  print(f"Skipping item with missing task_id or question: {item}")
81
  continue
82
  try:
83
- submitted_answer = agent(question_text)
84
- answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
85
- results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
86
- except Exception as e:
87
- print(f"Error running agent on task {task_id}: {e}")
88
- results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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 = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
 
 
 
 
 
96
  status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
97
  print(status_update)
98
 
@@ -133,7 +354,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
133
  print(status_message)
134
  results_df = pd.DataFrame(results_log)
135
  return status_message, results_df
136
- except Exception as e:
137
  status_message = f"An unexpected error occurred during submission: {e}"
138
  print(status_message)
139
  results_df = pd.DataFrame(results_log)
@@ -146,11 +367,9 @@ with gr.Blocks() as demo:
146
  gr.Markdown(
147
  """
148
  **Instructions:**
149
-
150
  1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
151
  2. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
152
  3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
153
-
154
  ---
155
  **Disclaimers:**
156
  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).
@@ -162,20 +381,19 @@ with gr.Blocks() as demo:
162
 
163
  run_button = gr.Button("Run Evaluation & Submit All Answers")
164
 
165
- status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
 
 
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") # Get SPACE_ID at startup
179
 
180
  if space_host_startup:
181
  print(f"✅ SPACE_HOST found: {space_host_startup}")
@@ -183,14 +401,18 @@ if __name__ == "__main__":
183
  else:
184
  print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
185
 
186
- if space_id_startup: # Print repo URLs if SPACE_ID is found
187
  print(f"✅ SPACE_ID found: {space_id_startup}")
188
  print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
189
- print(f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
 
 
190
  else:
191
- print("ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
 
 
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
+ """app.py"""
2
+
3
  import os
4
+ import re
5
+ import pathlib
6
+ import tempfile
7
+ from pathlib import Path
8
+ from typing import Union, Optional
9
+
10
+ import openai
11
  import requests
12
+ import gradio as gr
13
  import pandas as pd
14
+ from tabulate import tabulate
15
+ from smolagents import (
16
+ OpenAIServerModel,
17
+ DuckDuckGoSearchTool,
18
+ CodeAgent,
19
+ WikipediaSearchTool,
20
+ )
21
+ from smolagents.tools import PipelineTool, Tool
22
+
23
 
24
  # (Keep Constants as is)
25
  # --- Constants ---
26
  DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
27
 
28
+
29
+ class SpeechToTextTool(PipelineTool):
30
+ """
31
+ Transcribes an audio file to text using the OpenAI Whisper API.
32
+ Only local file paths are supported.
33
+ """
34
+
35
+ default_checkpoint = "openai/whisper-1" # purely informational here
36
+ description = (
37
+ "This tool sends an audio file to OpenAI Whisper and returns the "
38
+ "transcribed text."
39
+ )
40
+ name = "transcriber"
41
+ inputs = {
42
+ "audio": {
43
+ "type": "string",
44
+ "description": "Absolute or relative path to a local audio file.",
45
+ }
46
+ }
47
+ output_type = "string"
48
+
49
+ # ──────────────────────────────────────────────────────────────────────────
50
+ # Public interface
51
+ # ──────────────────────────────────────────────────────────────────────────
52
+ def __call__(self, audio: str) -> str:
53
+ """
54
+ Convenience wrapper so the tool can be used like a regular function:
55
+ text = SpeechToTextTool()(path_to_audio)
56
+ """
57
+ return self._transcribe(audio)
58
+
59
+ # ──────────────────────────────────────────────────────────────────────────
60
+ # Internal helpers
61
+ # ──────────────────────────────────────────────────────────────────────────
62
+ @staticmethod
63
+ def _transcribe(audio_path: str) -> str:
64
+ # ----- validation ----------------------------------------------------
65
+ if not isinstance(audio_path, str):
66
+ raise TypeError(
67
+ "Parameter 'audio' must be a string containing the file path."
68
+ )
69
+ path = Path(audio_path).expanduser().resolve()
70
+ if not path.is_file():
71
+ raise FileNotFoundError(f"No such audio file: {path}")
72
+
73
+ # ----- API call ------------------------------------------------------
74
+ with path.open("rb") as fp:
75
+ response = openai.audio.transcriptions.create(
76
+ file=fp,
77
+ model="whisper-1", # currently the only Whisper model
78
+ response_format="text", # returns plain text instead of JSON
79
+ )
80
+
81
+ # For response_format="text", `response` is already the raw transcript
82
+ return response
83
+
84
+
85
+ class ExcelToTextTool(Tool):
86
+ """Render an Excel worksheet as Markdown text."""
87
+
88
+ # ------------------------------------------------------------------
89
+ # Required smol‑agents metadata
90
+ # ------------------------------------------------------------------
91
+ name = "excel_to_text"
92
+ description = (
93
+ "Read an Excel file and return a Markdown table of the requested sheet. "
94
+ "Accepts either the sheet name or the zero-based index."
95
+ )
96
+
97
+ inputs = {
98
+ "excel_path": {
99
+ "type": "string",
100
+ "description": "Path to the Excel file (.xlsx / .xls).",
101
+ },
102
+ "sheet_name": {
103
+ "type": "string",
104
+ "description": (
105
+ "Worksheet name or zero-based index *as a string* (optional; default first sheet)."
106
+ ),
107
+ "nullable": True,
108
+ },
109
+ }
110
+
111
+ output_type = "string"
112
+
113
+ def forward(
114
+ self,
115
+ excel_path: str,
116
+ sheet_name: Optional[str] = None,
117
+ ) -> str:
118
+ """Load *excel_path* and return the sheet as a Markdown table."""
119
+
120
+ path = pathlib.Path(excel_path).expanduser().resolve()
121
+ if not path.exists():
122
+ return f"Error: Excel file not found at {path}"
123
+
124
+ try:
125
+ # Interpret sheet identifier -----------------------------------
126
+ sheet: Union[str, int]
127
+ if sheet_name is None or sheet_name == "":
128
+ sheet = 0 # first sheet
129
+ else:
130
+ # If the user passed a numeric string (e.g. "1"), cast to int
131
+ sheet = int(sheet_name) if sheet_name.isdigit() else sheet_name
132
+
133
+ # Load worksheet ----------------------------------------------
134
+ df = pd.read_excel(path, sheet_name=sheet)
135
+
136
+ # Render to Markdown; fall back to tabulate if needed ---------
137
+ if hasattr(pd.DataFrame, "to_markdown"):
138
+ return df.to_markdown(index=False)
139
+
140
+ return tabulate(df, headers="keys", tablefmt="github", showindex=False)
141
+
142
+ except Exception as exc: # pylint: disable=broad-except
143
+ return f"Error reading Excel file: {exc}"
144
+
145
+
146
+ def download_file_if_any(base_api_url: str, task_id: str) -> str | None:
147
+ """
148
+ Try GET /files/{task_id}.
149
+ • On HTTP 200 → save to a temp dir and return local path.
150
+ • On 404 → return None.
151
+ • On other errors → raise so caller can log / handle.
152
+ """
153
+ url = f"{base_api_url}/files/{task_id}"
154
+ try:
155
+ resp = requests.get(url, timeout=30)
156
+ if resp.status_code == 404:
157
+ return None # no file
158
+ resp.raise_for_status() # raise on 4xx/5xx ≠ 404
159
+ except requests.exceptions.HTTPError as e:
160
+ # propagate non-404 errors (403, 500, …)
161
+ raise e
162
+
163
+ # ▸ Save bytes to a named file inside the system temp dir
164
+ # Try to keep original extension from Content-Disposition if present.
165
+ cdisp = resp.headers.get("content-disposition", "")
166
+ filename = task_id # default base name
167
+ if "filename=" in cdisp:
168
+ m = re.search(r'filename="([^"]+)"', cdisp)
169
+ if m:
170
+ filename = m.group(1) # keep provided name
171
+
172
+ tmp_dir = Path(tempfile.gettempdir()) / "gaia_files"
173
+ tmp_dir.mkdir(exist_ok=True)
174
+ file_path = tmp_dir / filename
175
+ with open(file_path, "wb") as f:
176
+ f.write(resp.content)
177
+ return str(file_path)
178
+
179
+
180
  class BasicAgent:
181
+ """Basic Agent for the evaluation task."""
182
+
183
  def __init__(self):
184
+ self.agent = CodeAgent(
185
+ model=OpenAIServerModel(model_id="gpt-4o"),
186
+ tools=[
187
+ DuckDuckGoSearchTool(),
188
+ WikipediaSearchTool(),
189
+ SpeechToTextTool(),
190
+ ExcelToTextTool(),
191
+ ],
192
+ add_base_tools=True,
193
+ additional_authorized_imports=["pandas", "numpy", "csv", "subprocess"],
194
+ )
195
+
196
  print("BasicAgent initialized.")
197
+
198
  def __call__(self, question: str) -> str:
199
  print(f"Agent received question (first 50 chars): {question[:50]}...")
200
+ fixed_answer = self.agent.run(question)
201
+ print(f"Agent returning answer: {fixed_answer}")
202
  return fixed_answer
203
 
204
+
205
+ def run_and_submit_all(profile: gr.OAuthProfile | None):
206
  """
207
  Fetches all questions, runs the BasicAgent on them, submits all answers,
208
  and displays the results.
209
  """
210
  # --- Determine HF Space Runtime URL and Repo URL ---
211
+ space_id = "l3xv/Final_Assignment_Template"
212
 
213
  if profile:
214
+ username = f"{profile.username}"
215
  print(f"User logged in: {username}")
216
  else:
217
  print("User not logged in.")
 
221
  questions_url = f"{api_url}/questions"
222
  submit_url = f"{api_url}/submit"
223
 
 
224
  try:
225
  agent = BasicAgent()
226
+ except Exception as e: # pylint: disable=broad-except
227
  print(f"Error instantiating agent: {e}")
228
  return f"Error initializing agent: {e}", None
229
  # 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)
 
237
  response.raise_for_status()
238
  questions_data = response.json()
239
  if not questions_data:
240
+ print("Fetched questions list is empty.")
241
+ return "Fetched questions list is empty or invalid format.", None
242
  print(f"Fetched {len(questions_data)} questions.")
243
+ except requests.exceptions.JSONDecodeError as e:
244
+ print(f"Error decoding JSON response from questions endpoint: {e}")
245
+ print(f"Response text: {response.text[:500]}")
246
+ return f"Error decoding server response for questions: {e}", None
247
  except requests.exceptions.RequestException as e:
248
  print(f"Error fetching questions: {e}")
249
  return f"Error fetching questions: {e}", None
250
+ except Exception as e: # pylint: disable=broad-except
 
 
 
 
251
  print(f"An unexpected error occurred fetching questions: {e}")
252
  return f"An unexpected error occurred fetching questions: {e}", None
253
 
 
258
  for item in questions_data:
259
  task_id = item.get("task_id")
260
  question_text = item.get("question")
261
+
262
+ # ----------fetch any attached file ----------
263
+ try:
264
+ file_path = download_file_if_any(api_url, task_id)
265
+ except Exception as e: # pylint: disable=broad-except
266
+ file_path = None
267
+ print(f"[file fetch error] {task_id}: {e}")
268
+
269
+ # ---------- Build the prompt sent to the agent ----------
270
+ if file_path:
271
+ q_for_agent = (
272
+ f"{question_text}\n\n"
273
+ f"---\n"
274
+ f"A file was downloaded for this task and saved locally at:\n"
275
+ f"{file_path}\n"
276
+ f"---\n\n"
277
+ )
278
+ else:
279
+ q_for_agent = question_text
280
+
281
  if not task_id or question_text is None:
282
  print(f"Skipping item with missing task_id or question: {item}")
283
  continue
284
  try:
285
+ submitted_answer = agent(q_for_agent)
286
+ answers_payload.append(
287
+ {"task_id": task_id, "submitted_answer": submitted_answer}
288
+ )
289
+ results_log.append(
290
+ {
291
+ "Task ID": task_id,
292
+ "Question": question_text,
293
+ "Submitted Answer": submitted_answer,
294
+ }
295
+ )
296
+ except Exception as e: # pylint: disable=broad-except
297
+ print(f"Error running agent on task {task_id}: {e}")
298
+ results_log.append(
299
+ {
300
+ "Task ID": task_id,
301
+ "Question": question_text,
302
+ "Submitted Answer": f"AGENT ERROR: {e}",
303
+ }
304
+ )
305
 
306
  if not answers_payload:
307
  print("Agent did not produce any answers to submit.")
308
  return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
309
 
310
+ # 4. Prepare Submission
311
+ submission_data = {
312
+ "username": username.strip(),
313
+ "agent_code": agent_code,
314
+ "answers": answers_payload,
315
+ }
316
+
317
  status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
318
  print(status_update)
319
 
 
354
  print(status_message)
355
  results_df = pd.DataFrame(results_log)
356
  return status_message, results_df
357
+ except Exception as e: # pylint: disable=broad-except
358
  status_message = f"An unexpected error occurred during submission: {e}"
359
  print(status_message)
360
  results_df = pd.DataFrame(results_log)
 
367
  gr.Markdown(
368
  """
369
  **Instructions:**
 
370
  1. Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
371
  2. Log in to your Hugging Face account using the button below. This uses your HF username for submission.
372
  3. Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
 
373
  ---
374
  **Disclaimers:**
375
  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).
 
381
 
382
  run_button = gr.Button("Run Evaluation & Submit All Answers")
383
 
384
+ status_output = gr.Textbox(
385
+ label="Run Status / Submission Result", lines=5, interactive=False
386
+ )
387
  # Removed max_rows=10 from DataFrame constructor
388
  results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
389
 
390
+ run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
 
 
 
391
 
392
  if __name__ == "__main__":
393
+ print("\n" + "-" * 30 + " App Starting " + "-" * 30)
394
  # Check for SPACE_HOST and SPACE_ID at startup for information
395
  space_host_startup = os.getenv("SPACE_HOST")
396
+ space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
397
 
398
  if space_host_startup:
399
  print(f"✅ SPACE_HOST found: {space_host_startup}")
 
401
  else:
402
  print("ℹ️ SPACE_HOST environment variable not found (running locally?).")
403
 
404
+ if space_id_startup: # Print repo URLs if SPACE_ID is found
405
  print(f"✅ SPACE_ID found: {space_id_startup}")
406
  print(f" Repo URL: https://huggingface.co/spaces/{space_id_startup}")
407
+ print(
408
+ f" Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main"
409
+ )
410
  else:
411
+ print(
412
+ "ℹ️ SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined."
413
+ )
414
 
415
+ print("-" * (60 + len(" App Starting ")) + "\n")
416
 
417
  print("Launching Gradio Interface for Basic Agent Evaluation...")
418
+ demo.launch(debug=True, share=False)