Update app.py
Browse files
app.py
CHANGED
@@ -1,190 +1,29 @@
|
|
1 |
import os
|
2 |
import gradio as gr
|
3 |
import requests
|
4 |
-
import openai
|
5 |
-
from smolagents import OpenAIServerModel, DuckDuckGoSearchTool, CodeAgent, WikipediaSearchTool
|
6 |
-
from pathlib import Path
|
7 |
-
import tempfile
|
8 |
-
from smolagents.tools import PipelineTool, Tool
|
9 |
-
import pathlib
|
10 |
-
from typing import Union, Optional
|
11 |
import pandas as pd
|
12 |
-
from
|
13 |
-
|
14 |
|
15 |
# (Keep Constants as is)
|
16 |
# --- Constants ---
|
17 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
18 |
|
19 |
-
class SpeechToTextTool(PipelineTool):
|
20 |
-
"""
|
21 |
-
Transcribes an audio file to text using the OpenAI Whisper API.
|
22 |
-
Only local file paths are supported.
|
23 |
-
"""
|
24 |
-
default_checkpoint = "openai/whisper-1" # purely informational here
|
25 |
-
description = (
|
26 |
-
"This tool sends an audio file to OpenAI Whisper and returns the "
|
27 |
-
"transcribed text."
|
28 |
-
)
|
29 |
-
name = "transcriber"
|
30 |
-
inputs = {
|
31 |
-
"audio": {
|
32 |
-
"type": "string",
|
33 |
-
"description": "Absolute or relative path to a local audio file.",
|
34 |
-
}
|
35 |
-
}
|
36 |
-
output_type = "string"
|
37 |
-
|
38 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
39 |
-
# Public interface
|
40 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
41 |
-
def __call__(self, audio: str) -> str:
|
42 |
-
"""
|
43 |
-
Convenience wrapper so the tool can be used like a regular function:
|
44 |
-
text = SpeechToTextTool()(path_to_audio)
|
45 |
-
"""
|
46 |
-
return self._transcribe(audio)
|
47 |
-
|
48 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
49 |
-
# Internal helpers
|
50 |
-
# ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
51 |
-
@staticmethod
|
52 |
-
def _transcribe(audio_path: str) -> str:
|
53 |
-
# ----- validation ----------------------------------------------------
|
54 |
-
if not isinstance(audio_path, str):
|
55 |
-
raise TypeError(
|
56 |
-
"Parameter 'audio' must be a string containing the file path."
|
57 |
-
)
|
58 |
-
path = Path(audio_path).expanduser().resolve()
|
59 |
-
if not path.is_file():
|
60 |
-
raise FileNotFoundError(f"No such audio file: {path}")
|
61 |
-
|
62 |
-
# ----- API call ------------------------------------------------------
|
63 |
-
with path.open("rb") as fp:
|
64 |
-
response = openai.audio.transcriptions.create(
|
65 |
-
file=fp,
|
66 |
-
model="whisper-1", # currently the only Whisper model
|
67 |
-
response_format="text" # returns plain text instead of JSON
|
68 |
-
)
|
69 |
-
|
70 |
-
# For response_format="text", `response` is already the raw transcript
|
71 |
-
return response
|
72 |
-
|
73 |
-
class ExcelToTextTool(Tool):
|
74 |
-
"""Render an Excel worksheet as Markdown text."""
|
75 |
-
|
76 |
-
# ------------------------------------------------------------------
|
77 |
-
# Required smolβagents metadata
|
78 |
-
# ------------------------------------------------------------------
|
79 |
-
name = "excel_to_text"
|
80 |
-
description = (
|
81 |
-
"Read an Excel file and return a Markdown table of the requested sheet. "
|
82 |
-
"Accepts either the sheet name or the zero-based index."
|
83 |
-
)
|
84 |
-
|
85 |
-
inputs = {
|
86 |
-
"excel_path": {
|
87 |
-
"type": "string",
|
88 |
-
"description": "Path to the Excel file (.xlsx / .xls).",
|
89 |
-
},
|
90 |
-
"sheet_name": {
|
91 |
-
"type": "string",
|
92 |
-
"description": (
|
93 |
-
"Worksheet name or zeroβbased index *as a string* (optional; default first sheet)."
|
94 |
-
),
|
95 |
-
"nullable": True,
|
96 |
-
},
|
97 |
-
}
|
98 |
-
|
99 |
-
output_type = "string"
|
100 |
-
|
101 |
-
# ------------------------------------------------------------------
|
102 |
-
# Core logic
|
103 |
-
# ------------------------------------------------------------------
|
104 |
-
def forward(
|
105 |
-
self,
|
106 |
-
excel_path: str,
|
107 |
-
sheet_name: Optional[str] = None,
|
108 |
-
) -> str:
|
109 |
-
"""Load *excel_path* and return the sheet as a Markdown table."""
|
110 |
-
|
111 |
-
path = pathlib.Path(excel_path).expanduser().resolve()
|
112 |
-
if not path.exists():
|
113 |
-
return f"Error: Excel file not found at {path}"
|
114 |
-
|
115 |
-
try:
|
116 |
-
# Interpret sheet identifier -----------------------------------
|
117 |
-
sheet: Union[str, int]
|
118 |
-
if sheet_name is None or sheet_name == "":
|
119 |
-
sheet = 0 # first sheet
|
120 |
-
else:
|
121 |
-
# If the user passed a numeric string (e.g. "1"), cast to int
|
122 |
-
sheet = int(sheet_name) if sheet_name.isdigit() else sheet_name
|
123 |
-
|
124 |
-
# Load worksheet ----------------------------------------------
|
125 |
-
df = pd.read_excel(path, sheet_name=sheet)
|
126 |
-
|
127 |
-
# Render to Markdown; fall back to tabulate if needed ---------
|
128 |
-
if hasattr(pd.DataFrame, "to_markdown"):
|
129 |
-
return df.to_markdown(index=False)
|
130 |
-
from tabulate import tabulate # pragma: no cover β fallback path
|
131 |
-
|
132 |
-
return tabulate(df, headers="keys", tablefmt="github", showindex=False)
|
133 |
-
|
134 |
-
except Exception as exc: # broad catch keeps the agent chatβfriendly
|
135 |
-
return f"Error reading Excel file: {exc}"
|
136 |
-
|
137 |
-
|
138 |
-
def download_file_if_any(base_api_url: str, task_id: str) -> str | None:
|
139 |
-
"""
|
140 |
-
Try GET /files/{task_id}.
|
141 |
-
β’ On HTTP 200 β save to a temp dir and return local path.
|
142 |
-
β’ On 404 β return None.
|
143 |
-
β’ On other errors β raise so caller can log / handle.
|
144 |
-
"""
|
145 |
-
url = f"{base_api_url}/files/{task_id}"
|
146 |
-
try:
|
147 |
-
resp = requests.get(url, timeout=30)
|
148 |
-
if resp.status_code == 404:
|
149 |
-
return None # no file
|
150 |
-
resp.raise_for_status() # raise on 4xx/5xx β 404
|
151 |
-
except requests.exceptions.HTTPError as e:
|
152 |
-
# propagate non-404 errors (403, 500, β¦)
|
153 |
-
raise e
|
154 |
-
|
155 |
-
# βΈ Save bytes to a named file inside the system temp dir
|
156 |
-
# Try to keep original extension from Content-Disposition if present.
|
157 |
-
cdisp = resp.headers.get("content-disposition", "")
|
158 |
-
filename = task_id # default base name
|
159 |
-
if "filename=" in cdisp:
|
160 |
-
m = re.search(r'filename="([^"]+)"', cdisp)
|
161 |
-
if m:
|
162 |
-
filename = m.group(1) # keep provided name
|
163 |
-
|
164 |
-
tmp_dir = Path(tempfile.gettempdir()) / "gaia_files"
|
165 |
-
tmp_dir.mkdir(exist_ok=True)
|
166 |
-
file_path = tmp_dir / filename
|
167 |
-
with open(file_path, "wb") as f:
|
168 |
-
f.write(resp.content)
|
169 |
-
return str(file_path)
|
170 |
-
|
171 |
# --- Basic Agent Definition ---
|
172 |
# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
|
173 |
class BasicAgent:
|
174 |
def __init__(self):
|
175 |
-
self.agent = CodeAgent(
|
176 |
-
model=OpenAIServerModel(model_id="gpt-4o"),
|
177 |
-
tools=[DuckDuckGoSearchTool(), WikipediaSearchTool(), SpeechToTextTool(), ExcelToTextTool()],
|
178 |
-
add_base_tools=True,
|
179 |
-
additional_authorized_imports=['pandas','numpy','csv','subprocess']
|
180 |
-
)
|
181 |
-
|
182 |
print("BasicAgent initialized.")
|
183 |
-
|
|
|
|
|
|
|
|
|
|
|
184 |
def __call__(self, question: str) -> str:
|
185 |
print(f"Agent received question (first 50 chars): {question[:50]}...")
|
186 |
-
fixed_answer =
|
187 |
-
print(f"Agent returning answer: {fixed_answer}")
|
188 |
return fixed_answer
|
189 |
|
190 |
def run_and_submit_all( profile: gr.OAuthProfile | None):
|
@@ -193,7 +32,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
193 |
and displays the results.
|
194 |
"""
|
195 |
# --- Determine HF Space Runtime URL and Repo URL ---
|
196 |
-
space_id = "
|
197 |
|
198 |
if profile:
|
199 |
username= f"{profile.username}"
|
@@ -244,31 +83,11 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
244 |
for item in questions_data:
|
245 |
task_id = item.get("task_id")
|
246 |
question_text = item.get("question")
|
247 |
-
|
248 |
-
# ----------fetch any attached file ----------
|
249 |
-
try:
|
250 |
-
file_path = download_file_if_any(api_url, task_id)
|
251 |
-
except Exception as e:
|
252 |
-
file_path = None
|
253 |
-
print(f"[file fetch error] {task_id}: {e}")
|
254 |
-
|
255 |
-
# ---------- Build the prompt sent to the agent ----------
|
256 |
-
if file_path:
|
257 |
-
q_for_agent = (
|
258 |
-
f"{question_text}\n\n"
|
259 |
-
f"---\n"
|
260 |
-
f"A file was downloaded for this task and saved locally at:\n"
|
261 |
-
f"{file_path}\n"
|
262 |
-
f"---\n\n"
|
263 |
-
)
|
264 |
-
else:
|
265 |
-
q_for_agent = question_text
|
266 |
-
|
267 |
if not task_id or question_text is None:
|
268 |
print(f"Skipping item with missing task_id or question: {item}")
|
269 |
continue
|
270 |
try:
|
271 |
-
submitted_answer = agent(
|
272 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
273 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
274 |
except Exception as e:
|
@@ -361,7 +180,7 @@ if __name__ == "__main__":
|
|
361 |
print("\n" + "-"*30 + " App Starting " + "-"*30)
|
362 |
# Check for SPACE_HOST and SPACE_ID at startup for information
|
363 |
space_host_startup = os.getenv("SPACE_HOST")
|
364 |
-
space_id_startup = "
|
365 |
|
366 |
if space_host_startup:
|
367 |
print(f"β
SPACE_HOST found: {space_host_startup}")
|
|
|
1 |
import os
|
2 |
import gradio as gr
|
3 |
import requests
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
import pandas as pd
|
5 |
+
from smolagents import CodeAgent, DuckDuckGoSearchTool, HFApiModel
|
6 |
+
# Test comment
|
7 |
|
8 |
# (Keep Constants as is)
|
9 |
# --- Constants ---
|
10 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
# --- Basic Agent Definition ---
|
13 |
# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
|
14 |
class BasicAgent:
|
15 |
def __init__(self):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
print("BasicAgent initialized.")
|
17 |
+
token_key = os.getenv("HFAPI_KEY")
|
18 |
+
# Initialize the search tool
|
19 |
+
model = HFApiModel(model_id="Qwen/Qwen2.5-7B-Instruct", api_key=token_key)
|
20 |
+
search_tool = DuckDuckGoSearchTool()
|
21 |
+
|
22 |
+
self.agent = CodeAgent(model=model, tools=[search_tool])
|
23 |
def __call__(self, question: str) -> str:
|
24 |
print(f"Agent received question (first 50 chars): {question[:50]}...")
|
25 |
+
fixed_answer = "This is a default answer."
|
26 |
+
print(f"Agent returning fixed answer: {fixed_answer}")
|
27 |
return fixed_answer
|
28 |
|
29 |
def run_and_submit_all( profile: gr.OAuthProfile | None):
|
|
|
32 |
and displays the results.
|
33 |
"""
|
34 |
# --- Determine HF Space Runtime URL and Repo URL ---
|
35 |
+
space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
|
36 |
|
37 |
if profile:
|
38 |
username= f"{profile.username}"
|
|
|
83 |
for item in questions_data:
|
84 |
task_id = item.get("task_id")
|
85 |
question_text = item.get("question")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
if not task_id or question_text is None:
|
87 |
print(f"Skipping item with missing task_id or question: {item}")
|
88 |
continue
|
89 |
try:
|
90 |
+
submitted_answer = agent(question_text)
|
91 |
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
|
92 |
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
|
93 |
except Exception as e:
|
|
|
180 |
print("\n" + "-"*30 + " App Starting " + "-"*30)
|
181 |
# Check for SPACE_HOST and SPACE_ID at startup for information
|
182 |
space_host_startup = os.getenv("SPACE_HOST")
|
183 |
+
space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
|
184 |
|
185 |
if space_host_startup:
|
186 |
print(f"β
SPACE_HOST found: {space_host_startup}")
|