Spaces:
Sleeping
Sleeping
Update agent.py
Browse files
agent.py
CHANGED
@@ -7,11 +7,8 @@ import tempfile
|
|
7 |
import os
|
8 |
import whisper
|
9 |
|
10 |
-
|
11 |
SYSTEM_PROMPT = """
|
12 |
-
You are a
|
13 |
-
Then return only the answer without any explanation or formatting.
|
14 |
-
Do not say 'Final answer' or anything else. Just output the raw answer string.
|
15 |
"""
|
16 |
|
17 |
class GaiaAgent:
|
@@ -29,44 +26,42 @@ class GaiaAgent:
|
|
29 |
if results:
|
30 |
return results[0]['body']
|
31 |
except Exception as e:
|
32 |
-
return
|
33 |
return ""
|
34 |
|
35 |
def transcribe_audio(self, file_path: str) -> str:
|
36 |
try:
|
37 |
result = self.transcriber.transcribe(file_path)
|
38 |
return result['text']
|
39 |
-
except Exception
|
40 |
-
return
|
41 |
|
42 |
def handle_excel(self, file_path: str) -> str:
|
43 |
try:
|
44 |
df = pd.read_excel(file_path)
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
49 |
|
50 |
def __call__(self, question: str, files: dict = None) -> tuple[str, str]:
|
51 |
try:
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
for key in file_keys:
|
58 |
-
if key.endswith(".mp3"):
|
59 |
-
audio_txt = self.transcribe_audio(files[key])
|
60 |
-
prompt = f"{SYSTEM_PROMPT}\n\n{audio_txt}\n\n{question}"
|
61 |
break
|
62 |
-
elif
|
63 |
-
excel_result = self.handle_excel(
|
64 |
-
return excel_result, excel_result
|
65 |
-
|
66 |
-
|
67 |
-
else:
|
68 |
-
prompt = f"{SYSTEM_PROMPT}\n\n{question}"
|
69 |
|
|
|
70 |
inputs = self.tokenizer(prompt, return_tensors="pt", truncation=True).to(self.device)
|
71 |
outputs = self.model.generate(
|
72 |
**inputs,
|
@@ -77,7 +72,7 @@ class GaiaAgent:
|
|
77 |
)
|
78 |
output_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
79 |
final = output_text.strip()
|
80 |
-
return final,
|
81 |
except Exception as e:
|
82 |
return "ERROR", f"Agent failed: {e}"
|
83 |
|
|
|
7 |
import os
|
8 |
import whisper
|
9 |
|
|
|
10 |
SYSTEM_PROMPT = """
|
11 |
+
You are a helpful AI assistant. Think step by step to solve the problem. If the question requires reasoning, perform it. If it refers to a search or file, use the result provided. At the end, return ONLY the final answer string. No explanations.
|
|
|
|
|
12 |
"""
|
13 |
|
14 |
class GaiaAgent:
|
|
|
26 |
if results:
|
27 |
return results[0]['body']
|
28 |
except Exception as e:
|
29 |
+
return ""
|
30 |
return ""
|
31 |
|
32 |
def transcribe_audio(self, file_path: str) -> str:
|
33 |
try:
|
34 |
result = self.transcriber.transcribe(file_path)
|
35 |
return result['text']
|
36 |
+
except Exception:
|
37 |
+
return ""
|
38 |
|
39 |
def handle_excel(self, file_path: str) -> str:
|
40 |
try:
|
41 |
df = pd.read_excel(file_path)
|
42 |
+
df.columns = [col.lower() for col in df.columns]
|
43 |
+
if 'category' in df.columns and 'sales' in df.columns:
|
44 |
+
food_sales = df[df['category'].str.lower() != 'drink']['sales'].sum()
|
45 |
+
return f"{food_sales:.2f}"
|
46 |
+
except Exception:
|
47 |
+
return ""
|
48 |
+
return ""
|
49 |
|
50 |
def __call__(self, question: str, files: dict = None) -> tuple[str, str]:
|
51 |
try:
|
52 |
+
context = ""
|
53 |
+
if files:
|
54 |
+
for filename, filepath in files.items():
|
55 |
+
if filename.endswith(".mp3"):
|
56 |
+
context = self.transcribe_audio(filepath)
|
|
|
|
|
|
|
|
|
57 |
break
|
58 |
+
elif filename.endswith(".xlsx"):
|
59 |
+
excel_result = self.handle_excel(filepath)
|
60 |
+
return excel_result.strip(), excel_result.strip()
|
61 |
+
elif "http" in question.lower() or "wikipedia" in question.lower():
|
62 |
+
context = self.search(question)
|
|
|
|
|
63 |
|
64 |
+
prompt = f"{SYSTEM_PROMPT}\n\n{context}\n\nQuestion: {question.strip()}"
|
65 |
inputs = self.tokenizer(prompt, return_tensors="pt", truncation=True).to(self.device)
|
66 |
outputs = self.model.generate(
|
67 |
**inputs,
|
|
|
72 |
)
|
73 |
output_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
74 |
final = output_text.strip()
|
75 |
+
return final, final
|
76 |
except Exception as e:
|
77 |
return "ERROR", f"Agent failed: {e}"
|
78 |
|