updated error handling
Browse files- llm/inference.py +69 -2
llm/inference.py
CHANGED
@@ -12,7 +12,7 @@ nltk.download('averaged_perceptron_tagger')
|
|
12 |
|
13 |
client = InferenceClient(api_key=api_key)
|
14 |
|
15 |
-
|
16 |
def extract_product_info(text):
|
17 |
print(f'Extract function called!')
|
18 |
# Initialize result dictionary
|
@@ -57,7 +57,74 @@ def extract_product_info(text):
|
|
57 |
result["description"] = " ".join(description_parts)
|
58 |
print(f'extract function returned:\n{result}')
|
59 |
return result
|
60 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
61 |
|
62 |
|
63 |
def extract_info(text):
|
|
|
12 |
|
13 |
client = InferenceClient(api_key=api_key)
|
14 |
|
15 |
+
'''
|
16 |
def extract_product_info(text):
|
17 |
print(f'Extract function called!')
|
18 |
# Initialize result dictionary
|
|
|
57 |
result["description"] = " ".join(description_parts)
|
58 |
print(f'extract function returned:\n{result}')
|
59 |
return result
|
60 |
+
'''
|
61 |
+
def extract_product_info(text):
|
62 |
+
print(f"Extract function called with input: {text}")
|
63 |
+
|
64 |
+
# Initialize result dictionary
|
65 |
+
result = {"brand": None, "model": None, "description": None, "price": None}
|
66 |
+
|
67 |
+
try:
|
68 |
+
# Extract price using regex
|
69 |
+
price_match = re.search(r'\$\s?\d{1,3}(?:,\d{3})*(?:\.\d{2})?', text)
|
70 |
+
print(f"Price match: {price_match}")
|
71 |
+
if price_match:
|
72 |
+
result["price"] = price_match.group().replace("$", "").replace(",", "").strip()
|
73 |
+
# Remove the price part from the text to prevent interference
|
74 |
+
text = text.replace(price_match.group(), "").strip()
|
75 |
+
print(f"Text after removing price: {text}")
|
76 |
+
|
77 |
+
# Tokenize the remaining text
|
78 |
+
try:
|
79 |
+
tokens = nltk.word_tokenize(text)
|
80 |
+
print(f"Tokens: {tokens}")
|
81 |
+
except Exception as e:
|
82 |
+
print(f"Error during tokenization: {e}")
|
83 |
+
# Fall back to a simple split if tokenization fails
|
84 |
+
tokens = text.split()
|
85 |
+
print(f"Fallback tokens: {tokens}")
|
86 |
+
|
87 |
+
# POS tagging
|
88 |
+
try:
|
89 |
+
pos_tags = nltk.pos_tag(tokens)
|
90 |
+
print(f"POS Tags: {pos_tags}")
|
91 |
+
except Exception as e:
|
92 |
+
print(f"Error during POS tagging: {e}")
|
93 |
+
# If POS tagging fails, create dummy tags
|
94 |
+
pos_tags = [(word, "NN") for word in tokens]
|
95 |
+
print(f"Fallback POS Tags: {pos_tags}")
|
96 |
+
|
97 |
+
# Extract brand, model, and description
|
98 |
+
brand_parts = []
|
99 |
+
model_parts = []
|
100 |
+
description_parts = []
|
101 |
+
|
102 |
+
for word, tag in pos_tags:
|
103 |
+
if tag == 'NNP' or re.match(r'[A-Za-z0-9-]+', word):
|
104 |
+
if len(brand_parts) == 0: # Assume the first proper noun is the brand
|
105 |
+
brand_parts.append(word)
|
106 |
+
else: # Model number tends to follow the brand
|
107 |
+
model_parts.append(word)
|
108 |
+
else:
|
109 |
+
description_parts.append(word)
|
110 |
+
|
111 |
+
# Assign values to the result dictionary
|
112 |
+
if brand_parts:
|
113 |
+
result["brand"] = " ".join(brand_parts)
|
114 |
+
if model_parts:
|
115 |
+
result["model"] = " ".join(model_parts)
|
116 |
+
if description_parts:
|
117 |
+
result["description"] = " ".join(description_parts)
|
118 |
+
|
119 |
+
print(f"Extract function returned: {result}")
|
120 |
+
|
121 |
+
except Exception as e:
|
122 |
+
print(f"Unexpected error: {e}")
|
123 |
+
# Return a fallback result in case of a critical error
|
124 |
+
result["description"] = text
|
125 |
+
print(f"Fallback result: {result}")
|
126 |
+
|
127 |
+
return result
|
128 |
|
129 |
|
130 |
def extract_info(text):
|