Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -12,6 +12,7 @@ from tenacity import retry, stop_after_attempt, wait_fixed
|
|
12 |
import tempfile
|
13 |
from typing import Union, Dict, List
|
14 |
from contextlib import contextmanager
|
|
|
15 |
|
16 |
# Constants
|
17 |
DEFAULT_LANGUAGE = "English"
|
@@ -69,8 +70,15 @@ class OCRProcessor:
|
|
69 |
|
70 |
def _get_file_content(self, file_input: Union[str, bytes]) -> bytes:
|
71 |
if isinstance(file_input, str):
|
72 |
-
|
73 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
74 |
return file_input.read() if hasattr(file_input, 'read') else file_input
|
75 |
|
76 |
def ocr_pdf_url(self, pdf_url: str) -> str:
|
@@ -140,7 +148,7 @@ class OCRProcessor:
|
|
140 |
base64_url = f"data:image/jpeg;base64,{encoded_image}"
|
141 |
ocr_response = self._call_ocr_api({"type": "image_url", "image_url": base64_url})
|
142 |
markdown = self._extract_markdown(ocr_response)
|
143 |
-
|
144 |
chat_response = self._call_chat_complete(
|
145 |
model="pixtral-12b-latest",
|
146 |
messages=[{
|
@@ -156,12 +164,12 @@ class OCRProcessor:
|
|
156 |
response_format={"type": "json_object"},
|
157 |
temperature=0
|
158 |
)
|
159 |
-
|
160 |
# Ensure the response is a dictionary
|
161 |
response_content = chat_response.choices[0].message.content
|
162 |
if isinstance(response_content, list):
|
163 |
response_content = response_content[0] if response_content else "{}"
|
164 |
-
|
165 |
content = json.loads(response_content)
|
166 |
return self._format_structured_response(temp_path, content)
|
167 |
except Exception as e:
|
@@ -267,4 +275,4 @@ def create_interface():
|
|
267 |
return demo
|
268 |
|
269 |
if __name__ == "__main__":
|
270 |
-
create_interface().launch(share=True, debug=True)
|
|
|
12 |
import tempfile
|
13 |
from typing import Union, Dict, List
|
14 |
from contextlib import contextmanager
|
15 |
+
import requests
|
16 |
|
17 |
# Constants
|
18 |
DEFAULT_LANGUAGE = "English"
|
|
|
70 |
|
71 |
def _get_file_content(self, file_input: Union[str, bytes]) -> bytes:
|
72 |
if isinstance(file_input, str):
|
73 |
+
if file_input.startswith("http"):
|
74 |
+
# Handle URLs
|
75 |
+
response = requests.get(file_input)
|
76 |
+
response.raise_for_status()
|
77 |
+
return response.content
|
78 |
+
else:
|
79 |
+
# Handle local file paths
|
80 |
+
with open(file_input, "rb") as f:
|
81 |
+
return f.read()
|
82 |
return file_input.read() if hasattr(file_input, 'read') else file_input
|
83 |
|
84 |
def ocr_pdf_url(self, pdf_url: str) -> str:
|
|
|
148 |
base64_url = f"data:image/jpeg;base64,{encoded_image}"
|
149 |
ocr_response = self._call_ocr_api({"type": "image_url", "image_url": base64_url})
|
150 |
markdown = self._extract_markdown(ocr_response)
|
151 |
+
|
152 |
chat_response = self._call_chat_complete(
|
153 |
model="pixtral-12b-latest",
|
154 |
messages=[{
|
|
|
164 |
response_format={"type": "json_object"},
|
165 |
temperature=0
|
166 |
)
|
167 |
+
|
168 |
# Ensure the response is a dictionary
|
169 |
response_content = chat_response.choices[0].message.content
|
170 |
if isinstance(response_content, list):
|
171 |
response_content = response_content[0] if response_content else "{}"
|
172 |
+
|
173 |
content = json.loads(response_content)
|
174 |
return self._format_structured_response(temp_path, content)
|
175 |
except Exception as e:
|
|
|
275 |
return demo
|
276 |
|
277 |
if __name__ == "__main__":
|
278 |
+
create_interface().launch(share=True, debug=True)
|