Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -7,35 +7,24 @@ import fitz # PyMuPDF
|
|
7 |
import time
|
8 |
import os
|
9 |
|
10 |
-
#
|
11 |
-
GEMINI_API_KEY = st.secrets["GEMINI_API_KEY"]
|
12 |
-
GEMINI_MODEL = "gemini-2.0-flash" # Verify with Gemini API documentation
|
13 |
DOCUMENT_TYPES = ["Land Records", "Caste Certificates", "Property Registrations"]
|
14 |
|
15 |
-
#
|
16 |
-
class ExampleFile:
|
17 |
-
def __init__(self, file_path):
|
18 |
-
self.name = os.path.basename(file_path)
|
19 |
-
self.type = "image/jpeg"
|
20 |
-
with open(file_path, "rb") as f:
|
21 |
-
self._buffer = BytesIO(f.read())
|
22 |
-
|
23 |
-
def getvalue(self):
|
24 |
-
return self._buffer.getvalue()
|
25 |
-
|
26 |
-
# Initialize session state
|
27 |
def initialize_session_state():
|
28 |
if "processed_doc" not in st.session_state:
|
29 |
st.session_state.processed_doc = None
|
30 |
if "current_file" not in st.session_state:
|
31 |
st.session_state.current_file = None
|
|
|
|
|
32 |
|
33 |
# Reset session state
|
34 |
def reset_session_state():
|
35 |
st.session_state.processed_doc = None
|
36 |
st.session_state.current_file = None
|
37 |
|
38 |
-
# Encode file to base64
|
39 |
def encode_file(file):
|
40 |
try:
|
41 |
file_content = file.getvalue()
|
@@ -46,23 +35,27 @@ def encode_file(file):
|
|
46 |
|
47 |
# Query Gemini API
|
48 |
def query_gemini(prompt, image_b64):
|
|
|
|
|
|
|
|
|
49 |
try:
|
50 |
headers = {
|
51 |
-
"Authorization": f"Bearer {
|
52 |
"Content-Type": "application/json"
|
53 |
}
|
54 |
payload = {
|
55 |
-
"model":
|
56 |
"prompt": prompt,
|
57 |
"image": image_b64
|
58 |
}
|
59 |
response = requests.post(
|
60 |
-
"https://
|
61 |
headers=headers,
|
62 |
json=payload
|
63 |
)
|
64 |
response.raise_for_status()
|
65 |
-
return response.json().get("result", "")
|
66 |
except Exception as e:
|
67 |
st.error(f"Gemini API error: {str(e)}")
|
68 |
return None
|
@@ -71,7 +64,6 @@ def query_gemini(prompt, image_b64):
|
|
71 |
def process_document(file):
|
72 |
try:
|
73 |
with st.spinner("Analyzing document..."):
|
74 |
-
# Encode file to base64
|
75 |
image_b64 = encode_file(file)
|
76 |
if not image_b64:
|
77 |
return
|
@@ -84,12 +76,6 @@ def process_document(file):
|
|
84 |
st.session_state.doc_preview = Image.frombytes("RGB", [pix.width, pix.height], pix.samples)
|
85 |
elif file.type.startswith('image/'):
|
86 |
st.session_state.doc_preview = Image.open(BytesIO(file.getvalue()))
|
87 |
-
elif file.type == "text/plain":
|
88 |
-
text = file.getvalue().decode('utf-8')
|
89 |
-
img = Image.new('RGB', (800, 600), color=(73, 109, 137))
|
90 |
-
d = ImageDraw.Draw(img)
|
91 |
-
d.text((10, 10), text, fill=(255, 255, 0))
|
92 |
-
st.session_state.doc_preview = img
|
93 |
else:
|
94 |
st.error("Unsupported file format")
|
95 |
return
|
@@ -109,7 +95,7 @@ def process_document(file):
|
|
109 |
details = query_gemini(extract_prompt, image_b64)
|
110 |
|
111 |
# Verify authenticity
|
112 |
-
verify_prompt = "Analyze this document for signs of tampering or forgery.
|
113 |
verification = query_gemini(verify_prompt, image_b64)
|
114 |
|
115 |
st.session_state.processed_doc = {
|
@@ -134,24 +120,20 @@ def main():
|
|
134 |
# Sidebar Controls
|
135 |
with st.sidebar:
|
136 |
st.header("Document Controls")
|
137 |
-
|
|
|
|
|
|
|
138 |
uploaded_file = st.file_uploader(
|
139 |
"Upload Document",
|
140 |
-
type=["pdf", "jpg", "jpeg", "png"
|
141 |
key="uploaded_file"
|
142 |
)
|
143 |
|
144 |
-
if st.button("Use Example Image"):
|
145 |
-
example_file_path = "Caste-Certificate.jpg"
|
146 |
-
st.session_state.current_file = ExampleFile(example_file_path)
|
147 |
-
elif uploaded_file:
|
148 |
-
st.session_state.current_file = uploaded_file
|
149 |
-
else:
|
150 |
-
st.session_state.current_file = None
|
151 |
-
|
152 |
if st.button("Process Document"):
|
153 |
-
if
|
154 |
-
|
|
|
155 |
else:
|
156 |
st.error("Please select a document to process.")
|
157 |
|
@@ -181,7 +163,7 @@ def main():
|
|
181 |
st.subheader("Verification Analysis")
|
182 |
st.markdown(st.session_state.processed_doc['verification'])
|
183 |
else:
|
184 |
-
st.info("Please
|
185 |
|
186 |
if __name__ == "__main__":
|
187 |
-
main()
|
|
|
7 |
import time
|
8 |
import os
|
9 |
|
10 |
+
# Document categories
|
|
|
|
|
11 |
DOCUMENT_TYPES = ["Land Records", "Caste Certificates", "Property Registrations"]
|
12 |
|
13 |
+
# Set up session state
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
def initialize_session_state():
|
15 |
if "processed_doc" not in st.session_state:
|
16 |
st.session_state.processed_doc = None
|
17 |
if "current_file" not in st.session_state:
|
18 |
st.session_state.current_file = None
|
19 |
+
if "gemini_api_key" not in st.session_state:
|
20 |
+
st.session_state.gemini_api_key = None
|
21 |
|
22 |
# Reset session state
|
23 |
def reset_session_state():
|
24 |
st.session_state.processed_doc = None
|
25 |
st.session_state.current_file = None
|
26 |
|
27 |
+
# Encode file to base64
|
28 |
def encode_file(file):
|
29 |
try:
|
30 |
file_content = file.getvalue()
|
|
|
35 |
|
36 |
# Query Gemini API
|
37 |
def query_gemini(prompt, image_b64):
|
38 |
+
if not st.session_state.gemini_api_key:
|
39 |
+
st.error("API Key is missing. Please enter your Gemini API Key.")
|
40 |
+
return None
|
41 |
+
|
42 |
try:
|
43 |
headers = {
|
44 |
+
"Authorization": f"Bearer {st.session_state.gemini_api_key}",
|
45 |
"Content-Type": "application/json"
|
46 |
}
|
47 |
payload = {
|
48 |
+
"model": "gemini-1.5-pro", # Change based on the latest API model
|
49 |
"prompt": prompt,
|
50 |
"image": image_b64
|
51 |
}
|
52 |
response = requests.post(
|
53 |
+
"https://generativelanguage.googleapis.com/v1/models/gemini-1.5-pro:generateContent", # Corrected URL
|
54 |
headers=headers,
|
55 |
json=payload
|
56 |
)
|
57 |
response.raise_for_status()
|
58 |
+
return response.json().get("result", "No response received.")
|
59 |
except Exception as e:
|
60 |
st.error(f"Gemini API error: {str(e)}")
|
61 |
return None
|
|
|
64 |
def process_document(file):
|
65 |
try:
|
66 |
with st.spinner("Analyzing document..."):
|
|
|
67 |
image_b64 = encode_file(file)
|
68 |
if not image_b64:
|
69 |
return
|
|
|
76 |
st.session_state.doc_preview = Image.frombytes("RGB", [pix.width, pix.height], pix.samples)
|
77 |
elif file.type.startswith('image/'):
|
78 |
st.session_state.doc_preview = Image.open(BytesIO(file.getvalue()))
|
|
|
|
|
|
|
|
|
|
|
|
|
79 |
else:
|
80 |
st.error("Unsupported file format")
|
81 |
return
|
|
|
95 |
details = query_gemini(extract_prompt, image_b64)
|
96 |
|
97 |
# Verify authenticity
|
98 |
+
verify_prompt = "Analyze this document for signs of tampering or forgery. Provide verification results."
|
99 |
verification = query_gemini(verify_prompt, image_b64)
|
100 |
|
101 |
st.session_state.processed_doc = {
|
|
|
120 |
# Sidebar Controls
|
121 |
with st.sidebar:
|
122 |
st.header("Document Controls")
|
123 |
+
|
124 |
+
# API Key input
|
125 |
+
st.session_state.gemini_api_key = st.text_input("Enter Gemini API Key", type="password")
|
126 |
+
|
127 |
uploaded_file = st.file_uploader(
|
128 |
"Upload Document",
|
129 |
+
type=["pdf", "jpg", "jpeg", "png"],
|
130 |
key="uploaded_file"
|
131 |
)
|
132 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
133 |
if st.button("Process Document"):
|
134 |
+
if uploaded_file:
|
135 |
+
st.session_state.current_file = uploaded_file
|
136 |
+
process_document(uploaded_file)
|
137 |
else:
|
138 |
st.error("Please select a document to process.")
|
139 |
|
|
|
163 |
st.subheader("Verification Analysis")
|
164 |
st.markdown(st.session_state.processed_doc['verification'])
|
165 |
else:
|
166 |
+
st.info("Please upload a document and click 'Process Document' to start verification.")
|
167 |
|
168 |
if __name__ == "__main__":
|
169 |
+
main()
|