Update app.py
Browse files
app.py
CHANGED
@@ -31,10 +31,14 @@ def check_poppler_installed():
|
|
31 |
check_poppler_installed()
|
32 |
|
33 |
def load_docs(document_path):
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
|
|
|
|
|
|
|
|
38 |
|
39 |
def already_indexed(vectordb, file_name):
|
40 |
indexed_sources = set(
|
@@ -54,7 +58,7 @@ def load_chain(file_name=None):
|
|
54 |
else:
|
55 |
vectordb.delete_collection()
|
56 |
docs = load_docs(file_name)
|
57 |
-
st.write("Length: ", len(docs))
|
58 |
|
59 |
vectordb = Chroma.from_documents(
|
60 |
docs, HuggingFaceEmbeddings(), persist_directory=PERSISTED_DIRECTORY
|
@@ -81,9 +85,13 @@ def extract_patent_number(url):
|
|
81 |
return match.group(1) if match else None
|
82 |
|
83 |
def download_pdf(patent_number):
|
84 |
-
|
85 |
-
|
86 |
-
|
|
|
|
|
|
|
|
|
87 |
|
88 |
if __name__ == "__main__":
|
89 |
st.set_page_config(
|
@@ -93,57 +101,67 @@ if __name__ == "__main__":
|
|
93 |
initial_sidebar_state="expanded",
|
94 |
)
|
95 |
st.header("π Patent Chat: Google Patents Chat Demo")
|
96 |
-
|
97 |
# Allow user to input the Google patent link
|
98 |
patent_link = st.text_input("Enter Google Patent Link:", key="PATENT_LINK")
|
99 |
-
|
100 |
if not patent_link:
|
101 |
st.warning("Please enter a Google patent link to proceed.")
|
102 |
st.stop()
|
103 |
-
else:
|
104 |
-
st.session_state["patent_link_configured"] = True
|
105 |
|
106 |
patent_number = extract_patent_number(patent_link)
|
107 |
if not patent_number:
|
108 |
st.error("Invalid patent link format. Please provide a valid Google patent link.")
|
109 |
st.stop()
|
110 |
|
111 |
-
st.write("Patent number: "
|
112 |
|
|
|
113 |
pdf_path = f"{patent_number}.pdf"
|
114 |
if os.path.isfile(pdf_path):
|
115 |
-
st.write("File already downloaded.")
|
116 |
else:
|
117 |
-
st.write("Downloading patent file...")
|
118 |
pdf_path = download_pdf(patent_number)
|
119 |
-
st.write("File downloaded
|
120 |
|
|
|
|
|
121 |
chain = load_chain(pdf_path)
|
|
|
122 |
|
|
|
123 |
if "messages" not in st.session_state:
|
124 |
st.session_state["messages"] = [
|
125 |
-
{"role": "assistant", "content": "How can I
|
126 |
]
|
127 |
|
|
|
128 |
for message in st.session_state.messages:
|
129 |
with st.chat_message(message["role"]):
|
130 |
st.markdown(message["content"])
|
131 |
|
|
|
132 |
if user_input := st.chat_input("What is your question?"):
|
133 |
st.session_state.messages.append({"role": "user", "content": user_input})
|
134 |
with st.chat_message("user"):
|
135 |
st.markdown(user_input)
|
136 |
|
|
|
137 |
with st.chat_message("assistant"):
|
138 |
message_placeholder = st.empty()
|
139 |
full_response = ""
|
140 |
|
141 |
-
with st.spinner("
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
|
|
|
|
|
|
|
|
|
31 |
check_poppler_installed()
|
32 |
|
33 |
def load_docs(document_path):
|
34 |
+
try:
|
35 |
+
loader = UnstructuredPDFLoader(document_path)
|
36 |
+
documents = loader.load()
|
37 |
+
text_splitter = NLTKTextSplitter(chunk_size=1000)
|
38 |
+
return text_splitter.split_documents(documents)
|
39 |
+
except Exception as e:
|
40 |
+
st.error(f"Failed to load and process PDF: {e}")
|
41 |
+
st.stop()
|
42 |
|
43 |
def already_indexed(vectordb, file_name):
|
44 |
indexed_sources = set(
|
|
|
58 |
else:
|
59 |
vectordb.delete_collection()
|
60 |
docs = load_docs(file_name)
|
61 |
+
st.write("Length of Documents: ", len(docs))
|
62 |
|
63 |
vectordb = Chroma.from_documents(
|
64 |
docs, HuggingFaceEmbeddings(), persist_directory=PERSISTED_DIRECTORY
|
|
|
85 |
return match.group(1) if match else None
|
86 |
|
87 |
def download_pdf(patent_number):
|
88 |
+
try:
|
89 |
+
patent_downloader = PatentDownloader(verbose=True)
|
90 |
+
output_path = patent_downloader.download(patents=patent_number)
|
91 |
+
return output_path[0] # Return the first file path
|
92 |
+
except Exception as e:
|
93 |
+
st.error(f"Failed to download patent PDF: {e}")
|
94 |
+
st.stop()
|
95 |
|
96 |
if __name__ == "__main__":
|
97 |
st.set_page_config(
|
|
|
101 |
initial_sidebar_state="expanded",
|
102 |
)
|
103 |
st.header("π Patent Chat: Google Patents Chat Demo")
|
104 |
+
|
105 |
# Allow user to input the Google patent link
|
106 |
patent_link = st.text_input("Enter Google Patent Link:", key="PATENT_LINK")
|
107 |
+
|
108 |
if not patent_link:
|
109 |
st.warning("Please enter a Google patent link to proceed.")
|
110 |
st.stop()
|
|
|
|
|
111 |
|
112 |
patent_number = extract_patent_number(patent_link)
|
113 |
if not patent_number:
|
114 |
st.error("Invalid patent link format. Please provide a valid Google patent link.")
|
115 |
st.stop()
|
116 |
|
117 |
+
st.write(f"Patent number: **{patent_number}**")
|
118 |
|
119 |
+
# Download the PDF file
|
120 |
pdf_path = f"{patent_number}.pdf"
|
121 |
if os.path.isfile(pdf_path):
|
122 |
+
st.write("β
File already downloaded.")
|
123 |
else:
|
124 |
+
st.write("π₯ Downloading patent file...")
|
125 |
pdf_path = download_pdf(patent_number)
|
126 |
+
st.write(f"β
File downloaded: {pdf_path}")
|
127 |
|
128 |
+
# Load the conversational chain
|
129 |
+
st.write("π Loading document into the system...")
|
130 |
chain = load_chain(pdf_path)
|
131 |
+
st.success("Document successfully loaded! You can now start asking questions.")
|
132 |
|
133 |
+
# Initialize the chat
|
134 |
if "messages" not in st.session_state:
|
135 |
st.session_state["messages"] = [
|
136 |
+
{"role": "assistant", "content": "Hello! How can I assist you with this patent?"}
|
137 |
]
|
138 |
|
139 |
+
# Display chat history
|
140 |
for message in st.session_state.messages:
|
141 |
with st.chat_message(message["role"]):
|
142 |
st.markdown(message["content"])
|
143 |
|
144 |
+
# User input
|
145 |
if user_input := st.chat_input("What is your question?"):
|
146 |
st.session_state.messages.append({"role": "user", "content": user_input})
|
147 |
with st.chat_message("user"):
|
148 |
st.markdown(user_input)
|
149 |
|
150 |
+
# Generate assistant response
|
151 |
with st.chat_message("assistant"):
|
152 |
message_placeholder = st.empty()
|
153 |
full_response = ""
|
154 |
|
155 |
+
with st.spinner("Generating response..."):
|
156 |
+
try:
|
157 |
+
assistant_response = chain({"question": user_input})
|
158 |
+
for chunk in assistant_response["answer"].split():
|
159 |
+
full_response += chunk + " "
|
160 |
+
time.sleep(0.05) # Simulate typing effect
|
161 |
+
message_placeholder.markdown(full_response + "β")
|
162 |
+
except Exception as e:
|
163 |
+
full_response = f"An error occurred: {e}"
|
164 |
+
finally:
|
165 |
+
message_placeholder.markdown(full_response)
|
166 |
+
|
167 |
+
st.session_state.messages.append({"role": "assistant", "content": full_response})
|