Update app.py
Browse files
app.py
CHANGED
@@ -2,22 +2,10 @@ import sys
|
|
2 |
import os
|
3 |
import re
|
4 |
import time
|
|
|
5 |
import streamlit as st
|
6 |
import nltk
|
7 |
-
|
8 |
-
|
9 |
-
# Force NLTK to download 'punkt' into a virtual, in-memory location
|
10 |
-
try:
|
11 |
-
from nltk.data import load
|
12 |
-
print("Downloading 'punkt' tokenizer to memory...")
|
13 |
-
nltk.download("punkt")
|
14 |
-
load("tokenizers/punkt/english.pickle")
|
15 |
-
print("β
'punkt' successfully loaded into memory.")
|
16 |
-
except Exception as e:
|
17 |
-
print(f"Error loading 'punkt': {e}")
|
18 |
-
raise e
|
19 |
-
|
20 |
-
sys.path.append(os.path.abspath("."))
|
21 |
from langchain.chains import ConversationalRetrievalChain
|
22 |
from langchain.memory import ConversationBufferMemory
|
23 |
from langchain.llms import OpenAI
|
@@ -27,54 +15,52 @@ from langchain.embeddings import HuggingFaceEmbeddings
|
|
27 |
from langchain.text_splitter import NLTKTextSplitter
|
28 |
from patent_downloader import PatentDownloader
|
29 |
|
30 |
-
|
|
|
31 |
|
32 |
-
#
|
33 |
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
|
34 |
if not OPENAI_API_KEY:
|
35 |
-
st.error("Critical Error: OpenAI API key not found in
|
36 |
st.stop()
|
37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
def load_docs(document_path):
|
|
|
39 |
try:
|
40 |
-
loader = UnstructuredPDFLoader(
|
41 |
-
document_path,
|
42 |
-
mode="elements",
|
43 |
-
strategy="fast",
|
44 |
-
ocr_languages=None
|
45 |
-
)
|
46 |
documents = loader.load()
|
47 |
text_splitter = NLTKTextSplitter(chunk_size=1000)
|
48 |
return text_splitter.split_documents(documents)
|
49 |
except Exception as e:
|
50 |
-
st.error(f"Failed to
|
51 |
-
|
52 |
-
|
53 |
-
def already_indexed(vectordb, file_name):
|
54 |
-
indexed_sources = set(
|
55 |
-
x["source"] for x in vectordb.get(include=["metadatas"])["metadatas"]
|
56 |
-
)
|
57 |
-
return file_name in indexed_sources
|
58 |
|
59 |
-
def load_chain(file_name=None):
|
60 |
-
loaded_patent = st.session_state.get("LOADED_PATENT")
|
61 |
|
62 |
-
|
63 |
-
|
64 |
-
|
|
|
65 |
)
|
66 |
-
if loaded_patent == file_name or already_indexed(vectordb, file_name):
|
67 |
-
st.write("β
Already indexed.")
|
68 |
-
else:
|
69 |
-
vectordb.delete_collection()
|
70 |
-
docs = load_docs(file_name)
|
71 |
-
st.write("π Number of Documents: ", len(docs))
|
72 |
-
|
73 |
-
vectordb = Chroma.from_documents(
|
74 |
-
docs, HuggingFaceEmbeddings(), persist_directory=PERSISTED_DIRECTORY
|
75 |
-
)
|
76 |
-
vectordb.persist()
|
77 |
-
st.session_state["LOADED_PATENT"] = file_name
|
78 |
|
79 |
memory = ConversationBufferMemory(
|
80 |
memory_key="chat_history",
|
@@ -82,6 +68,7 @@ def load_chain(file_name=None):
|
|
82 |
input_key="question",
|
83 |
output_key="answer",
|
84 |
)
|
|
|
85 |
return ConversationalRetrievalChain.from_llm(
|
86 |
OpenAI(temperature=0, openai_api_key=OPENAI_API_KEY),
|
87 |
vectordb.as_retriever(search_kwargs={"k": 3}),
|
@@ -89,20 +76,8 @@ def load_chain(file_name=None):
|
|
89 |
memory=memory,
|
90 |
)
|
91 |
|
92 |
-
def extract_patent_number(url):
|
93 |
-
pattern = r"/patent/([A-Z]{2}\d+)"
|
94 |
-
match = re.search(pattern, url)
|
95 |
-
return match.group(1) if match else None
|
96 |
-
|
97 |
-
def download_pdf(patent_number):
|
98 |
-
try:
|
99 |
-
patent_downloader = PatentDownloader(verbose=True)
|
100 |
-
output_path = patent_downloader.download(patents=patent_number, output_path="/tmp")
|
101 |
-
return output_path[0]
|
102 |
-
except Exception as e:
|
103 |
-
st.error(f"Failed to download patent PDF: {e}")
|
104 |
-
st.stop()
|
105 |
|
|
|
106 |
if __name__ == "__main__":
|
107 |
st.set_page_config(
|
108 |
page_title="Patent Chat: Google Patents Chat Demo",
|
@@ -110,8 +85,10 @@ if __name__ == "__main__":
|
|
110 |
layout="wide",
|
111 |
initial_sidebar_state="expanded",
|
112 |
)
|
|
|
113 |
st.header("π Patent Chat: Google Patents Chat Demo")
|
114 |
|
|
|
115 |
patent_link = st.text_input("Enter Google Patent Link:", key="PATENT_LINK")
|
116 |
|
117 |
if not patent_link:
|
@@ -123,48 +100,64 @@ if __name__ == "__main__":
|
|
123 |
st.error("Invalid patent link format. Please provide a valid Google patent link.")
|
124 |
st.stop()
|
125 |
|
126 |
-
st.write(f"Patent
|
127 |
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
pdf_path = download_pdf(patent_number)
|
134 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
135 |
|
136 |
-
|
137 |
-
chain = load_chain(pdf_path)
|
138 |
st.success("π Document successfully loaded! You can now start asking questions.")
|
139 |
|
|
|
140 |
if "messages" not in st.session_state:
|
141 |
st.session_state["messages"] = [
|
142 |
{"role": "assistant", "content": "Hello! How can I assist you with this patent?"}
|
143 |
]
|
144 |
|
|
|
145 |
for message in st.session_state.messages:
|
146 |
with st.chat_message(message["role"]):
|
147 |
st.markdown(message["content"])
|
148 |
|
|
|
149 |
if user_input := st.chat_input("What is your question?"):
|
150 |
st.session_state.messages.append({"role": "user", "content": user_input})
|
|
|
151 |
with st.chat_message("user"):
|
152 |
st.markdown(user_input)
|
153 |
|
154 |
with st.chat_message("assistant"):
|
155 |
message_placeholder = st.empty()
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
time.sleep(0.05)
|
164 |
-
message_placeholder.markdown(full_response + "β")
|
165 |
-
except Exception as e:
|
166 |
-
full_response = f"An error occurred: {e}"
|
167 |
-
finally:
|
168 |
-
message_placeholder.markdown(full_response)
|
169 |
|
170 |
st.session_state.messages.append({"role": "assistant", "content": full_response})
|
|
|
2 |
import os
|
3 |
import re
|
4 |
import time
|
5 |
+
import tempfile
|
6 |
import streamlit as st
|
7 |
import nltk
|
8 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
from langchain.chains import ConversationalRetrievalChain
|
10 |
from langchain.memory import ConversationBufferMemory
|
11 |
from langchain.llms import OpenAI
|
|
|
15 |
from langchain.text_splitter import NLTKTextSplitter
|
16 |
from patent_downloader import PatentDownloader
|
17 |
|
18 |
+
# Download NLTK resources
|
19 |
+
nltk.download("punkt", quiet=True)
|
20 |
|
21 |
+
#fetch API key
|
22 |
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
|
23 |
if not OPENAI_API_KEY:
|
24 |
+
st.error("Critical Error: OpenAI API key not found in environment variables. Please configure it.")
|
25 |
st.stop()
|
26 |
|
27 |
+
|
28 |
+
def extract_patent_number(url):
|
29 |
+
"""Extracts patent number from a Google patent link."""
|
30 |
+
pattern = r"/patent/([A-Z]{2}\d+)"
|
31 |
+
match = re.search(pattern, url)
|
32 |
+
return match.group(1) if match else None
|
33 |
+
|
34 |
+
|
35 |
+
def download_pdf(patent_number):
|
36 |
+
"""Downloads patent PDF using a temporary directory."""
|
37 |
+
try:
|
38 |
+
with tempfile.TemporaryDirectory() as temp_dir:
|
39 |
+
patent_downloader = PatentDownloader(verbose=True)
|
40 |
+
output_path = patent_downloader.download(patents=patent_number, output_path=temp_dir)
|
41 |
+
return output_path[0]
|
42 |
+
except Exception as e:
|
43 |
+
st.error(f"Failed to download patent PDF: {e}")
|
44 |
+
return None
|
45 |
+
|
46 |
+
|
47 |
def load_docs(document_path):
|
48 |
+
"""Loads and splits PDF documents into chunks."""
|
49 |
try:
|
50 |
+
loader = UnstructuredPDFLoader(document_path)
|
|
|
|
|
|
|
|
|
|
|
51 |
documents = loader.load()
|
52 |
text_splitter = NLTKTextSplitter(chunk_size=1000)
|
53 |
return text_splitter.split_documents(documents)
|
54 |
except Exception as e:
|
55 |
+
st.error(f"Failed to process PDF: {e}")
|
56 |
+
return []
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
|
|
|
|
|
58 |
|
59 |
+
def load_chain(docs):
|
60 |
+
"""Creates a conversational retrieval chain using in-memory ChromaDB."""
|
61 |
+
vectordb = Chroma.from_documents(
|
62 |
+
docs, HuggingFaceEmbeddings(), persist_directory=None
|
63 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
64 |
|
65 |
memory = ConversationBufferMemory(
|
66 |
memory_key="chat_history",
|
|
|
68 |
input_key="question",
|
69 |
output_key="answer",
|
70 |
)
|
71 |
+
|
72 |
return ConversationalRetrievalChain.from_llm(
|
73 |
OpenAI(temperature=0, openai_api_key=OPENAI_API_KEY),
|
74 |
vectordb.as_retriever(search_kwargs={"k": 3}),
|
|
|
76 |
memory=memory,
|
77 |
)
|
78 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
79 |
|
80 |
+
# Streamlit UI
|
81 |
if __name__ == "__main__":
|
82 |
st.set_page_config(
|
83 |
page_title="Patent Chat: Google Patents Chat Demo",
|
|
|
85 |
layout="wide",
|
86 |
initial_sidebar_state="expanded",
|
87 |
)
|
88 |
+
|
89 |
st.header("π Patent Chat: Google Patents Chat Demo")
|
90 |
|
91 |
+
# Input for Google Patent Link
|
92 |
patent_link = st.text_input("Enter Google Patent Link:", key="PATENT_LINK")
|
93 |
|
94 |
if not patent_link:
|
|
|
100 |
st.error("Invalid patent link format. Please provide a valid Google patent link.")
|
101 |
st.stop()
|
102 |
|
103 |
+
st.write(f"π Patent Number: **{patent_number}**")
|
104 |
|
105 |
+
# Download or Upload PDF
|
106 |
+
st.write("π₯ Downloading patent PDF...")
|
107 |
+
pdf_path = None
|
108 |
+
|
109 |
+
try:
|
110 |
pdf_path = download_pdf(patent_number)
|
111 |
+
except Exception:
|
112 |
+
st.error("Automatic download failed. Please upload the PDF manually below.")
|
113 |
+
|
114 |
+
if not pdf_path:
|
115 |
+
uploaded_file = st.file_uploader("Upload the patent PDF file:", type="pdf")
|
116 |
+
if uploaded_file:
|
117 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp_file:
|
118 |
+
tmp_file.write(uploaded_file.read())
|
119 |
+
pdf_path = tmp_file.name
|
120 |
+
st.success("β
PDF successfully uploaded.")
|
121 |
+
else:
|
122 |
+
st.stop()
|
123 |
+
|
124 |
+
# Load and Process PDF
|
125 |
+
st.write("π Processing document...")
|
126 |
+
docs = load_docs(pdf_path)
|
127 |
+
|
128 |
+
if not docs:
|
129 |
+
st.error("No content found in the PDF. Exiting...")
|
130 |
+
st.stop()
|
131 |
|
132 |
+
chain = load_chain(docs)
|
|
|
133 |
st.success("π Document successfully loaded! You can now start asking questions.")
|
134 |
|
135 |
+
# Initialize chat history
|
136 |
if "messages" not in st.session_state:
|
137 |
st.session_state["messages"] = [
|
138 |
{"role": "assistant", "content": "Hello! How can I assist you with this patent?"}
|
139 |
]
|
140 |
|
141 |
+
# Display chat history
|
142 |
for message in st.session_state.messages:
|
143 |
with st.chat_message(message["role"]):
|
144 |
st.markdown(message["content"])
|
145 |
|
146 |
+
# Handle User Input
|
147 |
if user_input := st.chat_input("What is your question?"):
|
148 |
st.session_state.messages.append({"role": "user", "content": user_input})
|
149 |
+
|
150 |
with st.chat_message("user"):
|
151 |
st.markdown(user_input)
|
152 |
|
153 |
with st.chat_message("assistant"):
|
154 |
message_placeholder = st.empty()
|
155 |
+
with st.spinner("Generating response..."):
|
156 |
+
try:
|
157 |
+
assistant_response = chain({"question": user_input})
|
158 |
+
full_response = assistant_response.get("answer", "I'm sorry, I couldn't generate a response.")
|
159 |
+
except Exception as e:
|
160 |
+
full_response = f"An error occurred: {e}"
|
161 |
+
message_placeholder.markdown(full_response)
|
|
|
|
|
|
|
|
|
|
|
|
|
162 |
|
163 |
st.session_state.messages.append({"role": "assistant", "content": full_response})
|