Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -13,25 +13,25 @@ TOKEN=os.getenv('HF_TOKEN')
|
|
| 13 |
subprocess.run(["huggingface-cli", "login", "--token", TOKEN, "--add-to-git-credential"])
|
| 14 |
st.sidebar.title("Welcome to MBAL Chatbot")
|
| 15 |
class PDFChatbot:
|
| 16 |
-
|
| 17 |
self.azure_client = openai.OpenAI()
|
| 18 |
self.conversation_history = []
|
| 19 |
self.pdf_content = ""
|
| 20 |
|
| 21 |
-
|
| 22 |
"""Split text into smaller chunks for better processing."""
|
| 23 |
db = FAISS.load_local("mbal_faiss_db", embeddings=HuggingFaceEmbeddings(model_name='bkai-foundation-models/vietnamese-bi-encoder'), allow_dangerous_deserialization=True)
|
| 24 |
relevant_chunks = db.similarity_search(user_question, k=3)
|
| 25 |
relevant_chunks = [chunk.page_content for chunk in relevant_chunks]
|
| 26 |
return "\n\n".join(relevant_chunks)
|
| 27 |
-
|
| 28 |
"""Generate response using Azure OpenAI based on PDF content and user question."""
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
{
|
| 36 |
"role": "system",
|
| 37 |
"content": """You are an experienced insurance agent assistant who helps customers understand their insurance policies and coverage details. Follow these guidelines:
|
|
@@ -53,64 +53,64 @@ class PDFChatbot:
|
|
| 53 |
{relevant_context}
|
| 54 |
Customer Question: {user_question}
|
| 55 |
Please provide a helpful response based on the insurance document content above."""
|
| 56 |
-
|
| 57 |
-
|
| 58 |
# Add conversation history
|
| 59 |
-
|
| 60 |
messages.append(msg)
|
| 61 |
-
|
| 62 |
-
|
| 63 |
model="gpt-4o-mini",
|
| 64 |
messages=messages,
|
| 65 |
max_tokens=300, #TODO
|
| 66 |
temperature=0.7
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
def main():
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
| 84 |
-
|
| 85 |
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
# Main chat interface
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
# Chat input
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
user_question,
|
| 107 |
st.session_state.chatbot.pdf_content
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
else:
|
| 115 |
# Show example questions
|
| 116 |
st.subheader("Các câu hỏi bạn có thể hỏi:")
|
|
@@ -138,4 +138,4 @@ def main():
|
|
| 138 |
""")
|
| 139 |
|
| 140 |
if __name__ == "__main__":
|
| 141 |
-
|
|
|
|
| 13 |
subprocess.run(["huggingface-cli", "login", "--token", TOKEN, "--add-to-git-credential"])
|
| 14 |
st.sidebar.title("Welcome to MBAL Chatbot")
|
| 15 |
class PDFChatbot:
|
| 16 |
+
def __init__(self):
|
| 17 |
self.azure_client = openai.OpenAI()
|
| 18 |
self.conversation_history = []
|
| 19 |
self.pdf_content = ""
|
| 20 |
|
| 21 |
+
def get_relevant_context(self, user_question: str) -> List[str]:
|
| 22 |
"""Split text into smaller chunks for better processing."""
|
| 23 |
db = FAISS.load_local("mbal_faiss_db", embeddings=HuggingFaceEmbeddings(model_name='bkai-foundation-models/vietnamese-bi-encoder'), allow_dangerous_deserialization=True)
|
| 24 |
relevant_chunks = db.similarity_search(user_question, k=3)
|
| 25 |
relevant_chunks = [chunk.page_content for chunk in relevant_chunks]
|
| 26 |
return "\n\n".join(relevant_chunks)
|
| 27 |
+
def chat_with_pdf(self, user_question: str, pdf_content: str) -> str:
|
| 28 |
"""Generate response using Azure OpenAI based on PDF content and user question."""
|
| 29 |
+
try:
|
| 30 |
+
# Split PDF content into chunks
|
| 31 |
+
# Get relevant context for the question
|
| 32 |
+
relevant_context = self.get_relevant_context(user_question)
|
| 33 |
+
# Prepare messages for the chat
|
| 34 |
+
messages = [
|
| 35 |
{
|
| 36 |
"role": "system",
|
| 37 |
"content": """You are an experienced insurance agent assistant who helps customers understand their insurance policies and coverage details. Follow these guidelines:
|
|
|
|
| 53 |
{relevant_context}
|
| 54 |
Customer Question: {user_question}
|
| 55 |
Please provide a helpful response based on the insurance document content above."""
|
| 56 |
+
}
|
| 57 |
+
]
|
| 58 |
# Add conversation history
|
| 59 |
+
for msg in self.conversation_history[-2:]: # Keep last 6 messages for context
|
| 60 |
messages.append(msg)
|
| 61 |
+
# Get response from Azure OpenAI
|
| 62 |
+
response = self.azure_client.chat.completions.create(
|
| 63 |
model="gpt-4o-mini",
|
| 64 |
messages=messages,
|
| 65 |
max_tokens=300, #TODO
|
| 66 |
temperature=0.7
|
| 67 |
+
)
|
| 68 |
+
bot_response = response.choices[0].message.content
|
| 69 |
+
# Update conversation history
|
| 70 |
+
self.conversation_history.append({"role": "user", "content": user_question})
|
| 71 |
+
self.conversation_history.append({"role": "assistant", "content": bot_response})
|
| 72 |
+
return bot_response
|
| 73 |
+
except Exception as e:
|
| 74 |
+
return f"Error generating response: {str(e)}"
|
| 75 |
def main():
|
| 76 |
+
# st.set_page_config(page_title="Insurance PDF Chatbot", page_icon="🛡️", layout="wide")
|
| 77 |
+
st.title("🛡️ Insurance Policy Assistant")
|
| 78 |
+
st.markdown("Upload your insurance policy PDF and ask questions about your coverage, claims, deductibles, and more!")
|
| 79 |
+
# Initialize chatbot
|
| 80 |
+
if 'chatbot' not in st.session_state:
|
| 81 |
+
st.session_state.chatbot = PDFChatbot()
|
| 82 |
+
st.session_state.pdf_processed = False
|
| 83 |
+
st.session_state.chat_history = []
|
| 84 |
+
# Sidebar for PDF upload and settings
|
| 85 |
|
| 86 |
+
# Clear conversation
|
| 87 |
+
if st.button("Xóa lịch sử"):
|
| 88 |
+
st.session_state.chatbot.conversation_history = []
|
| 89 |
+
st.session_state.chat_history = []
|
| 90 |
+
st.rerun()
|
| 91 |
# Main chat interface
|
| 92 |
+
if st.session_state.pdf_processed:
|
| 93 |
+
st.header("💬 Ask About Your Insurance Policy")
|
| 94 |
+
# Display chat history
|
| 95 |
+
for i, (question, answer) in enumerate(st.session_state.chat_history):
|
| 96 |
+
with st.container():
|
| 97 |
+
st.markdown(f"**You:** {question}")
|
| 98 |
+
st.markdown(f"**Insurance Assistant:** {answer}")
|
| 99 |
+
st.divider()
|
| 100 |
# Chat input
|
| 101 |
+
user_question = st.chat_input("Hãy đặt những câu hỏi về hợp đồng bảo hiểm cơ bản...")
|
| 102 |
+
if user_question:
|
| 103 |
+
with st.spinner("Analyzing your policy..."):
|
| 104 |
+
# Get response from chatbot
|
| 105 |
+
response = st.session_state.chatbot.chat_with_pdf(
|
| 106 |
user_question,
|
| 107 |
st.session_state.chatbot.pdf_content
|
| 108 |
+
)
|
| 109 |
+
# Add to chat history
|
| 110 |
+
st.session_state.chat_history.append((user_question, response))
|
| 111 |
+
# Display the new response
|
| 112 |
+
st.markdown(f"**You:** {user_question}")
|
| 113 |
+
st.markdown(f"**Insurance Assistant:** {response}")
|
| 114 |
else:
|
| 115 |
# Show example questions
|
| 116 |
st.subheader("Các câu hỏi bạn có thể hỏi:")
|
|
|
|
| 138 |
""")
|
| 139 |
|
| 140 |
if __name__ == "__main__":
|
| 141 |
+
main()
|