Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -64,13 +64,15 @@ def get_document_text(uploaded_files):
|
|
64 |
|
65 |
# Split text into chunks
|
66 |
def get_chunks(documents):
|
67 |
-
text_splitter = CharacterTextSplitter(separator="\n", chunk_size=
|
68 |
-
|
69 |
-
|
70 |
# Create vectorstore
|
|
|
71 |
def get_vectorstore(chunks):
|
72 |
embeddings = OpenAIEmbeddings()
|
73 |
-
|
|
|
74 |
|
75 |
# Create a conversational chain
|
76 |
def get_conversationchain(vectorstore):
|
@@ -78,7 +80,7 @@ def get_conversationchain(vectorstore):
|
|
78 |
memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)
|
79 |
conversation_chain = ConversationalRetrievalChain.from_llm(
|
80 |
llm=llm,
|
81 |
-
retriever=vectorstore.as_retriever(search_type="similarity",search_kwargs={"k":
|
82 |
condense_question_prompt=CUSTOM_QUESTION_PROMPT,
|
83 |
memory=memory,
|
84 |
combine_docs_chain_kwargs={'prompt': prompt}
|
@@ -100,6 +102,25 @@ def handle_question(question):
|
|
100 |
else:
|
101 |
st.markdown(f"**Bot:** {msg.content}")
|
102 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
103 |
# Main Streamlit app
|
104 |
def main():
|
105 |
st.set_page_config(page_title="Chat with Documents", page_icon="π")
|
|
|
64 |
|
65 |
# Split text into chunks
|
66 |
def get_chunks(documents):
|
67 |
+
text_splitter = CharacterTextSplitter(separator="\n", chunk_size=600, chunk_overlap=200, length_function=len)
|
68 |
+
chunks = [chunk for doc in documents for chunk in text_splitter.split_text(doc.page_content)]
|
69 |
+
return chunks
|
70 |
# Create vectorstore
|
71 |
+
|
72 |
def get_vectorstore(chunks):
|
73 |
embeddings = OpenAIEmbeddings()
|
74 |
+
vectorstore = FAISS.from_texts(texts=chunks, embedding=embeddings)
|
75 |
+
return vectorstore
|
76 |
|
77 |
# Create a conversational chain
|
78 |
def get_conversationchain(vectorstore):
|
|
|
80 |
memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)
|
81 |
conversation_chain = ConversationalRetrievalChain.from_llm(
|
82 |
llm=llm,
|
83 |
+
retriever=vectorstore.as_retriever(search_type="similarity",search_kwargs={"k": 10}),
|
84 |
condense_question_prompt=CUSTOM_QUESTION_PROMPT,
|
85 |
memory=memory,
|
86 |
combine_docs_chain_kwargs={'prompt': prompt}
|
|
|
102 |
else:
|
103 |
st.markdown(f"**Bot:** {msg.content}")
|
104 |
|
105 |
+
def handle_question(question):
|
106 |
+
if not st.session_state.conversation:
|
107 |
+
st.warning("Please process your documents first.")
|
108 |
+
return
|
109 |
+
|
110 |
+
# Get the response from the conversation chain
|
111 |
+
response = st.session_state.conversation({'question': question})
|
112 |
+
|
113 |
+
# Update chat history
|
114 |
+
st.session_state.chat_history = response['chat_history']
|
115 |
+
|
116 |
+
# Display chat history
|
117 |
+
for i, msg in enumerate(st.session_state.chat_history):
|
118 |
+
if i % 2 == 0:
|
119 |
+
st.markdown(f"**You:** {msg.content}")
|
120 |
+
else:
|
121 |
+
st.markdown(f"**Bot:** {msg.content}")
|
122 |
+
|
123 |
+
|
124 |
# Main Streamlit app
|
125 |
def main():
|
126 |
st.set_page_config(page_title="Chat with Documents", page_icon="π")
|