Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,136 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
import streamlit as st
|
3 |
+
from langchain_community.embeddings import HuggingFaceBgeEmbeddings
|
4 |
+
from pinecone import Pinecone
|
5 |
+
from langchain.vectorstores import Pinecone as LangchainPinecone
|
6 |
+
from langchain_groq import ChatGroq
|
7 |
+
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder, PromptTemplate
|
8 |
+
from langchain.chains import create_retrieval_chain
|
9 |
+
from langchain.chains.combine_documents import create_stuff_documents_chain
|
10 |
+
from langchain_community.chat_message_histories import StreamlitChatMessageHistory
|
11 |
+
from langchain_core.chat_history import BaseChatMessageHistory
|
12 |
+
from langchain_core.runnables.history import RunnableWithMessageHistory
|
13 |
+
from langchain.chains import create_history_aware_retriever
|
14 |
+
import time
|
15 |
+
import os
|
16 |
+
|
17 |
+
|
18 |
+
# Embedding setup
|
19 |
+
model_name = "BAAI/bge-small-en"
|
20 |
+
model_kwargs = {"device": "cpu"}
|
21 |
+
encode_kwargs = {"normalize_embeddings": True}
|
22 |
+
embeddings = HuggingFaceBgeEmbeddings(
|
23 |
+
model_name=model_name, model_kwargs=model_kwargs, encode_kwargs=encode_kwargs
|
24 |
+
)
|
25 |
+
|
26 |
+
# Pinecone setup
|
27 |
+
pc = Pinecone(api_key="pcsk_5yLpy7_7DWbGm2s2HTf1NCbo4zFB8KLEZFLT54q3poTUoEFMbf1B9ShUZqpsT7EPnE3Pjw")
|
28 |
+
text_field = "text"
|
29 |
+
index_name = "contentengine"
|
30 |
+
index = pc.Index(index_name)
|
31 |
+
vectorstore = LangchainPinecone(index, embeddings.embed_query, text_field)
|
32 |
+
|
33 |
+
# Retriever setup
|
34 |
+
retriever = vectorstore.as_retriever(
|
35 |
+
search_type="similarity_score_threshold",
|
36 |
+
search_kwargs={"k": 1, "score_threshold": 0.5},
|
37 |
+
)
|
38 |
+
|
39 |
+
|
40 |
+
|
41 |
+
llm = ChatGroq(model="llama3-8b-8192", api_key='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', max_tokens=4096)
|
42 |
+
|
43 |
+
|
44 |
+
# Retriever prompt setup
|
45 |
+
retriever_prompt = """
|
46 |
+
Given a chat history and the latest user question which might reference context in the chat history,
|
47 |
+
formulate a standalone question which can be understood without the chat history.
|
48 |
+
Do NOT answer the question, just reformulate it if needed and otherwise return it as is.
|
49 |
+
|
50 |
+
Chat History:
|
51 |
+
{chat_history}
|
52 |
+
|
53 |
+
User Question: {input}
|
54 |
+
|
55 |
+
Standalone question:
|
56 |
+
"""
|
57 |
+
contextualize_q_prompt = ChatPromptTemplate.from_messages([
|
58 |
+
("system", retriever_prompt),
|
59 |
+
MessagesPlaceholder(variable_name="chat_history"),
|
60 |
+
("human", "{input}"),
|
61 |
+
])
|
62 |
+
|
63 |
+
history_aware_retriever = create_history_aware_retriever(llm, retriever, contextualize_q_prompt)
|
64 |
+
|
65 |
+
from langchain_core.prompts import PromptTemplate
|
66 |
+
|
67 |
+
template = """
|
68 |
+
Context: This Content Engine is designed to analyze and compare key information across multiple Form 10-K filings for major companies, specifically Alphabet, Tesla, and Uber. The system uses Retrieval-Augmented Generation (RAG) to retrieve and summarize insights, highlight differences, and answer user queries on various financial and operational topics, such as risk factors, revenue, and business models.
|
69 |
+
|
70 |
+
Chat History: {chat_history}
|
71 |
+
Context: {context}
|
72 |
+
Human: {input}
|
73 |
+
|
74 |
+
Answer:
|
75 |
+
"""
|
76 |
+
|
77 |
+
# Define the PromptTemplate with specified input variables
|
78 |
+
custom_rag_prompt = PromptTemplate(template=template, input_variables=["chat_history", "context", "input"])
|
79 |
+
|
80 |
+
question_answering_chain = create_stuff_documents_chain(llm, custom_rag_prompt)
|
81 |
+
rag_chain = create_retrieval_chain(history_aware_retriever, question_answering_chain)
|
82 |
+
|
83 |
+
# ======================================================= Streamlit UI =======================================================
|
84 |
+
|
85 |
+
st.title("Chat with Content Engine")
|
86 |
+
|
87 |
+
# Initialize chat history
|
88 |
+
if "chat_history" not in st.session_state:
|
89 |
+
st.session_state.chat_history = StreamlitChatMessageHistory(key="chat_messages")
|
90 |
+
|
91 |
+
# Message history setup
|
92 |
+
def get_chat_history():
|
93 |
+
return st.session_state.chat_history
|
94 |
+
|
95 |
+
# Conversational_rag_chain to use the Streamlit chat history
|
96 |
+
conversational_rag_chain = RunnableWithMessageHistory(
|
97 |
+
rag_chain,
|
98 |
+
get_chat_history,
|
99 |
+
input_messages_key="input",
|
100 |
+
history_messages_key="chat_history",
|
101 |
+
output_messages_key="answer"
|
102 |
+
)
|
103 |
+
|
104 |
+
# Function to interact with the chatbot
|
105 |
+
def chat_with_bot(query: str) -> str:
|
106 |
+
result = conversational_rag_chain.invoke(
|
107 |
+
{"input": query},
|
108 |
+
config={
|
109 |
+
"configurable": {"session_id": "streamlit_session"}
|
110 |
+
},
|
111 |
+
)
|
112 |
+
return result["answer"]
|
113 |
+
|
114 |
+
# Display chat messages from history
|
115 |
+
for message in st.session_state.chat_history.messages:
|
116 |
+
with st.chat_message(message.type):
|
117 |
+
st.markdown(message.content)
|
118 |
+
|
119 |
+
# Accept user input
|
120 |
+
if user_input := st.chat_input("Enter your question here..."):
|
121 |
+
|
122 |
+
# Display user message in chat message container
|
123 |
+
with st.chat_message("human"):
|
124 |
+
st.markdown(user_input)
|
125 |
+
|
126 |
+
# Display assistant response in chat message container
|
127 |
+
with st.chat_message("ai"):
|
128 |
+
with st.spinner("Thinking..."):
|
129 |
+
response = chat_with_bot(user_input)
|
130 |
+
message_placeholder = st.empty()
|
131 |
+
full_response = "⚠️ **_Reminder: Please double-check information._** \n\n"
|
132 |
+
for chunk in response:
|
133 |
+
full_response += chunk
|
134 |
+
time.sleep(0.01)
|
135 |
+
message_placeholder.markdown(full_response + ":white_circle:", unsafe_allow_html=True)
|
136 |
+
|