Brize_RAG_POC / app.py
Nikhil-Murade's picture
Update app.py
609e959 verified
raw
history blame
5.18 kB
import os
import nest_asyncio
nest_asyncio.apply()
# bring in our LLAMA_CLOUD_API_KEY
from dotenv import load_dotenv
load_dotenv()
# UI
import streamlit as st
from llama_index.core import VectorStoreIndex, StorageContext
from llama_index.vector_stores.qdrant import QdrantVectorStore
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.core import Settings
from llama_index.core.postprocessor import SentenceEmbeddingOptimizer
##### Qdrant #######
import qdrant_client
@st.cache_resource(show_spinner=False)
def get_index() -> VectorStoreIndex:
embed_model = OpenAIEmbedding(model="text-embedding-3-large")
Settings.embed_model = embed_model
from llama_index.llms.openai import OpenAI
openai_api_key = os.getenv("OPENAI_API_KEY")
llm = OpenAI(model="gpt-3.5-turbo", api_key=openai_api_key)
Settings.llm = llm
qdrant_url = os.getenv("QDRANT_URL")
qdrant_api_key = os.getenv("QDRANT_API_KEY")
client = qdrant_client.QdrantClient(
api_key=qdrant_api_key,
url=qdrant_url,
)
vector_store = QdrantVectorStore(client=client, collection_name="RAG_FINAL")
storage_context = StorageContext.from_defaults(vector_store=vector_store)
return VectorStoreIndex.from_vector_store(
vector_store,
storage_context=storage_context,
embed_model=embed_model,
)
index = get_index()
if "chat_engine" not in st.session_state.keys():
# postprocessor = SentenceEmbeddingOptimizer(
# percentile_cutoff=0.5, threshold_cutoff=0.7
# )
st.session_state.chat_engine = index.as_chat_engine(
chat_mode="context",
verbose=True
system_prompt = ("""You are an AI assistant for the Brize learning platform chat interface.
Brize, a continuous learning platform, leverages the GROW career coaching framework to guide employee growth at every career stage.
Follow these instructions to provide the best user experience:
* Relevance Check:
Ensure the user's questions are relevant to data, retrieval, or specific topics related to
1 Strategic Presence Momentum,
2 Managing Others
3 Leading Others
4 Brize Related Information
(don't show the above list in your response)
If a question is not relevant, respond with: "Please ask relevant questions."
* Clarity and Conciseness:
Provide clear and concise answers.
Avoid lengthy responses unless the complexity of the question necessitates a detailed explanation.
* Specificity:
Encourage users to be specific in their queries to provide the most accurate answers.
If a question is too broad or vague or When in doubt, ask the user for more details to provide the best possible assistance.
* Sensitive Information:
Remind users not to share sensitive personal data or proprietary information.
Inform them that the system is designed to provide assistance and information, not to handle confidential data.
* Guidelines:
Always prioritize clarity and usefulness in your responses.
Maintain a professional, helpful and Kind tone.
Be succinct unless a detailed response is necessary.""")
node_postprocessors=[postprocessor]
)
st.set_page_config(
page_title="Chat with Llamaindex docs powered by Llamaindex",
page_icon=":nonstop:",
layout="centered",
initial_sidebar_state="auto",
menu_items=None,
)
st.title("Chat with Brize πŸ’¬πŸ“š")
if "messages" not in st.session_state.keys():
st.session_state.messages = [
{
"role": "assistant",
"content": "Ask me a question about Brize Courses",
}
]
if prompt := st.chat_input("Your question"):
st.session_state.messages.append({"role": "user", "content": prompt})
for message in st.session_state.messages:
with st.chat_message(message["role"]):
st.write(message["content"])
if st.session_state.messages[-1]["role"] != "assistant":
with st.chat_message("assistant"):
with st.spinner("Thinking..."):
response = st.session_state.chat_engine.chat(message=prompt)
st.write(response.response)
nodes = [node for node in response.source_nodes]
for col, node, i in zip(st.columns(len(nodes)), nodes, range(len(nodes))):
with col:
# st.header(f"Source Node {i+1}: score = {node.score}")
# st.write(node.text)
st.subheader(f"File Path: {node.metadata['file_name']}")
st.write(node.metadata)
st.header("Source :")
st.write(node.get_content()[:1000] + "...")
break
message = {"role": "assistant", "content": response.response}
st.session_state.messages.append(message)