app.py

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 💬📚")

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)