app.py

import streamlit as st
from llama_index.core import VectorStoreIndex
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.core.node_parser import SentenceSplitter
from llama_index.core.ingestion import IngestionPipeline
import chromadb
from llama_index.vector_stores.chroma import ChromaVectorStore
from llama_index.llms.ollama import Ollama

from llama_index.llms.huggingface import HuggingFaceLLM

from llama_index.core import Settings
import time

# Ustawienia strony
st.title("Aplikacja z LlamaIndex")

db = chromadb.PersistentClient(path="./abc")
chroma_collection = db.get_or_create_collection("pomoc_ukrainie")
vector_store = ChromaVectorStore(chroma_collection=chroma_collection)
embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5", device="cpu")

# Utwórz indeks
index = VectorStoreIndex.from_vector_store(vector_store, embed_model=embed_model)

# Utwórz silnik zapytań

# Load the correct tokenizer and LLM
from transformers import AutoTokenizer

llm = HuggingFaceLLM(
    model_name="PrunaAI/eryk-mazus-polka-1.1b-bnb-4bit-smashed", # Mały model 1.3B
    tokenizer=AutoTokenizer.from_pretrained("eryk-mazus/polka-1.1b"),
    trust_remote_code=True,
    device_map="auto"
    # model_kwargs={
    #     "quantization_config": quantization_config,
    # }
)

#Query Engine
query_engine = index.as_query_engine(
    llm=llm,
    response_mode='compact')




# app gui
# Store LLM generated responses
if "messages" not in st.session_state.keys():
    st.session_state.messages = [{"role": "assistant", "content": "Zadaj mi pytanie..."}]

# Display chat messages
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.write(message["content"])

# User-provided prompt
if input := st.chat_input():
    st.session_state.messages.append({"role": "user", "content": input})
    with st.chat_message("user"):
        st.write(input)

# Generate a new response if last message is not from assistant
if st.session_state.messages[-1]["role"] != "assistant":
    with st.chat_message("assistant"):
        with st.spinner("Czekaj, odpowiedź jest generowana.."):
            start_time = time.time() # Start timing
            response = query_engine.query(input)
            end_time = time.time() # End timing
            generation_time = end_time - start_time
            
            # Zbuduj treść wiadomości z odpowiedzią i score
            content = str(response.response)  # Upewnij się, że response jest stringiem
            if hasattr(response, 'source_nodes') and response.source_nodes:  # Sprawdź, czy source_nodes istnieje
                # Dodaj score pierwszego węzła (jeśli istnieje)
                content += f"\nScore: {response.source_nodes[0].score:.4f}"  # Dodaj score

            # Add generation time
            content += f"\nCzas generowania: {generation_time:.2f} sekund"
            
            st.write(content)  # Wyświetl całą treść w Streamlit

    message = {"role": "assistant", "content": content}  # Zapisz całą treść w wiadomości
    st.session_state.messages.append(message)