app.py

import streamlit as st
import torch
from transformers import AutoTokenizer, AutoModel, AutoModelForCausalLM
import faiss
import numpy as np
import os
import pickle
import warnings
warnings.filterwarnings("ignore", category=FutureWarning, module="transformers")

# Model combinations with speed ratings and estimated time savings
MODEL_COMBINATIONS = {
    "Fastest (30 seconds)": {
        "embedding": "sentence-transformers/all-MiniLM-L6-v2",
        "generation": "distilgpt2",
        "free": True,
        "time_saved": "2.5 minutes"
    },
    "Balanced (1 minute)": {
        "embedding": "sentence-transformers/all-MiniLM-L12-v2",
        "generation": "facebook/opt-350m",
        "free": True,
        "time_saved": "2 minutes"
    },
    "High Quality (2 minutes)": {
        "embedding": "sentence-transformers/all-mpnet-base-v2",
        "generation": "gpt2",
        "free": True,
        "time_saved": "1 minute"
    },
    "Premium Speed (15 seconds)": {
        "embedding": "sentence-transformers/all-MiniLM-L6-v2",
        "generation": "microsoft/phi-1_5",
        "free": False,
        "time_saved": "2.75 minutes"
    },
    "Premium Quality (1.5 minutes)": {
        "embedding": "openai-embedding-ada-002",
        "generation": "meta-llama/Llama-2-7b-chat-hf",
        "free": False,
        "time_saved": "1.5 minutes"
    }
}

@st.cache_resource
def load_models(model_combination):
    try:
        embedding_tokenizer = AutoTokenizer.from_pretrained(MODEL_COMBINATIONS[model_combination]['embedding'])
        embedding_model = AutoModel.from_pretrained(MODEL_COMBINATIONS[model_combination]['embedding'])
        generation_tokenizer = AutoTokenizer.from_pretrained(MODEL_COMBINATIONS[model_combination]['generation'])
        generation_model = AutoModelForCausalLM.from_pretrained(MODEL_COMBINATIONS[model_combination]['generation'])
        return embedding_tokenizer, embedding_model, generation_tokenizer, generation_model
    except Exception as e:
        st.error(f"Error loading models: {str(e)}")
        return None, None, None, None

@st.cache_data
def load_and_process_text(file_path):
    try:
        with open(file_path, 'r', encoding='utf-8') as file:
            text = file.read()
        chunks = [text[i:i+512] for i in range(0, len(text), 512)]
        return chunks
    except Exception as e:
        st.error(f"Error loading text file: {str(e)}")
        return []

@st.cache_data
def create_embeddings(chunks, _embedding_model):
    embeddings = []
    for chunk in chunks:
        inputs = tokenizer(chunk, return_tensors="pt", padding=True, truncation=True, max_length=512)
        with torch.no_grad():
            outputs = _embedding_model(**inputs)
        embeddings.append(outputs.last_hidden_state.mean(dim=1).squeeze().numpy())
    return np.array(embeddings)

@st.cache_resource
def create_faiss_index(embeddings):
    index = faiss.IndexFlatL2(embeddings.shape[1])
    index.add(embeddings)
    return index

def generate_response(query, tokenizer, generation_model, embedding_model, index, chunks):
    inputs = tokenizer(query, return_tensors="pt", padding=True, truncation=True, max_length=512)
    with torch.no_grad():
        outputs = embedding_model(**inputs)
    query_embedding = outputs.last_hidden_state.mean(dim=1).squeeze().numpy()
    
    k = 3
    _, I = index.search(query_embedding.reshape(1, -1), k)
    
    context = " ".join([chunks[i] for i in I[0]])
    
    prompt = f"As the Muse of A.R. Ammons, respond to this query: {query}\nContext: {context}\nMuse:"
    
    input_ids = tokenizer.encode(prompt, return_tensors="pt")
    output = generation_model.generate(input_ids, max_new_tokens=100, num_return_sequences=1, temperature=0.7)
    response = tokenizer.decode(output[0], skip_special_tokens=True)
    
    muse_response = response.split("Muse:")[-1].strip()
    return muse_response

def save_data(chunks, embeddings, index):
    with open('chunks.pkl', 'wb') as f:
        pickle.dump(chunks, f)
    np.save('embeddings.npy', embeddings)
    faiss.write_index(index, 'faiss_index.bin')

def load_data():
    if os.path.exists('chunks.pkl') and os.path.exists('embeddings.npy') and os.path.exists('faiss_index.bin'):
        with open('chunks.pkl', 'rb') as f:
            chunks = pickle.load(f)
        embeddings = np.load('embeddings.npy')
        index = faiss.read_index('faiss_index.bin')
        return chunks, embeddings, index
    return None, None, None

# Streamlit UI
st.set_page_config(page_title="A.R. Ammons' Muse Chatbot", page_icon="🎭")

st.title("A.R. Ammons' Muse Chatbot 🎭")
st.markdown("""
    <style>
    .big-font {
        font-size:20px !important;
        font-weight: bold;
    }
    </style>
    """, unsafe_allow_html=True)
st.markdown('<p class="big-font">Chat with the Muse of A.R. Ammons. Ask questions or discuss poetry!</p>', unsafe_allow_html=True)

# Model selection
if 'model_combination' not in st.session_state:
    st.session_state.model_combination = "Fastest (30 seconds)"

# Create a list of model options, with non-free models at the end
free_models = [k for k, v in MODEL_COMBINATIONS.items() if v['free']]
non_free_models = [k for k, v in MODEL_COMBINATIONS.items() if not v['free']]
all_models = free_models + non_free_models

# Custom CSS to grey out non-free options
st.markdown("""
<style>
    .stSelectbox div[role="option"][aria-selected="false"]:nth-last-child(-n+2) {
        color: grey !important;
    }
</style>
""", unsafe_allow_html=True)

selected_model = st.selectbox(
    "Choose a model combination:",
    all_models,
    index=all_models.index(st.session_state.model_combination),
    format_func=lambda x: f"{x} {'(Not Free)' if not MODEL_COMBINATIONS[x]['free'] else ''}"
)

# Prevent selection of non-free models
if not MODEL_COMBINATIONS[selected_model]['free']:
    st.warning("Premium models are not available in the free version.")
    st.stop()

st.session_state.model_combination = selected_model

st.info(f"Potential time saved compared to slowest option: {MODEL_COMBINATIONS[selected_model]['time_saved']}")

if st.button("Load Selected Models"):
    with st.spinner("Loading models and data..."):
        embedding_tokenizer, embedding_model, generation_tokenizer, generation_model = load_models(st.session_state.model_combination)
        chunks = load_and_process_text('ammons_muse.txt')
        embeddings = create_embeddings(chunks, embedding_model)
        index = create_faiss_index(embeddings)
    
    st.session_state.models_loaded = True
    st.success("Models loaded successfully!")

if 'models_loaded' not in st.session_state or not st.session_state.models_loaded:
    st.warning("Please load the models before chatting.")
    st.stop()

# Initialize chat history
if 'messages' not in st.session_state:
    st.session_state.messages = []

# Display chat messages from history on app rerun
for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.markdown(message["content"])

# React to user input
if prompt := st.chat_input("What would you like to ask the Muse?"):
    st.chat_message("user").markdown(prompt)
    st.session_state.messages.append({"role": "user", "content": prompt})

    with st.spinner("The Muse is contemplating..."):
        try:
            response = generate_response(prompt, tokenizer, generation_model, embedding_model, index, chunks)
        except Exception as e:
            response = f"I apologize, but I encountered an error: {str(e)}"
    
    with st.chat_message("assistant"):
        st.markdown(response)
    st.session_state.messages.append({"role": "assistant", "content": response})

# Add a button to clear chat history
if st.button("Clear Chat History"):
    st.session_state.messages = []
    st.experimental_rerun()

# Add a footer
st.markdown("---")
st.markdown("*Powered by the spirit of A.R. Ammons and the magic of AI*")