app.py

import streamlit as st
import requests
import json
import os
import datetime
from huggingface_hub import InferenceClient

# Constants
SPACE_URL = "https://qf6hn3tcwcf7pc7p.us-east-1.aws.endpoints.huggingface.cloud"
HF_API_KEY = os.getenv("HF_API_KEY")  # Retrieve the Hugging Face API key from system variables
EOS_TOKEN = "<|end|>"
CHAT_HISTORY_DIR = "chat_histories"
IMAGE_PATH = "DubsChat.png"
IMAGE_PATH_2 = "Reboot AI.png"
DUBS_PATH = "Dubs.png"

# Ensure the directory exists
try:
    os.makedirs(CHAT_HISTORY_DIR, exist_ok=True)
except OSError as e:
    st.error(f"Failed to create chat history directory: {e}")

# Streamlit Configurations
st.set_page_config(page_title="DUBSChat", page_icon=IMAGE_PATH, layout="wide")
st.logo(IMAGE_PATH_2,size="large")

# -------------------------
#    Chat Template
# -------------------------


def format_chat_template(history, user_input):
    """
    Formats the chat template by combining the chat history and user input.
    """
    CHAT_TEMPLATE =f"""
    <|system|> You are Dubs, a helpful assistant created by RebootAI.<|end|> \n {history} <|user|> \n {user_input}<|end|> \n <|assistant|> """
    return CHAT_TEMPLATE.format(history=history, user_input=user_input)

# -------------------------
#    Generate Chat History
# -------------------------
def format_chat_history(messages):
    """
    Converts the chat messages into a string compatible with the chat template.
    Ensures no duplicate <|assistant|> tokens in the history.
    """
    history = ""
    for message in messages:
        if message["role"] == "user":
            history += f"<|user|>{message['content']}<|end|>\n"
        elif message["role"] == "assistant":
            history += f"<|assistant|>{message['content']}<|end|>\n"
    return history.strip()  # Remove any trailing newlines

# -------------------------
#    Utility Functions
# -------------------------
def save_chat_history(session_name, messages):
    """
    Save the chat history to a JSON file.
    """
    file_path = os.path.join(CHAT_HISTORY_DIR, f"{session_name}.json")
    try:
        with open(file_path, "w") as f:
            json.dump(messages, f)
    except IOError as e:
        st.error(f"Failed to save chat history: {e}")


def load_chat_history(file_name):
    """
    Load the chat history from a JSON file.
    """
    file_path = os.path.join(CHAT_HISTORY_DIR, file_name)
    try:
        with open(file_path, "r") as f:
            return json.load(f)
    except (FileNotFoundError, json.JSONDecodeError):
        st.error("Failed to load chat history. Starting with a new session.")
        return []


def get_saved_sessions():
    """
    Get the list of saved chat sessions.
    """
    return [f.replace(".json", "") for f in os.listdir(CHAT_HISTORY_DIR) if f.endswith(".json")]

# -------------------------
#    Sidebar Configuration
# -------------------------
with st.sidebar:
    if st.button("New Chat"):
        st.session_state["messages"] = [
            {"role": "system", "content": "Your name is Dubs, a helpful assistant created by RebootAI."},
        ]
        st.session_state["session_name"] = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
        save_chat_history(st.session_state["session_name"], st.session_state["messages"])
        st.success("Chat reset and new session started.")
    
    saved_sessions = get_saved_sessions()
    if saved_sessions:
        selected_session = st.radio("Past Sessions:", saved_sessions)
        if st.button("Load Session"):
            st.session_state["messages"] = load_chat_history(f"{selected_session}.json")
            st.session_state["session_name"] = selected_session
            st.success(f"Loaded session: {selected_session}")
    else:
        st.write("No past sessions available.")

# -------------------------
# Chat History Initialization
# -------------------------
if "messages" not in st.session_state:
    st.session_state["messages"] = [
        {"role": "system", "content": "You are Dubs, a helpful assistant created by RebootAI"}
    ]
if "session_name" not in st.session_state:
    st.session_state["session_name"] = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S")

# -------------------------
#        Main Chat UI
# -------------------------
st.image(IMAGE_PATH, width=250)
st.markdown("Empowering you with a Sustainable AI")
st.markdown("DubsChat is currently best suited at assisting you with Coding Problems")

# Display existing chat history
for message in st.session_state["messages"]:
    if message["role"] == "user":
        st.chat_message("user").write(message["content"])
    elif message["role"] == "assistant":
        st.chat_message("assistant", avatar=DUBS_PATH).write(message["content"])

client = InferenceClient(SPACE_URL, token=HF_API_KEY)

# -------------------------
#    Streaming Logic
# -------------------------
def stream_response(prompt_text):
    """
    Stream text from the HF Inference Endpoint using the InferenceClient.
    Yields each partial chunk of text as it arrives.
    """
    gen_kwargs = {
        "max_new_tokens": 4096,
        "top_k": 30,
        "top_p": 0.9,
        "temperature": 0.2,
        "repetition_penalty": 1.02,
        "stop_sequences": ["<|end|>"]
    }

    stream = client.text_generation(prompt_text, stream=True, details=True, **gen_kwargs)

    for response in stream:
        if response.token.special:
            continue
        yield response.token.text

# -------------------------
#       User Input
# -------------------------
prompt = st.chat_input()

if prompt:
    # 1) Add the user's message to session state
    st.session_state["messages"].append({"role": "user", "content": prompt})
    st.chat_message("user").write(prompt)

    # 2) Format chat history and user input for the template
    chat_history = format_chat_history(st.session_state["messages"][:-1])  # Exclude the current user input
    model_input = format_chat_template(chat_history, prompt)

    # 3) Generate the assistant's response
    with st.spinner("Dubs is thinking... Woof Woof! 🐾"):
        msg = ""
        with st.chat_message("assistant", avatar=DUBS_PATH):
            response_stream = stream_response(model_input)
            msg = st.write_stream(response_stream)

    # 4) Add the assistant's response to session state
    st.session_state["messages"].append({"role": "assistant", "content": msg})

    # 5) Persist the updated chat history
    save_chat_history(st.session_state["session_name"], st.session_state["messages"])