app.py

import os
import threading
import asyncio
import time
import discord
from dotenv import load_dotenv
from huggingface_hub import hf_hub_download
from llama_cpp import Llama

# Load environment variables (from Hugging Face Secrets and .env if available)
load_dotenv()
DISCORD_TOKEN = os.getenv("DISCORD_TOKEN")

if not DISCORD_TOKEN:
    raise ValueError("Discord bot token is missing. Set DISCORD_TOKEN in the environment variables.")

# Model details: Using the Q6_K_L variant from bartowski’s GGUF collection.
MODEL_REPO = "bartowski/agentica-org_DeepScaleR-1.5B-Preview-GGUF"
MODEL_FILENAME = "agentica-org_DeepScaleR-1.5B-Preview-Q6_K_L.gguf"
MODEL_PATH = f"./{MODEL_FILENAME}"

# Download the model file if it doesn't exist locally.
if not os.path.exists(MODEL_PATH):
    print("Model file not found locally. Downloading now...")
    MODEL_PATH = hf_hub_download(repo_id=MODEL_REPO, filename=MODEL_FILENAME)
    print(f"Model downloaded to {MODEL_PATH}")
else:
    print(f"Model found locally at {MODEL_PATH}")

# Initialize the model using llama-cpp-python.
print("Initializing model...")
llm = Llama(model_path=MODEL_PATH, n_threads=4)
print("Model initialization complete.")

# Define a function to generate responses using the model.
def generate_response(prompt):
    try:
        # Reduce max_tokens to 50 to speed up generation.
        output = llm(prompt=prompt, max_tokens=50, temperature=0.7, top_p=0.9, echo=False)
        response = output["text"]
        # Enforce bot identity: replace internal model name with "Shiv Yantra AI"
        response = response.replace("DeepScaleR", "Shiv Yantra AI")
        return response
    except Exception as e:
        print(f"Error in generate_response: {e}")
        return "Error processing your request."

# ----------------------------
# Discord Bot Setup
# ----------------------------
intents = discord.Intents.default()
intents.message_content = True  # Required to read message content
client = discord.Client(intents=intents)

@client.event
async def on_ready():
    print(f"Discord bot logged in as {client.user}")

@client.event
async def on_message(message):
    if message.author == client.user:
        return  # Skip messages from the bot itself
    user_input = message.content.strip()
    if user_input:
        try:
            # Run generate_response in a separate thread to avoid blocking.
            ai_response = await asyncio.to_thread(generate_response, user_input)
        except Exception as e:
            print(f"Error during generation in on_message: {e}")
            ai_response = "Error processing your request."
        await message.channel.send(ai_response)

def run_discord_bot():
    client.run(DISCORD_TOKEN)

# ----------------------------
# Start Services Concurrently
# ----------------------------
if __name__ == "__main__":
    print("Starting Discord bot...")
    threading.Thread(target=run_discord_bot, daemon=True).start()
    print("Discord bot started. Keeping main thread alive.")
    # Use a sleep loop instead of busy-waiting.
    while True:
        time.sleep(60)