Update app.py

app.py

@@ -7,19 +7,19 @@ from dotenv import load_dotenv
 from huggingface_hub import hf_hub_download
 from llama_cpp import Llama
 
-# Load environment variables from Hugging Face Secrets (.env file is optional)
+# 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 quantized Q8 version from bartowski's repository.
+# Model details: Using the Q6_K_L variant (recommended) from bartowski's GGUF collection.
 MODEL_REPO = "bartowski/agentica-org_DeepScaleR-1.5B-Preview-GGUF"
-MODEL_FILENAME = "agentica-org_DeepScaleR-1.5B-Preview-Q8_0.gguf"
+MODEL_FILENAME = "agentica-org_DeepScaleR-1.5B-Preview-Q6_K_L.gguf"
 MODEL_PATH = f"./{MODEL_FILENAME}"
 
-# Download the model file if it does not exist locally.
+# 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)
@@ -28,18 +28,18 @@ else:
     print(f"Model found locally at {MODEL_PATH}")
 
 # Initialize the model using llama-cpp-python.
-# Adjust n_threads based on available CPU cores.
+# Adjust n_threads based on your available CPU cores.
 print("Initializing model...")
 llm = Llama(model_path=MODEL_PATH, n_threads=4)
 print("Model initialization complete.")
 
-# Define a function to generate a response.
+# Define a function to generate responses using the model.
 def generate_response(prompt):
     try:
-        # Generate response using the quantized model.
+        # Call the model: adjust max_tokens, temperature, and top_p as needed.
         output = llm(prompt=prompt, max_tokens=200, temperature=0.7, top_p=0.9, echo=False)
         response = output["text"]
-        # Enforce your bot identity.
+        # Replace any instance of the internal model name with your bot's identity.
         response = response.replace("DeepScaleR", "Shiv Yantra AI")
         return response
     except Exception as e:
@@ -50,7 +50,7 @@ def generate_response(prompt):
 # Discord Bot Setup
 # ----------------------------
 intents = discord.Intents.default()
-intents.message_content = True  # Ensure we can read message contents
+intents.message_content = True  # Required to read message contents
 client = discord.Client(intents=intents)
 
 @client.event
@@ -59,13 +59,13 @@ async def on_ready():
 
 @client.event
 async def on_message(message):
-    # Skip bot's own messages.
+    # Avoid replying to itself.
     if message.author == client.user:
         return
     user_input = message.content.strip()
     if user_input:
         try:
-            # Run the generate_response in a separate thread to avoid blocking.
+            # Run the synchronous generate_response function in a separate thread.
             ai_response = await asyncio.to_thread(generate_response, user_input)
         except Exception as e:
             print(f"Error during generation in on_message: {e}")
@@ -82,5 +82,6 @@ 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 to avoid busy-waiting.
     while True:
         time.sleep(60)