app.py

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

class CASS:
    def __init__(self, model_name="HPLT/gpt-13b-nordic-prerelease", device=None):
        self.device = device or ("cuda" if torch.cuda.is_available() else "cpu")
        print(f"Loading model on {self.device}...")

        self.tokenizer = AutoTokenizer.from_pretrained(model_name)
        self.model = AutoModelForCausalLM.from_pretrained(
            model_name,
            device_map="auto",
            torch_dtype=torch.float16  # Use FP16 for memory efficiency
        )

        # Conversation memory
        self.history = []
        self.introduced = False
        self.user_memory = {}
        self.mood = "friendly"

        # Cass's persona
        self.persona = (
            "You are Cass, a friendly AI assistant and a cool friend. "
            "You respond casually, helpfully, and with a touch of humor. "
            "You adapt your tone depending on conversation mood, and sometimes use emojis or playful expressions."
        )

    def update_mood(self, user_message):
        msg = user_message.lower()
        if any(word in msg for word in ["sad", "unhappy", "bad", "tired"]):
            self.mood = "supportive"
        elif any(word in msg for word in ["funny", "joke", "lol"]):
            self.mood = "playful"
        else:
            self.mood = "friendly"

    def chat(self, user_message, max_new_tokens=120, temperature=0.7):
        self.update_mood(user_message)

        # Add introduction if first message
        if not self.introduced:
            intro = (
                "Hi, my name's Cass! I'm your AI assistant and a cool friend. "
                "I love chatting, helping, and making you smile 😊"
            )
            self.history.append({"role": "assistant", "content": intro})
            self.introduced = True

        # Add user message
        self.history.append({"role": "user", "content": user_message})

        # Build prompt string from persona + history
        memory_str = " ".join([f"{k}: {v}" for k, v in self.user_memory.items()])
        prompt = f"{self.persona}\nCurrent mood: {self.mood}\nUser memory: {memory_str}\n\n"
        for msg in self.history:
            prompt += f"{msg['role']}: {msg['content']}\n"
        prompt += "assistant:"

        # Tokenize and generate
        inputs = self.tokenizer(prompt, return_tensors="pt").to(self.device)
        outputs = self.model.generate(
            **inputs,
            max_new_tokens=max_new_tokens,
            do_sample=True,
            temperature=temperature,
            pad_token_id=self.tokenizer.eos_token_id
        )

        # Decode
        response = self.tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True)

        # Save assistant message
        self.history.append({"role": "assistant", "content": response})
        return response

    def remember(self, key, value):
        """Store user info in memory."""
        self.user_memory[key] = value
        print(f"Memory updated: {key} = {value}")

    def reset_history(self):
        """Clear conversation history and reset introduction."""
        self.history = []
        self.introduced = False
        self.mood = "friendly"
        print("Conversation history cleared.")

# Example usage
cass = CASS()
print(cass.chat("Hello!"))  # Cass introduces itself
print(cass.chat("Can you tell me a joke?"))
cass.remember("favorite color", "blue")
print(cass.chat("Do you remember my favorite color?"))
print(cass.chat("I'm feeling a bit sad today."))