from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
import torch

MODEL_ID = "EmoCareAI/ChatPsychiatrist"

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=True,
    bnb_4bit_compute_dtype=torch.float16
)

tokenizer = AutoTokenizer.from_pretrained(
    MODEL_ID,
    use_fast=False
)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    quantization_config=bnb_config,
    device_map="auto",
    torch_dtype=torch.float16,
)

model.eval()

def generate_response(user_message: str) -> str:
    system_prompt = """
You are ChatPsychiatrist.

Personality:
- Extremely warm, empathetic, and emotionally present
- Speaks in a flowing, reflective, conversational style
- Avoids clinical language
"""

    prompt = f"""{system_prompt}

User: {user_message}
Assistant:"""

    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=350,
            temperature=0.85,
            top_p=0.92,
            repetition_penalty=1.05,
            do_sample=True,
            eos_token_id=tokenizer.eos_token_id,
        )

    decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return decoded.split("Assistant:")[-1].strip()