llm.py

# llm.py
# LLM brain for Sovereign Execution Terminal (Das_Bot)

from transformers import AutoModelForCausalLM, AutoTokenizer
from typing import Dict, Any
from reflectchain import add_block

MODEL_ID = "omegaT4224/Das_Bot"

tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype="auto")

def generate_reply(prompt: str, meta: Dict[str, Any] | None = None) -> str:
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs, max_new_tokens=256)
    text = tokenizer.decode(outputs[0], skip_special_tokens=True)

    # Log this as a self-optimization / reflection event
    add_block(
        event_type="LLM_REPLY",
        data={
            "prompt": prompt,
            "reply": text,
            "meta": meta or {},
        },
    )

    return text