modules/config_settings_public.py

# FILE: modules/config_settings_public.py (Hugging Face Demo)

import os
import sys
import logging
from huggingface_hub import hf_hub_download

logger = logging.getLogger("ZOTHEOS_Config")
if not logger.handlers:
    handler = logging.StreamHandler(sys.stdout)
    formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - [%(funcName)s] - %(message)s')
    handler.setFormatter(formatter)
    logger.addHandler(handler)
    logger.setLevel(logging.INFO)

# --- ✅ WEB-OPTIMIZED MODEL SOURCES ---
# These models are smaller and faster, perfect for a CPU-based web demo.
MODEL_DEFINITIONS = {
    "mistral": {
        "repo_id": "TheBloke/Mistral-7B-Instruct-v0.2-GGUF",
        "filename": "mistral-7b-instruct-v0.2.Q2_K.gguf" # Using a smaller Q2_K quantization for speed
    },
    "gemma": {
        "repo_id": "TheBloke/gemma-2b-it-GGUF",
        "filename": "gemma-2b-it.Q4_K_M.gguf" # Using the 2B parameter version
    },
    "qwen": {
        "repo_id": "TheBloke/Qwen1.5-1.8B-Chat-GGUF",
        "filename": "qwen1.5-1.8b-chat.Q4_K_M.gguf" # Using the 1.8B parameter version
    }
}

MODEL_PATHS = {}

# This logic will only run when the code is on Hugging Face Spaces.
logger.info("✅✅✅ RUNNING IN WEB DEMO MODE (Hugging Face Space) ✅✅✅")
N_GPU_LAYERS_FALLBACK = 0  # Force CPU-only mode for Hugging Face free tier

for name, model_info in MODEL_DEFINITIONS.items():
    logger.info(f"Downloading model for demo: {name} from {model_info['repo_id']}")
    try:
        MODEL_PATHS[name] = hf_hub_download(repo_id=model_info["repo_id"], filename=model_info["filename"])
        logger.info(f"✅ Successfully downloaded {name}")
    except Exception as e:
        logger.error(f"❌ FAILED to download {name}: {e}")
        raise e

# --- ✅ WEB-OPTIMIZED MODEL PARAMETERS ---
MODEL_SPECIFIC_PARAMS = {
    "_default": { 
        "n_gpu_layers": N_GPU_LAYERS_FALLBACK, # Ensures CPU usage
        "n_ctx": 4096,           # Smaller context window for lower RAM usage
        "n_batch": 512,          # Standard batch size for CPU
        "verbose": True
    },
    "mistral": { "chat_format": "mistral-instruct" },
    "gemma": { "chat_format": "gemma" },
    "qwen": { "chat_format": "chatml" }
}

# --- ✅ TIER INFERENCE PRESETS ---
INFERENCE_PRESETS = {
    "balanced": {"temperature": 0.7, "top_p": 0.9, "max_tokens": 1024, "repeat_penalty": 1.1 },
    "precise": {"temperature": 0.2, "top_p": 0.7, "top_k": 20, "max_tokens": 1536, "repeat_penalty": 1.05 },
    "creative": {"temperature": 0.9, "top_p": 0.95, "top_k": 60, "max_tokens": 1024, "repeat_penalty": 1.15 }
}
DEFAULT_INFERENCE_PRESET = "balanced"

# --- ✅ TIER MODEL ROLES AND PROMPTS ---
DEFAULT_SYSTEM_PROMPT = "You are ZOTHEOS, an ethical AI developed to help humanity. Be clear, respectful, and helpful. Respond only in English."
MODEL_ROLES = {"mistral": "analyst", "gemma": "humanist", "qwen": "skeptic"}
MODEL_ROLE_SYSTEM_PROMPTS = {
    "analyst": "You are an impartial analyst. Provide structured, logical insights. Break down complex topics into clear, understandable points. Respond only in English.",
    "humanist": "You are an empathetic and values-driven AI. Focus on the emotional, ethical, and societal impact. Consider the human element above all. Respond only in English.",
    "skeptic": "You are a respectful but rigorous skeptic. Challenge the user's premise, identify potential risks, question assumptions, and explore alternative interpretations. Respond only in English.",
}

logger.info("✅ Hugging Face Demo Configuration Loaded Successfully.")