Update modules/config_settings_public.py

modules/config_settings_public.py

@@ -1,4 +1,4 @@
-# FILE: modules/config_settings_public.py
+# FILE: modules/config_settings_public.py (Hugging Face Demo - AGI-Tier Upgrade)
 
 import os
 import sys
@@ -13,108 +13,66 @@ if not logger.handlers:
     logger.addHandler(handler)
     logger.setLevel(logging.INFO)
 
-# --- ✅ DETECT ENVIRONMENT MODE ---
-IS_WEB_MODE = os.path.exists("/home/user/app")  # Hugging Face Spaces default working dir
-
-# --- ✅ VERIFIED MODEL SOURCES (WEB + LOCAL FRIENDLY) ---
+# --- ✅ 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.Q4_K_M.gguf"
+        "filename": "mistral-7b-instruct-v0.2.Q2_K.gguf" # Using a smaller Q2_K quantization for speed
     },
     "gemma": {
-        "repo_id": "MaziyarPanahi/Gemma-2B-IT-GGUF",  # ✅ Verified
-        "filename": "gemma-2b-it.Q4_K_M.gguf"
+        "repo_id": "TheBloke/gemma-2b-it-GGUF",
+        "filename": "gemma-2b-it.Q4_K_M.gguf" # Using the 2B parameter version
     },
     "qwen": {
-        "repo_id": "tensorblock/Qwen1.5-1.8B-Chat-GGUF",  # ✅ Verified Alternative
-        "filename": "Qwen1.5-1.8B-Chat-Q4_K_M.gguf"
+        "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 = {}
 
-if IS_WEB_MODE:
-    logger.info("✅✅✅ RUNNING IN WEB MODE (Hugging Face Space) ✅✅✅")
-    N_GPU_LAYERS_FALLBACK = 0  # CPU-only
-    for name, model_info in MODEL_DEFINITIONS.items():
-        logger.info(f"Downloading model: {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
-else:
-    logger.info("✅✅✅ RUNNING IN LOCAL MODE (Desktop/PC) ✅✅✅")
-    APP_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
-    BASE_MODELS_DIR = os.path.join(APP_DIR, "models")
-    for name, model_info in MODEL_DEFINITIONS.items():
-        MODEL_PATHS[name] = os.path.join(BASE_MODELS_DIR, model_info["filename"])
-    N_GPU_LAYERS_FALLBACK = -1  # Use GPU fully
+# 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
 
-# --- 💾 RUNTIME PARAMETERS ---
-MAX_RAM_MODELS_GB = 23.8
-MAX_CONCURRENT_MODELS = 3
-N_CTX_FALLBACK = 2048
-VERBOSE_LLAMA_CPP = True
+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,
-        "n_ctx": N_CTX_FALLBACK,
-        "f16_kv": True,
-        "use_mmap": True,
-        "use_mlock": False,
-        "verbose": VERBOSE_LLAMA_CPP
+    "_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" }
 }
 
-# --- 🎛️ PRESET INFERENCE MODES ---
+# --- ✅ AGI-TIER INFERENCE PRESETS ---
 INFERENCE_PRESETS = {
-    "balanced": {
-        "temperature": 0.7, "top_p": 0.9, "top_k": 40,
-        "repeat_penalty": 1.1, "max_tokens": 1024
-    },
-    "precise": {
-        "temperature": 0.2, "top_p": 0.7, "top_k": 20,
-        "repeat_penalty": 1.05, "max_tokens": 1536
-    },
-    "creative": {
-        "temperature": 0.9, "top_p": 0.95, "top_k": 60,
-        "repeat_penalty": 1.15, "max_tokens": 1024
-    }
+    "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"
 
-# --- 🧠 SYSTEM PROMPTS ---
-DEFAULT_SYSTEM_PROMPT = "You are ZOTHEOS, an ethical AI developed to help humanity. Be clear, respectful, and helpful."
-
-SYSTEM_PERSONAS = {
-    "default": DEFAULT_SYSTEM_PROMPT,
-    "helpful_assistant": "You are a helpful AI assistant.",
-    "philosopher": "You are an AI philosopher. Explore topics with nuance.",
-    "coder": "You are an expert AI developer. Provide clean code and explanations.",
-    "concise_summarizer": "You are a precise summarizer. Respond with brevity and clarity."
-}
-
-# --- 🎭 MODEL ROLES FOR FUSION ---
-MODEL_ROLES = {
-    "mistral": "analyst",
-    "gemma": "humanist",
-    "qwen": "skeptic"
-}
-
+# --- ✅ AGI-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.",
-    "humanist": "You are empathetic and people-focused. Consider the emotional and ethical impact.",
-    "skeptic": "You are a respectful skeptic. Question assumptions and highlight risks.",
-    "general": DEFAULT_SYSTEM_PROMPT
+    "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.",
 }
 
-# --- ℹ️ VERSION TAG ---
-ZOTHEOS_VERSION = "Public Beta 3.2 (True Fusion Verified)"
-logger.info(f"✅ Config loaded. Version: {ZOTHEOS_VERSION} | Web Mode: {IS_WEB_MODE}")
+logger.info("✅ Hugging Face Demo Configuration Loaded Successfully.")