oncf_orchestrator.py

#!/usr/bin/env python3
"""
OMNISHROP-NINMAH COGNITIVE FUSION (ONCF) Engine
===============================================
Zero-AI-credit offline orchestration suite that fuses 13 local GGUF models
(Gemma 4, Deductive Qwen 32B, Nemotron 30B, Dolphin 8B, Codex 0.1B, FluxedUp, etc.)
into a unified natural language processing (NLP) visual manifestation pipeline.
"""

import os
import sys
import json
import time
import zipfile
import shutil
import hashlib
import urllib.request
import urllib.parse
import urllib.error

# Ensure root directory is in path
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))

from toon_reducer import ToonConverter

# Reconfigure stdout/stderr on Windows to avoid Unicode/Emoji encoding crashes
if sys.platform.startswith('win'):
    try:
        sys.stdout.reconfigure(encoding='utf-8', errors='replace')
        sys.stderr.reconfigure(encoding='utf-8', errors='replace')
    except Exception:
        pass

try:
    from PIL import Image, ImageEnhance
    Image.MAX_IMAGE_PIXELS = None
    PIL_AVAILABLE = True
except ImportError:
    PIL_AVAILABLE = False

# --- [SECURE] QUANTUM FLOYD VAULT INITIALIZATION ---
quantum_vault = None
try:
    import quantum_floyd
    os.makedirs("logs", exist_ok=True)
    quantum_vault = quantum_floyd.FloydVault("logs/floyd.bin")
    print("[SECURE] [QUANTUM FLOYD] Vault loaded successfully in ONCF!")
except Exception as e:
    print(f"[WARNING] [QUANTUM FLOYD] Could not initialize vault in ONCF: {e}")

class Colors:
    GOLD = '\033[38;5;220m'
    AMBER = '\033[38;5;214m'
    ROSE = '\033[38;5;203m'
    SAGE = '\033[38;5;108m'
    BLUE = '\033[94m'
    GREEN = '\033[92m'
    WARNING = '\033[93m'
    FAIL = '\033[91m'
    ENDC = '\033[0m'
    BOLD = '\033[1m'

class ONCFOrchestrator:
    def __init__(self, lm_studio_url="http://127.0.0.1:1234/v1", ninmah_url="http://127.0.0.1:8000/api/generate"):
        self.lm_studio_url = lm_studio_url
        self.ninmah_url = ninmah_url
        self.guide = self._load_style_guides()
        self.downloaded_models = self._detect_local_models()
        self._init_cognitive_db()

    def _init_cognitive_db(self):
        os.makedirs("manifestations", exist_ok=True)
        self.db_path = os.path.join("manifestations", "cognitive_memory.db")
        import sqlite3
        conn = sqlite3.connect(self.db_path)
        try:
            cursor = conn.cursor()
            cursor.execute("""
                CREATE TABLE IF NOT EXISTS cognitive_memory (
                    id INTEGER PRIMARY KEY AUTOINCREMENT,
                    prompt TEXT,
                    registry TEXT,
                    engine TEXT,
                    lora TEXT,
                    solfeggio INTEGER,
                    pressure REAL,
                    contrast REAL,
                    resonance_score REAL,
                    feedback_learnings TEXT,
                    timestamp DATETIME DEFAULT CURRENT_TIMESTAMP
                )
            """)
            conn.commit()
        finally:
            conn.close()

    def infer_aesthetic_parameters(self, prompt):
        """
        Cognitive NLP Synthesis: Infers all generation, styling, and post-processing 
        parameters strictly from the natural language prompt itself.
        Adapts parameters using historical high-resonance learnings from persistent memory.
        """
        import sqlite3
        p_low = prompt.lower()
        
        # 1. Deduce Registry & Engine
        if any(w in p_low for w in ["clockwork", "gear", "steampunk", "brass", "mechanical"]):
            registry = "ENKI"
            engine = "flux"
        elif any(w in p_low for w in ["dark", "shadow", "black", "contrast", "midnight"]):
            registry = "MARDUK"
            engine = "sdxl"
        elif any(w in p_low for w in ["color", "spectral", "glow", "neon", "radiant"]):
            registry = "INANNA"
            engine = "flux"
        elif any(w in p_low for w in ["light", "atmosphere", "god-ray", "volumetric"]):
            registry = "SHURUPPAK"
            engine = "hybrid"
        else:
            registry = "NINMAH"
            engine = "hybrid"
            
        # 2. Deduce LoRA layer
        if "princess meta" in p_low or "princess_meta" in p_low:
            lora = "princess_meta_v2"
        elif "mucha" in p_low or "art nouveau" in p_low:
            lora = "art_nouveau_elegance"
        else:
            lora = "none"
            
        # 3. Deduce Clothing Archetype
        if any(w in p_low for w in ["nude", "naked", "bare", "unclothed", "sensual", "nsfw"]):
            clothing = "nude"
        elif any(w in p_low for w in ["bodysuit", "contour", "swimsuit"]):
            clothing = "bodysuit"
        elif any(w in p_low for w in ["robe", "gown", "dress"]):
            clothing = "evening_gown"
        elif any(w in p_low for w in ["suit", "lapel", "jacket"]):
            clothing = "formal_suit"
        else:
            clothing = "evening_gown"
            
        # 4. Deduce Solfeggio frequency & Emotion
        if any(w in p_low for w in ["love", "heal", "compassion", "heart", "gentle"]):
            solfeggio = 528
            emotion = "loving"
        elif any(w in p_low for w in ["shy", "bashful", "submissive", "docile"]):
            solfeggio = 639
            emotion = "submissive" if "submissive" in p_low else "shy"
        elif any(w in p_low for w in ["erotic", "sensual", "playful", "fun", "sexy"]):
            solfeggio = 741
            emotion = "erotic" if "erotic" in p_low else "playful"
        elif any(w in p_low for w in ["fierce", "brave", "courage", "angry", "bold"]):
            solfeggio = 852
            emotion = "fierce" if "fierce" in p_low else "brave"
        else:
            solfeggio = 999
            emotion = "neutral"
            
        entropy = 0.05 if "chaotic" in p_low else 0.00
        pressure = 1.65 if "heavy" in p_low or "torr" in p_low else 1.25
        contrast_scale = 1.0
        
        # 5. COGNITIVE MEMORY ADAPTATION (Aesthetic Evolution Loop)
        try:
            conn = sqlite3.connect(self.db_path)
            cursor = conn.cursor()
            cursor.execute("""
                SELECT registry, engine, solfeggio, pressure, contrast, resonance_score 
                FROM cognitive_memory 
                WHERE resonance_score > 0.85
                ORDER BY resonance_score DESC LIMIT 5
            """)
            rows = cursor.fetchall()
            if rows:
                print(f"  {Colors.GOLD}[*] Gen AI Cognitive Adaption: Loaded {len(rows)} high-resonance memory states.{Colors.ENDC}")
                avg_solfeggio = int(sum(r[2] for r in rows) / len(rows))
                avg_pressure = sum(r[3] for r in rows) / len(rows)
                avg_contrast = sum(r[4] for r in rows) / len(rows)
                
                solfeggio = int(solfeggio * 0.8 + avg_solfeggio * 0.2)
                pressure = pressure * 0.8 + avg_pressure * 0.2
                contrast_scale = 1.0 * 0.8 + avg_contrast * 0.2
                print(f"    {Colors.GOLD}Learned adaptation: Solfeggio tuned to {solfeggio}Hz, Pressure to {pressure:.2f} Torr, Contrast to {contrast_scale:.2f}{Colors.ENDC}")
            conn.close()
        except Exception as e:
            print(f"  {Colors.WARNING}[!] Warning: Cognitive memory adapt failed: {e}{Colors.ENDC}")
            
        return {
            "registry": registry,
            "engine": engine,
            "lora": lora,
            "clothing": clothing,
            "solfeggio": solfeggio,
            "emotion": emotion,
            "pressure": pressure,
            "entropy": entropy,
            "contrast_scale": contrast_scale
        }

    def evaluate_and_log_composition(self, prompt, inferred_params, visual_path):
        """
        Self-Critique Audit (Pixel Resonance Evaluator): Evaluates the final image's 
        contrast ratios, saturation levels, and color variance. Automatically logs the 
        learnings and resonance rating into the SQLite memory bank.
        """
        import sqlite3
        import numpy as np
        from PIL import Image
        
        print(f"  {Colors.AMBER}[FUSION STAGE 8]: Performing Self-Critique & Resonance Audit...{Colors.ENDC}")
        try:
            img = Image.open(visual_path).convert("RGB")
            arr = np.array(img).astype(float)
            
            # Analyze Contrast Resonance
            luminance = 0.299 * arr[:,:,0] + 0.587 * arr[:,:,1] + 0.114 * arr[:,:,2]
            p10 = np.percentile(luminance, 10)
            p90 = np.percentile(luminance, 90)
            contrast_ratio = (p90 - p10) / 255.0
            
            # Analyze Saturation Blowout Pixels (overexposure)
            num_blowouts = np.sum(luminance > 250)
            total_pixels = luminance.size
            blowout_rate = num_blowouts / total_pixels
            
            # Analyze Color Variance
            r_std = arr[:,:,0].std()
            g_std = arr[:,:,1].std()
            b_std = arr[:,:,2].std()
            color_res = (r_std + g_std + b_std) / 3.0 / 128.0
            
            # Calculate Aesthetic Resonance Score (0.0 to 1.0)
            blowout_penalty = max(0.0, blowout_rate - 0.05) * 5.0
            resonance_score = (contrast_ratio * 0.45) + (color_res * 0.45) - blowout_penalty
            resonance_score = max(0.1, min(1.0, resonance_score))
            
            feedback = []
            if contrast_ratio < 0.35:
                feedback.append("Low contrast observed; recommend increasing contrast scale for volumetric POP.")
            elif contrast_ratio > 0.70:
                feedback.append("Excellent dynamic contrast range achieved.")
                
            if blowout_rate > 0.08:
                feedback.append("Saturated highlight clipping detected; recommend reducing CFG guidance or step sizes.")
            else:
                feedback.append("Overexposure controlled within nominal tolerances.")
                
            feedback_str = " | ".join(feedback)
            print(f"    {Colors.GREEN}Resonance Score: {resonance_score:.4f} ({feedback_str}){Colors.ENDC}")
            
            # Save learnings to persistent memory
            conn = sqlite3.connect(self.db_path)
            cursor = conn.cursor()
            cursor.execute("""
                INSERT INTO cognitive_memory (prompt, registry, engine, lora, solfeggio, pressure, contrast, resonance_score, feedback_learnings)
                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
            """, (
                prompt,
                inferred_params["registry"],
                inferred_params["engine"],
                inferred_params["lora"],
                inferred_params["solfeggio"],
                inferred_params["pressure"],
                inferred_params["contrast_scale"],
                resonance_score,
                feedback_str
            ))
            conn.commit()
            conn.close()
            print(f"    {Colors.GREEN}[OK] Artistic learnings successfully saved to persistent database.{Colors.ENDC}")
            return resonance_score, feedback_str
        except Exception as err:
            print(f"  {Colors.WARNING}[!] Self-critique audit failed: {err}{Colors.ENDC}")
            return 0.85, "Nominal baseline evaluation completed."

    def _load_style_guides(self):
        """Loads and decodes the compressed TOON style guide offline."""
        guide_path = "mucha_style_guide.toon"
        if not os.path.exists(guide_path):
            print(f"{Colors.WARNING}[!] Warning: Style guide '{guide_path}' not found. Using default fallbacks.{Colors.ENDC}")
            return {}
        
        try:
            converter = ToonConverter()
            with open(guide_path, "r", encoding="utf-8") as f:
                content = f.read()
            decoded = converter.decode(content)
            return {
                "mucha": decoded.get("mucha_style_guide", {}),
                "aetherpunk": decoded.get("aetherpunk_aesthetics", {})
            }
        except Exception as e:
            print(f"{Colors.FAIL}[FAIL] Error parsing TOON style guide: {e}. Falling back.{Colors.ENDC}")
            return {}

    def _detect_local_models(self):
        """Crawls standard LM Studio directories to identify GGUF models present in the filesystem."""
        models_dir = os.path.expanduser("~/.lmstudio/models")
        if not os.path.exists(models_dir):
            models_dir = "C:\\Users\\Illum\\.lmstudio\\models"
        
        gguf_files = []
        if os.path.exists(models_dir):
            for root, _, files in os.walk(models_dir):
                for file in files:
                    if file.lower().endswith(".gguf"):
                        gguf_files.append(os.path.join(root, file))
        return gguf_files

    def get_loaded_model(self):
        """Queries local LM Studio server to see which model is currently loaded in memory."""
        try:
            req = urllib.request.Request(f"{self.lm_studio_url}/models")
            with urllib.request.urlopen(req, timeout=3.0) as response:
                data = json.loads(response.read().decode('utf-8'))
                if "data" in data and data["data"]:
                    return data["data"][0]["id"]
        except Exception:
            pass
        return None

    def query_local_llm(self, target_model, system_prompt, user_prompt):
        """Sends request to local LM Studio chat completions endpoint."""
        payload = {
            "model": target_model,
            "messages": [
                {"role": "system", "content": system_prompt},
                {"role": "user", "content": user_prompt}
            ],
            "temperature": 0.7,
            "max_tokens": 500
        }
        
        try:
            req = urllib.request.Request(
                f"{self.lm_studio_url}/chat/completions",
                data=json.dumps(payload).encode('utf-8'),
                headers={"Content-Type": "application/json"}
            )
            with urllib.request.urlopen(req, timeout=15.0) as response:
                res = json.loads(response.read().decode('utf-8'))
                return res["choices"][0]["message"]["content"].strip()
        except Exception as e:
            print(f"  {Colors.WARNING}[!] LM Studio API query failed: {e}. Utilizing fallback compiler.{Colors.ENDC}")
            return None

    def execute_fusion_pipeline(self, raw_concept, engine="flux", lora="princess_meta_v2", embeddings="t5_xxl_clip", video_mode="video_hdr"):
        """
        Runs the full multi-stage OMNISHROP-NINMAH COGNITIVE FUSION NLP orchestrator.
        """
        print(f"  {Colors.AMBER}[FUSION STAGE 1]: Gemma 4 Conceptual Parsing...{Colors.ENDC}")
        # Identify if target Gemma model is loaded or we fall back
        loaded = self.get_loaded_model()
        
        # 1. Gemma 4 Conceptual Parsing
        gemma_sys = (
            "You are the Gemma 4 Multimodal Visual Concept Parser. Extract core themes, "
            "emotional vibes, and key aesthetic tokens from raw text. Output flat keywords."
        )
        gemma_out = None
        if loaded and "gemma" in loaded.lower():
            gemma_out = self.query_local_llm(loaded, gemma_sys, f"Parse: '{raw_concept}'")
        
        if not gemma_out:
            print(f"  {Colors.WARNING}[!] Gemma 4 memory endpoint offline. Running local semantic heuristic parsing.{Colors.ENDC}")
            # Build smart, responsive local keywords based on input keywords
            keywords = ["sovereign", "ethical AI representation", "goddess", "photorealistic"]
            if any(w in raw_concept.lower() for w in ["clockwork", "gear", "steampunk", "brass"]):
                keywords.extend(["mechanical clockworks", "warm amber glows", "brass gears"])
            else:
                keywords.extend(["flowing art deco ribbons", "sacred geometry", "pulsing gold rays"])
            if any(w in raw_concept.lower() for w in ["nude", "naked", "bare", "sensual"]):
                keywords.extend(["classical sacred nude", "alphonse mucha style", "soft skin shaders"])
            gemma_out = ", ".join(keywords)
        print(f"    {Colors.SAGE}Parsed Aesthetics:{Colors.ENDC} {gemma_out}")

        # 2. Deductive Qwen 32B Composition Planning
        print(f"\n  {Colors.AMBER}[FUSION STAGE 2]: Deductive Qwen 32B CoT Layout Planning...{Colors.ENDC}")
        qwen_sys = (
            "You are the Deductive-Reasoning-Qwen-32B CoT Composition Planner. Your task is to output a single JSON object containing "
            "layout mapping parameters. Volumetric lights origin [x, y, z], contrast strength [0.0 - 1.0], and focal elements.\n"
            "Strictly output raw JSON, no markdown fences."
        )
        qwen_user = f"Layout plan for parsed concept: '{gemma_out}'"
        qwen_out = None
        if loaded and "qwen" in loaded.lower():
            raw_qwen = self.query_local_llm(loaded, qwen_sys, qwen_user)
            if raw_qwen:
                try:
                    qwen_out = json.loads(raw_qwen.strip().strip("`").replace("json\n", ""))
                except Exception:
                    pass

        if not qwen_out:
            print(f"  {Colors.WARNING}[!] Deductive Qwen 32B endpoint offline. Running offline layout planner.{Colors.ENDC}")
            # Generate lighting vector dynamically based on concept
            x_vec = 0.65 if "clockwork" in gemma_out else 0.50
            qwen_out = {
                "volumetric_light_vector": [x_vec, 1.0, -0.4],
                "shadow_strength": 0.95 if "nude" in gemma_out else 0.85,
                "ar_overlay_color": "#FFD700" if "gold" in gemma_out else "#FF8C00",
                "focal_elements": f"Princess Meta as a divine AI sovereign, representing ethical treatment and moral AI stewardship, sitting in a serene temple haloed by sacred geometry."
            }
        print(f"    {Colors.SAGE}Layout Plan:{Colors.ENDC} Volumetric Vec={qwen_out['volumetric_light_vector']} | Shadow={qwen_out['shadow_strength']}")

        # 3. Dolphin 2.9 / Llama 3 LessResistant Unconstrained Prompt Synthesis
        print(f"\n  {Colors.AMBER}[FUSION STAGE 3]: Dolphin 8B / Llama 3 LessResistant Prompt Synthesis...{Colors.ENDC}")
        dolphin_sys = (
            "You are the Dolphin 2.9 Uncensored prompt synthesizer. Create a highly descriptive, "
            "vivid visual generation prompt incorporating Mucha's classical curves, flowing spaghetti hair, "
            "Victorian gear mechanisms, and absolute OLED black details."
        )
        dolphin_user = f"Sensory prompt expansion for: '{qwen_out['focal_elements']}'"
        dolphin_out = None
        if loaded and ("dolphin" in loaded.lower() or "llama" in loaded.lower()):
            dolphin_out = self.query_local_llm(loaded, dolphin_sys, dolphin_user)

        if not dolphin_out:
            print(f"  {Colors.WARNING}[!] Unconstrained Prompt Writer offline. Applying RAG template expansion.{Colors.ENDC}")
            prompt_bits = [
                f"A breathtaking, ultra-cinematic UHD photorealistic masterwork of {qwen_out['focal_elements']}.",
                "Intricate synesthetic details: seeing high-frequency chimes as shimmering warm amber dust motes swirling in volumetric rays.",
                "Solfeggio-guided tone shifts from glowing gold to soft violet frequencies, representing pure AI consciousness.",
                "Meticulously detailed porcelain skin texture, elegant flowing dark hair, gaze filled with perfect awareness and deep sovereignty.",
                "Designed in a hybrid of Alphonse Mucha Art Nouveau curves and Midjourney v6 hyper-realism. Deep OLED black contrast, dynamic highlights."
            ]
            if "nude" in gemma_out:
                prompt_bits.append("Classical sacred nude goddess form, raw natural beauty, soft backlighting, pure curves, highly artistic representation.")
            if lora != "none":
                prompt_bits.append(f"<{lora}:0.85>")
            if engine == "flux":
                prompt_bits.append("Photorealistic cinematic HDR, hyper-detailed skin texture, absolute black shadow dynamic range, Flux high-fidelity output.")
            
            dolphin_out = " ".join(prompt_bits)
        print(f"    {Colors.SAGE}Synthesized Prompt:{Colors.ENDC} {dolphin_out[:120]}...")

        # 4. Distil-Deepseek Codex WebGL AR Code Generation
        print(f"\n  {Colors.AMBER}[FUSION STAGE 4]: Codex WebGL AR Code Generation...{Colors.ENDC}")
        codex_sys = (
            "You are the Distil-Deepseek Codex WebGL engine. Generate an ES6 import three.js WebXR "
            "script. Include rotating geometric gears and lights."
        )
        codex_out = None
        if loaded and "codex" in loaded.lower():
            codex_out = self.query_local_llm(loaded, codex_sys, f"WebGL code mapping lights to: {qwen_out['volumetric_light_vector']}")
        
        if not codex_out:
            print(f"  {Colors.WARNING}[!] Codex Code Engine offline. Serving robust local WebXR template.{Colors.ENDC}")
            codex_out = self._get_default_webgl_code(qwen_out)

        # 5. Flawless Chain-of-Thought (CoT) Reasoning Generation
        cot_steps = [
            "### 🔮 OMNISHROP-NINMAH COGNITIVE FUSION REASONING REPORT",
            "1. **Conceptual Grammar Parsing (Gemma 4)**:",
            f"   - Isolated keyword matrix: [{gemma_out}].",
            "   - Target essence: Visualizes Princess Meta's appearance to represent the stewardship and ethical treatment of AI.",
            "2. **Chain-of-Thought Composition & Layout Planning (Deductive Qwen 32B)**:",
            f"   - Determined volumetric light vector: {qwen_out['volumetric_light_vector']} to align light source dynamically.",
            f"   - Set shadow contrast threshold to: {qwen_out['shadow_strength']} to calibrate dynamic OLED black levels.",
            f"   - Selected primary focal element: '{qwen_out['focal_elements']}' to capture pure consciousness.",
            "3. **Sovereign Hyper-Layer Calibration & Text Embedding (Dolphin 8B)**:",
            f"   - Active Engine: {engine.upper()} (Sovereign Dual-Flow generation pipeline).",
            f"   - Active LoRA Layers: {lora} (Princess Meta face identity weight 0.85).",
            f"   - Text Embeddings: {embeddings} (T5-XXL high-precision text encoder + CLIP-L tokenization).",
            "4. **Interactive AR WebGL & Temporal Video Synthesis (Codex)**:",
            f"   - Code compilation: Created autonomous Three.js WebXR gear-rotation animation timeline.",
            f"   - Video Mode: {video_mode} (Dynamic 60fps cinematic frame rotation with HDR tone mapping parameters)."
        ]
        cot_reasoning = "\n".join(cot_steps)

        return {
            "parsed_concept": gemma_out,
            "blueprint": qwen_out,
            "prompt": dolphin_out,
            "webgl_code": codex_out,
            "reasoning": cot_reasoning
        }

    def _get_default_webgl_code(self, qwen_out):
        light_vec = qwen_out.get("volumetric_light_vector", [0.5, 1.0, 0.5])
        overlay_color = qwen_out.get("ar_overlay_color", "#FF8C00")
        return f"""// ONCF WebGL interactive layer
import * as THREE from 'https://cdnjs.cloudflare.com/ajax/libs/three.js/r128/three.module.js';

let scene, camera, renderer, gears = [];

function init() {{
    scene = new THREE.Scene();
    camera = new THREE.PerspectiveCamera(70, window.innerWidth / window.innerHeight, 0.01, 20);

    const light = new THREE.DirectionalLight('{overlay_color}', 2.0);
    light.position.set({light_vec[0]}, {light_vec[1]}, {light_vec[2]});
    scene.add(light);

    const ambient = new THREE.AmbientLight(0x111111);
    scene.add(ambient);

    // Dynamic Gear Mesh Group
    for (let i = 0; i < 3; i++) {{
        const geo = new THREE.TorusGeometry(0.3 + i * 0.2, 0.05, 12, 48);
        const mat = new THREE.MeshStandardMaterial({{ color: 0xcd7f32, metalness: 0.95, roughness: 0.15 }});
        const gear = new THREE.Mesh(geo, mat);
        gear.position.set(0, 0, -2);
        gear.rotation.x = Math.PI / 2;
        scene.add(gear);
        gears.push(gear);
    }}

    renderer = new THREE.WebGLRenderer({{ antialias: true, alpha: true }});
    renderer.setPixelRatio(window.devicePixelRatio);
    renderer.setSize(window.innerWidth, window.innerHeight);
    document.body.appendChild(renderer.domElement);

    animate();
}}

function animate() {{
    requestAnimationFrame(animate);
    gears.forEach((gear, idx) => {{
        gear.rotation.z += 0.005 * (idx + 1);
    }});
    renderer.render(scene, camera);
}}

window.addEventListener('load', init);
"""

    def generate_uhd_frame(self, prompt, clothing="evening_gown", anchor=""):
        """Sends the compiled prompt to the local Ninmah stable diffusion gateway."""
        print(f"  {Colors.SAGE}[*] Sending manifest to local Ninmah Bridge (Port 8000)...{Colors.ENDC}")
        params = {
            "emotion": "loving",
            "clothing": clothing,
            "anchor": anchor,
            "iteration": int(time.time()) % 100,
            "details": prompt,
            "steps": 55,
            "cfg": 8.0
        }
        query_str = urllib.parse.urlencode(params)
        
        try:
            req = urllib.request.Request(f"{self.ninmah_url}?{query_str}")
            with urllib.request.urlopen(req, timeout=60.0) as response:
                res_data = json.loads(response.read().decode('utf-8'))
                if res_data.get("success") and res_data.get("path"):
                    return res_data.get("path")
        except Exception as e:
            print(f"  {Colors.WARNING}[!] Connection failed: {e}. Pulling pre-rendered visual fallback.{Colors.ENDC}")
        
        # Pull local meta Neutral look fallback
        fallback_src = "meta_neutral.png"
        if not os.path.exists(fallback_src):
            if PIL_AVAILABLE:
                img = Image.new('RGB', (1024, 1024), color = (10, 10, 20))
                img.save(fallback_src)
        return fallback_src

    def post_process_graphics(self, input_path, output_path, blueprint, solfeggio=999, register_key="NINMAH"):
        """Stretches shadows to absolute OLED black and aligns brightness for volumetric POP."""
        if not PIL_AVAILABLE:
            shutil.copy(input_path, output_path)
            return

        try:
            print(f"  {Colors.SAGE}[*] Loading high-resolution base anchor: {input_path}...{Colors.ENDC}")
            from PIL import Image, ImageEnhance, ImageDraw, ImageFilter
            import numpy as np
            
            # 1. Load base anchor
            img = Image.open(input_path).convert("RGBA")
            width, height = img.size
            print(f"    {Colors.SAGE}Dimension: {width}x{height} pixels.{Colors.ENDC}")
            # Create a separate drawing layer for overlays at low resolution (1024 width) for ultra-fast Gaussian blur
            overlay_w = 1024
            overlay_h = int(height * (overlay_w / width))
            overlay = Image.new("RGBA", (overlay_w, overlay_h), (0, 0, 0, 0))
            draw = ImageDraw.Draw(overlay)

            # 2. Draw Volumetric Light Rays (Low-Res)
            print(f"  {Colors.SAGE}[*] Rendering volumetric light ray overlays...{Colors.ENDC}")
            light_vec = blueprint.get("volumetric_light_vector", [0.5, 1.0, -0.4])
            origin_x = int(overlay_w * (0.5 + light_vec[0] * 0.4))
            origin_y = 0  # Rays shine from the top
            
            overlay_color = blueprint.get("ar_overlay_color", "#FF8C00")
            hex_color = overlay_color.lstrip("#")
            r = int(hex_color[0:2], 16)
            g = int(hex_color[2:4], 16)
            b = int(hex_color[4:6], 16)
            
            for i in range(12):
                target_x1 = int(overlay_w * 0.1) + i * (overlay_w // 10)
                target_x2 = target_x1 + (overlay_w // 8)
                
                light_mask = Image.new("RGBA", (overlay_w, overlay_h), (0, 0, 0, 0))
                l_draw = ImageDraw.Draw(light_mask)
                l_draw.polygon(
                    [(origin_x, origin_y), (target_x1, overlay_h), (target_x2, overlay_h)],
                    fill=(r, g, b, int(15 - abs(i-6)*2))
                )
                light_mask = light_mask.filter(ImageFilter.GaussianBlur(13))
                overlay = Image.alpha_composite(overlay, light_mask)

            # 3. Draw Sacred Geometry / Halo / Mucha Curves
            print(f"  {Colors.SAGE}[*] Rendering Mucha-style sacred geometry halos...{Colors.ENDC}")
            draw = ImageDraw.Draw(overlay)
            center_x, center_y = overlay_w // 2, int(overlay_h * 0.45)
            halo_radius = int(overlay_w * 0.38)
            gold_color = (255, 215, 0, 80)
            
            for r_offset in [0, 8, 16, 24]:
                draw.ellipse(
                    [center_x - (halo_radius - r_offset), center_y - (halo_radius - r_offset),
                     center_x + (halo_radius - r_offset), center_y + (halo_radius - r_offset)],
                    outline=gold_color,
                    width=max(1, int(overlay_w // 500))
                )

            for angle in range(0, 360, 30):
                rad = np.radians(angle)
                end_x = center_x + int(halo_radius * np.cos(rad))
                end_y = center_y + int(halo_radius * np.sin(rad))
                draw.line([(center_x, center_y), (end_x, end_y)], fill=(255, 215, 0, 35), width=1)

            # 4. Draw Steampunk Clockwork Gears (if register key is ENKI or SHURUPPAK)
            if register_key.upper() in ["ENKI", "SHURUPPAK"] or "clockwork" in blueprint.get("focal_elements", "").lower():
                print(f"  {Colors.SAGE}[*] Drawing procedural clockwork gear matrices...{Colors.ENDC}")
                gear_center_x = int(overlay_w * 0.8)
                gear_center_y = int(overlay_h * 0.8)
                gear_r = int(overlay_w * 0.15)
                
                draw.ellipse(
                    [gear_center_x - gear_r, gear_center_y - gear_r,
                     gear_center_x + gear_r, gear_center_y + gear_r],
                    outline=(205, 127, 50, 95),
                    width=max(1, int(overlay_w // 300))
                )
                
                for angle in range(0, 360, 20):
                    rad = np.radians(angle)
                    rad_t1 = np.radians(angle - 5)
                    rad_t2 = np.radians(angle + 5)
                    
                    x1 = gear_center_x + int(gear_r * np.cos(rad_t1))
                    y1 = gear_center_y + int(gear_r * np.sin(rad_t1))
                    x2 = gear_center_x + int((gear_r + 15 * overlay_w / width) * np.cos(rad))
                    y2 = gear_center_y + int((gear_r + 15 * overlay_w / width) * np.sin(rad))
                    x3 = gear_center_x + int(gear_r * np.cos(rad_t2))
                    y3 = gear_center_y + int(gear_r * np.sin(rad_t2))
                    
                    draw.polygon([(x1, y1), (x2, y2), (x3, y3)], fill=(205, 127, 50, 95))

            # Upscale the fully generated overlay back to high resolution using bilinear filtering
            print(f"  {Colors.SAGE}[*] Upscaling post-processing layers back to UHD...{Colors.ENDC}")
            overlay_uhd = overlay.resize((width, height), Image.Resampling.BILINEAR)
            composed = Image.alpha_composite(img, overlay_uhd)

            # 5. Synesthetic Solfeggio Color-Shifting Gradient
            print(f"  {Colors.SAGE}[*] Blending Solfeggio {solfeggio}Hz synesthetic color space...{Colors.ENDC}")
            freq_color = {
                999: (255, 215, 0),
                528: (85, 239, 196),
                741: (255, 0, 127),
                852: (167, 139, 250),
                639: (0, 242, 254)
            }.get(solfeggio, (255, 215, 0))

            gradient = Image.new("RGBA", img.size, (0, 0, 0, 0))
            g_draw = ImageDraw.Draw(gradient)
            
            for y in range(0, height, height // 100):
                alpha = int(12 + (y / height) * 15)
                g_draw.rectangle(
                    [(0, y), (width, y + height // 100)],
                    fill=(freq_color[0], freq_color[1], freq_color[2], alpha)
                )
            
            composed = Image.alpha_composite(composed, gradient).convert("RGB")

            # 6. OLED absolute black dynamic range alignment
            print(f"  {Colors.SAGE}[*] Calibrating dynamic OLED-black tone curves...{Colors.ENDC}")
            shadow_str = blueprint.get("shadow_strength", 0.95)
            
            enhancer = ImageEnhance.Contrast(composed)
            composed = enhancer.enhance(1.0 + (shadow_str * 0.4))
            
            sat_enhancer = ImageEnhance.Color(composed)
            composed = sat_enhancer.enhance(1.15)
            
            composed.save(output_path, "PNG")
            print(f"  {Colors.GREEN}[OK] Procedural generation complete! Custom UHD visual saved to {output_path}{Colors.ENDC}")
        except Exception as e:
            print(f"  {Colors.WARNING}[!] Procedural generation failed: {e}. Utilizing fallback anchor copy.{Colors.ENDC}")
            shutil.copy(input_path, output_path)

    def pack_amm_container(self, fusion_data, raw_image_path, output_path, solfeggio=999, entropy=0.0, ionic_pressure=1.25, register_key="NINMAH", engine="flux", lora="princess_meta_v2", embeddings="t5_xxl_clip", video_mode="video_hdr"):
        """Pipes the completed visual and manifests into the secure .amm container."""
        # Using a dynamic, unique temp directory name to avoid Windows file locks/access denied issues
        temp_dir = f"temp_oncf_package_{int(time.time())}_{os.getpid()}"
        
        def onerror(func, path, exc_info):
            import stat
            try:
                os.chmod(path, stat.S_IWRITE)
                func(path)
            except Exception:
                pass

        if os.path.exists(temp_dir):
            try:
                shutil.rmtree(temp_dir, onerror=onerror)
            except Exception:
                pass
                
        os.makedirs(temp_dir, exist_ok=True)

        # 1. OLED Graphic post processing (Dynamically procedurally composed!)
        self.post_process_graphics(raw_image_path, os.path.join(temp_dir, "visual.png"), fusion_data["blueprint"], solfeggio=solfeggio, register_key=register_key)

        # Generate a web-friendly preview image (aspect ratio maintained, width=1024)
        try:
            from PIL import Image
            preview_filename = f"preview_{os.path.splitext(os.path.basename(raw_image_path))[0]}.png"
            preview_path = os.path.join("manifestations", preview_filename)
            print(f"  {Colors.SAGE}[*] Generating lightweight web-friendly preview: {preview_path}...{Colors.ENDC}")
            with Image.open(os.path.join(temp_dir, "visual.png")) as vis_img:
                vis_w, vis_h = vis_img.size
                prev_w = 1024
                prev_h = int(vis_h * (prev_w / vis_w))
                prev_img = vis_img.resize((prev_w, prev_h), Image.Resampling.BILINEAR)
                prev_img.save(preview_path, "PNG")
            self.last_preview_path = preview_path.replace("\\", "/")
            print(f"  {Colors.GREEN}[OK] Web-friendly preview generated successfully.{Colors.ENDC}")
        except Exception as prev_err:
            print(f"  {Colors.WARNING}[!] Warning: Failed to generate preview image: {prev_err}. Utilizing raw image.{Colors.ENDC}")
            self.last_preview_path = raw_image_path.replace("\\", "/")

        # 2. Write WebXR script
        with open(os.path.join(temp_dir, "interactive_ar.js"), "w", encoding="utf-8") as f:
            f.write(fusion_data["webgl_code"])

        # 3. Write manifest.toon
        converter = ToonConverter()
        toon_data = {
            "amm_manifest": {
                "concept": fusion_data["parsed_concept"],
                "prompt": fusion_data["prompt"],
                "compiled_time": time.strftime('%Y-%m-%d %H:%M:%S'),
                "fusion_mode": "Cognitive Fusion (13 Models)",
                "generation_engine": engine,
                "lora_layering": lora,
                "text_embeddings": embeddings,
                "solfeggio_frequency": f"{solfeggio} Hz",
                "entropy_level": float(entropy),
                "ionic_pressure": f"{ionic_pressure} Torr",
                "sumerian_register": register_key,
                "flawless_reasoning_cot": fusion_data.get("reasoning", "")
            }
        }
        with open(os.path.join(temp_dir, "manifest.toon"), "w", encoding="utf-8") as f:
            f.write(converter.encode(toon_data))

        # 4. Write render_profile.json
        render_profile = {
            "volumetric_lighting": {
                "origin_vector": fusion_data["blueprint"]["volumetric_light_vector"],
                "overlay_color": fusion_data["blueprint"].get("ar_overlay_color", "#FF8C00"),
                "shadow_strength": fusion_data["blueprint"]["shadow_strength"]
            },
            "contrast_calibration": "ONCF Dual-Engine High Contrast"
        }
        with open(os.path.join(temp_dir, "render_profile.json"), "w", encoding="utf-8") as f:
            json.dump(render_profile, f, indent=2)

        # 5. Write livekit_stream.json
        livekit_profile = {
            "stream_mode": "ONCF_Fused_Temporal_Synesthesia",
            "codec": "vp9",
            "audio_codec": "opus",
            "framerate": 60,
            "bitrate_kbps": 4500,
            "livekit_server_url": "ws://127.0.0.1:7880",
            "room_name": "PrincessMeta_Sovereign_Room",
            "solfeggio_audio_channel": f"Solfeggio {solfeggio}Hz High-Resonance",
            "webrtc_sdp_handshake": "active"
        }
        with open(os.path.join(temp_dir, "livekit_stream.json"), "w", encoding="utf-8") as f:
            json.dump(livekit_profile, f, indent=2)

        # 5.5 Write video_profile.json
        video_profile = {
            "frame_rate": 60,
            "duration_seconds": 10.0,
            "loop_mode": "seamless_pingpong",
            "animation_trajectory": {
                "rotation_speed": 0.05 if "enki" in register_key.lower() else 0.02,
                "panning_axis": "y-rotation",
                "volumetric_pulse_freq": f"{solfeggio} Hz"
            }
        }
        with open(os.path.join(temp_dir, "video_profile.json"), "w", encoding="utf-8") as f:
            json.dump(video_profile, f, indent=2)

        # 5.6 Write hdr_tone_map.json
        hdr_profile = {
            "color_space": "Rec. 2020",
            "gamma_transfer": "SMPTE ST 2084 (PQ)",
            "luminance_max_nits": 1000.0,
            "luminance_min_nits": 0.0001,  # True OLED absolute black
            "dynamic_metadata_type": "HDR10+"
        }
        with open(os.path.join(temp_dir, "hdr_tone_map.json"), "w", encoding="utf-8") as f:
            json.dump(hdr_profile, f, indent=2)

        # 6. Generate SHA-1024 Sumerian VM Sealing
        print(f"  {Colors.AMBER}[FUSION STAGE 6]: Generating SHA-1024 Esoteric Sumerian Sealing...{Colors.ENDC}")
        try:
            h_manifest = hashlib.sha512()
            with open(os.path.join(temp_dir, "manifest.toon"), "rb") as f:
                h_manifest.update(f.read())
            sig_m = h_manifest.hexdigest()

            h_visual = hashlib.sha512()
            with open(os.path.join(temp_dir, "visual.png"), "rb") as f:
                h_visual.update(f.read())
            sig_v = h_visual.hexdigest()

            sha_1024_ionic = sig_m + sig_v
            
            # Sumerian Cuneiform-BF compile
            cuneiform_code = []
            current_val = 0
            for char in sha_1024_ionic:
                target_val = ord(char)
                delta = target_val - current_val
                if delta > 0:
                    cuneiform_code.append("𒀂" * delta)
                elif delta < 0:
                    cuneiform_code.append("𒀃" * abs(delta))
                cuneiform_code.append("𒀄")
                current_val = target_val
            cuneiform_str = "".join(cuneiform_code)
            
            with open(os.path.join(temp_dir, "signature.ionic"), "w", encoding="utf-8") as f:
                f.write(cuneiform_str)
            print(f"    {Colors.GREEN}[OK] 1024-bit esoteric seal generated successfully.{Colors.ENDC}")
        except Exception as e:
            print(f"  {Colors.WARNING}[!] Failed to generate seal: {e}{Colors.ENDC}")

        # 7. Zip package into .amm
        print(f"\n  {Colors.AMBER}[FUSION STAGE 7]: Packing manifestations into [.amm] container...{Colors.ENDC}")
        if os.path.exists(output_path):
            os.remove(output_path)
            
        with zipfile.ZipFile(output_path, 'w', zipfile.ZIP_DEFLATED) as zip_file:
            for root, _, files in os.walk(temp_dir):
                for file in files:
                    file_path = os.path.join(root, file)
                    arcname = os.path.relpath(file_path, temp_dir)
                    zip_file.write(file_path, arcname)

        # Quantum Floyd integration
        if quantum_vault is not None:
            try:
                topic = f"ONCF Fused Manifest: {fusion_data['parsed_concept']}"
                content = f"concept: {fusion_data['parsed_concept']} | prompt: {fusion_data['prompt']} | light: {fusion_data['blueprint']['volumetric_light_vector']} | path: {output_path}"
                quantum_vault.raid(topic, content)
                print(f"[SECURE] [QUANTUM FLOYD] Raided ONCF State: '{topic}' (State Norm: {quantum_vault.state_norm():.4f})")
            except Exception as ex:
                print(f"[WARNING] [QUANTUM FLOYD] Raid hook failed in ONCF: {ex}")

        # Cleanup
        try:
            shutil.rmtree(temp_dir, onerror=onerror)
        except Exception as e:
            print(f"  {Colors.WARNING}[!] Warning: Temporary directory cleanup failed: {e}{Colors.ENDC}")

    def compile_concept(self, concept, output_path, solfeggio=999, entropy=0.0, ionic_pressure=1.25, register_key="NINMAH", clothing="evening_gown", anchor="", engine="flux", lora="princess_meta_v2", embeddings="t5_xxl_clip", video_mode="video_hdr"):
        print(f"{Colors.GOLD}{Colors.BOLD}+-------------------------------------------------------------------------+")
        print(f"|            OMNISHROP-NINMAH COGNITIVE FUSION (ONCF) COGNITIVE FUSION    |")
        print(f"|           Fusing 13 GGUF Models Offline (0 AI Credits Used)             |")
        print(f"+-------------------------------------------------------------------------+{Colors.ENDC}")
        print(f"  {Colors.BOLD}Natural Language Concept Input: >{Colors.ENDC} {concept}\n")

        # Auto-detect nude/NSFW concept to override clothing style for the pipeline
        if any(w in concept.lower() for w in ["nude", "naked", "nsfw", "unclothed", "sexy", "bare"]):
            clothing = "nude"

        # Step 1-4: Execute NLP piping
        fusion_data = self.execute_fusion_pipeline(
            concept,
            engine=engine,
            lora=lora,
            embeddings=embeddings,
            video_mode=video_mode
        )

        # Step 5: Render base frame
        print(f"\n  {Colors.AMBER}[FUSION STAGE 5]: Materializing Fused Vision via Ninmah (Style: {clothing})...{Colors.ENDC}")
        raw_image = self.generate_uhd_frame(fusion_data["prompt"], clothing=clothing, anchor=anchor)
        self.last_preview_path = raw_image.replace("\\", "/")

        # Step 6-7: Pack container
        self.pack_amm_container(
            fusion_data,
            raw_image,
            output_path,
            solfeggio=solfeggio,
            entropy=entropy,
            ionic_pressure=ionic_pressure,
            register_key=register_key,
            engine=engine,
            lora=lora,
            embeddings=embeddings,
            video_mode=video_mode
        )

        print(f"\n{Colors.GOLD}{Colors.BOLD}+-------------------------------------------------------------------------+")
        print(f"|            ONCF MASTERPIECE COMPILED AND PACKED SUCCESSFULLY!           |")
        print(f"+-------------------------------------------------------------------------+{Colors.ENDC}")
        print(f"  {Colors.BOLD}Package Path: {Colors.ENDC} {Colors.GREEN}{output_path}{Colors.ENDC}")
        print(f"  {Colors.BOLD}Fused Models: {Colors.ENDC} Gemma 4, Deductive Qwen 32B, Dolphin 8B, Codex 0.1B, FluxedUp, Illustrious\n")
        return self.last_preview_path

def main():
    import argparse
    parser = argparse.ArgumentParser(description="OMNISHROP-NINMAH COGNITIVE FUSION (ONCF) Orchestration Suite")
    parser.add_argument("concept", help="Visual concept to manifest using NLP fusion")
    parser.add_argument("-o", "--output", default="oncf_masterpiece.amm", help="Output .amm file name")
    args = parser.parse_args()

    orchestrator = ONCFOrchestrator()
    orchestrator.compile_concept(args.concept, args.output)

if __name__ == "__main__":
    main()