"""
IRIS — Intelligent Real-time Inference System for Visual Assistance
Next-Gen Cyber Assistive HUD & Real-time Spatial Navigation
Gradio Application for Hugging Face Spaces (Compatible with ZeroGPU & CPU)
"""

import os
import base64
import tempfile
from pathlib import Path

# Redirect YOLO config/cache to writable directory
os.environ["YOLO_CONFIG_DIR"] = "/tmp/Ultralytics"

# Hugging Face ZeroGPU compatibility
try:
    import spaces
except ImportError:
    class spaces:
        @staticmethod
        def GPU(func=None, duration=60):
            if func is None:
                def decorator(f):
                    return f
                return decorator
            return func

import cv2
import numpy as np
from PIL import Image
from ultralytics import YOLO
from gtts import gTTS
import gradio as gr

from priority_engine import PriorityEngine

# ── Config ───────────────────────────────────────────────────────────────────
CONF_THRESHOLD = 0.35
MODEL_NAME     = "yolo11n.pt"

# Initialize model & priority engine
model    = YOLO(MODEL_NAME)
priority = PriorityEngine()

# Colors for bounding boxes (RGB)
POSITION_COLORS = {
    "left":   (255, 140, 0),    # Orange
    "center": (16, 185, 129),   # Emerald Green (Directly Ahead)
    "right":  (59, 130, 246),   # Electric Blue
}


def get_position(center_x, frame_width):
    if center_x < frame_width / 3:
        return "left"
    elif center_x < 2 * frame_width / 3:
        return "center"
    return "right"


def generate_tts_audio(text: str):
    """Generate an MP3 audio file using gTTS."""
    if not text or len(text.strip()) == 0:
        return None
    try:
        clean_text = (
            text.replace("🔊", "")
            .replace("⚠️", "")
            .replace("—", ", ")
            .replace("–", ", ")
            .strip()
        )
        tts = gTTS(text=clean_text, lang="en", slow=False)
        tmp = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
        tts.save(tmp.name)
        return tmp.name
    except Exception as e:
        print(f"[TTS Error] {e}")
        return None


def render_spatial_hud(detections):
    """Render a 3-zone spatial radar HUD and obstacle badges."""
    left_items   = [d for d in detections if d["position"] == "left"]
    center_items = [d for d in detections if d["position"] == "center"]
    right_items  = [d for d in detections if d["position"] == "right"]

    def make_zone(items, zone_name, color):
        if not items:
            return f"""
            <div class="hud-zone" style="border-top: 3px solid {color}33;">
                <div class="hud-zone-header">
                    <span class="hud-zone-dot" style="background:{color};"></span>
                    <span class="hud-zone-name">{zone_name}</span>
                </div>
                <div class="hud-zone-val val-clear">CLEAR</div>
            </div>
            """
        top_obj = items[0]["object"].capitalize()
        conf    = int(items[0]["confidence"] * 100)
        extra   = f" +{len(items)-1}" if len(items) > 1 else ""
        is_warn = zone_name == "CENTER PATH"
        val_cls = "val-alert" if is_warn else "val-warn"
        return f"""
        <div class="hud-zone active" style="border-top: 3px solid {color};">
            <div class="hud-zone-header">
                <span class="hud-zone-dot pulse" style="background:{color};"></span>
                <span class="hud-zone-name">{zone_name}</span>
            </div>
            <div class="hud-zone-val {val_cls}">{top_obj} ({conf}%){extra}</div>
        </div>
        """

    zones_html = f"""
    <div class="hud-zones-bar">
        {make_zone(left_items, "LEFT ZONE", "#f97316")}
        {make_zone(center_items, "CENTER PATH", "#10b981")}
        {make_zone(right_items, "RIGHT ZONE", "#3b82f6")}
    </div>
    """

    if detections:
        badges = []
        for d in detections:
            c = POSITION_COLORS.get(d["position"], (200, 200, 200))
            rgb_str = f"rgb({c[0]},{c[1]},{c[2]})"
            badges.append(f"""
            <div class="det-chip">
                <span class="det-indicator" style="background:{rgb_str}; box-shadow: 0 0 8px {rgb_str}aa;"></span>
                <span class="det-title">{d['object'].capitalize()}</span>
                <span class="det-dir" style="color:{rgb_str};">{d['position'].upper()}</span>
                <span class="det-score">{int(d['confidence']*100)}%</span>
            </div>
            """)
        chips_html = f"""
        <div class="hud-breakdown">
            <div class="hud-subhead">RADAR DETECTIONS ({len(detections)})</div>
            <div class="chips-flex">
                {"".join(badges)}
            </div>
        </div>
        """
    else:
        chips_html = """
        <div class="hud-breakdown">
            <div class="hud-subhead">RADAR DETECTIONS</div>
            <div class="all-clear-banner">
                <span class="check-icon">✓</span> Navigation corridor is clear of detected obstacles.
            </div>
        </div>
        """

    return zones_html + chips_html


@spaces.GPU
def analyze_frame(image):
    """
    Run YOLO inference and Priority Engine on an input image.
    Returns:
        annotated_image, instruction_text, audio_filepath, detections_hud_html
    """
    if image is None:
        placeholder = """
        <div class="hud-zones-bar">
            <div class="hud-zone"><div class="hud-zone-val val-clear">STANDBY</div></div>
        </div>
        """
        return None, "STANDBY — Waiting for visual input.", None, placeholder

    h, w, _ = image.shape
    results = model(image, conf=CONF_THRESHOLD, verbose=False)

    detections = []
    annotated  = image.copy()

    for result in results:
        for box in result.boxes:
            conf = float(box.conf[0])
            if conf < CONF_THRESHOLD:
                continue

            cls_id   = int(box.cls[0])
            label    = model.names[cls_id]
            x1, y1, x2, y2 = box.xyxy[0].tolist()
            center_x = (x1 + x2) / 2
            pos      = get_position(center_x, w)

            detections.append({
                "object":     label,
                "confidence": round(conf, 2),
                "position":   pos,
                "bbox":       [round(x1, 1), round(y1, 1), round(x2, 1), round(y2, 1)],
            })

            # Draw bounding box and label in RGB
            color = POSITION_COLORS.get(pos, (255, 255, 255))
            cv2.rectangle(annotated, (int(x1), int(y1)), (int(x2), int(y2)), color, 2)

            tag = f"{label.upper()} {int(conf * 100)}%"
            (tw, th), _ = cv2.getTextSize(tag, cv2.FONT_HERSHEY_SIMPLEX, 0.55, 2)
            ty = max(int(y1) - 8, th + 6)
            cv2.rectangle(annotated, (int(x1), ty - th - 4), (int(x1) + tw + 6, ty + 2), color, -1)
            cv2.putText(
                annotated, tag, (int(x1) + 3, ty - 2),
                cv2.FONT_HERSHEY_SIMPLEX, 0.55, (10, 15, 25), 2
            )

    detections.sort(key=lambda d: d["confidence"], reverse=True)
    instruction = priority.pick(detections)
    audio_path  = generate_tts_audio(instruction)
    hud_html    = render_spatial_hud(detections)

    return annotated, f"🔊 {instruction}", audio_path, hud_html


@spaces.GPU
def analyze_webcam_input(b64_data, fallback_image):
    """Decodes live frame from browser canvas base64 or fallback Gradio snapshot."""
    image = None
    if b64_data and len(b64_data) > 100:
        try:
            raw = base64.b64decode(b64_data.split(",")[-1])
            arr = np.frombuffer(raw, np.uint8)
            bgr = cv2.imdecode(arr, cv2.IMREAD_COLOR)
            if bgr is not None:
                image = cv2.cvtColor(bgr, cv2.COLOR_BGR2RGB)
        except Exception as e:
            print(f"[Live frame decode error] {e}")

    if image is None and fallback_image is not None:
        image = fallback_image

    if image is None:
        placeholder = """
        <div class="all-clear-banner" style="color: #f59e0b; border-color: rgba(245,158,11,0.3);">
            ⚠️ Camera stream inactive. Click "Start Camera" above.
        </div>
        """
        return None, "STANDBY — Please start the camera stream.", None, placeholder

    return analyze_frame(image)


@spaces.GPU(duration=120)
def analyze_video(video_path):
    """Sample key frames from an uploaded video and generate navigation guidance."""
    if not video_path:
        return None, "No video provided", None

    cap = cv2.VideoCapture(video_path)
    total = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
    fps = cap.get(cv2.CAP_PROP_FPS) or 30

    sample_step = max(1, total // 6)
    sampled_frames = []
    frame_idx = 0
    all_instructions = []

    while True:
        ret, frame = cap.read()
        if not ret:
            break
        if frame_idx % sample_step == 0:
            frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
            ann, inst, _, _ = analyze_frame(frame_rgb)
            sampled_frames.append(ann)
            sec = round(frame_idx / fps, 1)
            all_instructions.append(f"⏱️ **{sec}s**: {inst.replace('🔊 ', '')}")
        frame_idx += 1

    cap.release()

    timeline = "\n\n".join(all_instructions)
    first_instruction = all_instructions[0].split(": ")[-1] if all_instructions else "Video analysis complete."
    overall_tts = generate_tts_audio(first_instruction)

    return sampled_frames, timeline, overall_tts


# ── Sequential Live Navigation Controller (Speaks 1st, then scans next) ─────
live_nav_js = """
() => {
    if (typeof window.isLiveNavActive === 'undefined') {
        window.isLiveNavActive = false;
        window.liveNavTimer = null;
    }

    window.isLiveNavActive = !window.isLiveNavActive;

    const btn = document.querySelector("#live_nav_toggle_btn button, #live_nav_toggle_btn");
    const status = document.getElementById("live_status_indicator");

    function triggerScan() {
        if (!window.isLiveNavActive) return;
        const scanBtn = document.querySelector("#scan_frame_btn button, #scan_frame_btn");
        if (scanBtn && !scanBtn.disabled) {
            scanBtn.click();
        }
    }

    // Called as soon as YOLO finishes processing a frame
    window.onFrameOutputReceived = function(instructionText) {
        if (!instructionText) {
            if (window.isLiveNavActive) {
                window.liveNavTimer = setTimeout(triggerScan, 1000);
            }
            return;
        }

        const clean = instructionText.replace("🔊", "").replace("⚠️", "").replace("STANDBY", "").trim();
        if (!clean || clean.includes("Please start") || clean.includes("Waiting for") || clean.includes("Camera stream")) {
            if (window.isLiveNavActive) {
                window.liveNavTimer = setTimeout(triggerScan, 1200);
            }
            return;
        }

        if (!('speechSynthesis' in window)) {
            // Fallback if browser doesn't support Web Speech
            if (window.isLiveNavActive) {
                window.liveNavTimer = setTimeout(triggerScan, 2500);
            }
            return;
        }

        // Cancel previous utterance
        window.speechSynthesis.cancel();

        const u = new SpeechSynthesisUtterance(clean);
        u.rate = 0.95;
        u.pitch = 1.0;

        let hasFinished = false;
        function onSpeechComplete() {
            if (hasFinished) return;
            hasFinished = true;

            // Frame output has finished speaking!
            // Wait 500ms pause, then trigger the next frame scan:
            if (window.isLiveNavActive) {
                window.liveNavTimer = setTimeout(triggerScan, 500);
            }
        }

        u.onend = onSpeechComplete;
        u.onerror = onSpeechComplete;

        // Safety timeout in case browser onend event is dropped
        const safeTimeout = Math.max(3500, clean.length * 120);
        setTimeout(onSpeechComplete, safeTimeout);

        window.speechSynthesis.speak(u);
    };

    if (window.isLiveNavActive) {
        if (btn) {
            btn.innerText = "⏹️ Stop Real-Time Navigation";
            btn.style.setProperty("background", "linear-gradient(135deg, #ef4444, #dc2626)", "important");
            btn.classList.add("active");
        }
        if (status) {
            status.innerHTML = "<span style='color:#10b981; font-weight:700;'>● ACTIVE</span> — Sequential speech: speaks full alert before scanning next frame.";
        }

        // Trigger the initial frame scan
        triggerScan();
    } else {
        if (btn) {
            btn.innerText = "▶️ Start Real-Time Navigation (Sequential Voice)";
            btn.style.setProperty("background", "linear-gradient(135deg, #10b981, #059669)", "important");
            btn.classList.remove("active");
        }
        if (status) {
            status.innerHTML = "<span style='color:#64748b;'>● STANDBY</span> — Click above to begin continuous scanning.";
        }
        if (window.liveNavTimer) {
            clearTimeout(window.liveNavTimer);
            window.liveNavTimer = null;
        }
        if ('speechSynthesis' in window) {
            window.speechSynthesis.cancel();
        }
    }
}
"""

# Extract current live frame from HTML5 video element with no shutter lag
js_extract_video = """
(b64, fallback_img) => {
    const video = document.querySelector("#webcam_viewport video");
    if (video && video.videoWidth > 0) {
        const canvas = document.createElement("canvas");
        canvas.width = video.videoWidth;
        canvas.height = video.videoHeight;
        const ctx = canvas.getContext("2d");
        ctx.drawImage(video, 0, 0, canvas.width, canvas.height);
        return [canvas.toDataURL("image/jpeg", 0.85), null];
    }
    return [b64, fallback_img];
}
"""

# ── Ultra-Professional Dark Theme CSS ────────────────────────────────────────
custom_css = """
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link href="https://fonts.googleapis.com/css2?family=Plus+Jakarta+Sans:wght@400;500;600;700;800&family=JetBrains+Mono:wght@500;700&display=swap" rel="stylesheet">

<style>
:root {
    --iris-bg: #07090e;
    --iris-card: rgba(15, 18, 28, 0.75);
    --iris-border: rgba(255, 255, 255, 0.08);
    --iris-cyan: #06b6d4;
    --iris-indigo: #6366f1;
    --iris-emerald: #10b981;
    --iris-amber: #f59e0b;
    --iris-rose: #f43f5e;
}

body, .gradio-container {
    background-color: var(--iris-bg) !important;
    font-family: 'Plus Jakarta Sans', -apple-system, BlinkMacSystemFont, sans-serif !important;
    color: #f1f5f9 !important;
    max-width: 1200px !important;
    margin: 0 auto !important;
}

/* Header & HUD Navbar */
.iris-brand-hero {
    display: flex;
    align-items: center;
    justify-content: space-between;
    padding: 18px 24px;
    margin-bottom: 20px;
    background: linear-gradient(135deg, rgba(20, 25, 40, 0.85), rgba(10, 13, 22, 0.95));
    border: 1px solid var(--iris-border);
    border-radius: 20px;
    backdrop-filter: blur(20px);
    box-shadow: 0 10px 30px -10px rgba(0, 0, 0, 0.5), inset 0 1px 0 rgba(255, 255, 255, 0.1);
}

.brand-left {
    display: flex;
    align-items: center;
    gap: 16px;
}

.brand-logo-hex {
    width: 46px;
    height: 46px;
    border-radius: 14px;
    background: linear-gradient(135deg, #6366f1, #06b6d4);
    display: flex;
    align-items: center;
    justify-content: center;
    box-shadow: 0 0 20px rgba(99, 102, 241, 0.5);
}

.brand-text h1 {
    font-size: 1.5rem;
    font-weight: 800;
    letter-spacing: -0.02em;
    background: linear-gradient(135deg, #ffffff 40%, #a5b4fc 100%);
    -webkit-background-clip: text;
    -webkit-text-fill-color: transparent;
    margin: 0;
    line-height: 1.2;
}

.brand-text p {
    font-size: 0.8rem;
    color: #94a3b8;
    margin: 0;
    font-weight: 500;
    letter-spacing: 0.05em;
    text-transform: uppercase;
}

.brand-badges {
    display: flex;
    gap: 8px;
}

.tech-badge {
    display: inline-flex;
    align-items: center;
    gap: 6px;
    padding: 6px 12px;
    border-radius: 9999px;
    font-size: 0.75rem;
    font-weight: 600;
    font-family: 'JetBrains Mono', monospace;
    background: rgba(255, 255, 255, 0.05);
    border: 1px solid var(--iris-border);
    color: #cbd5e1;
}

.badge-pulse {
    width: 8px;
    height: 8px;
    border-radius: 50%;
    background: var(--iris-emerald);
    box-shadow: 0 0 10px var(--iris-emerald);
    animation: pulseGlow 1.8s infinite;
}

@keyframes pulseGlow {
    0%, 100% { transform: scale(1); opacity: 1; }
    50% { transform: scale(1.3); opacity: 0.7; }
}

/* Tabs */
.tabs {
    border: none !important;
    background: transparent !important;
}

.tab-nav {
    border: none !important;
    background: rgba(15, 20, 32, 0.6) !important;
    padding: 6px !important;
    border-radius: 14px !important;
    margin-bottom: 20px !important;
    border: 1px solid var(--iris-border) !important;
    display: flex !important;
    gap: 6px !important;
}

.tab-nav button {
    border-radius: 10px !important;
    font-weight: 600 !important;
    font-size: 0.9rem !important;
    padding: 10px 18px !important;
    color: #94a3b8 !important;
    transition: all 0.2s ease !important;
    border: none !important;
    background: transparent !important;
}

.tab-nav button.selected {
    background: linear-gradient(135deg, rgba(99, 102, 241, 0.25), rgba(6, 182, 212, 0.2)) !important;
    color: #ffffff !important;
    border: 1px solid rgba(99, 102, 241, 0.4) !important;
    box-shadow: 0 4px 12px rgba(0, 0, 0, 0.3) !important;
}

/* Glassmorphic Panels */
.block {
    background: var(--iris-card) !important;
    border: 1px solid var(--iris-border) !important;
    border-radius: 18px !important;
    backdrop-filter: blur(12px) !important;
}

/* Urgent Instruction Banner */
.instruction-box textarea {
    font-size: 1.35rem !important;
    font-weight: 800 !important;
    line-height: 1.4 !important;
    color: #38bdf8 !important;
    background: linear-gradient(135deg, #0c1527, #131c31) !important;
    border: 2px solid rgba(56, 189, 248, 0.4) !important;
    border-radius: 14px !important;
    text-align: center !important;
    box-shadow: 0 0 25px rgba(56, 189, 248, 0.15) !important;
    padding: 16px !important;
}

/* Action Buttons */
.btn-primary-action {
    background: linear-gradient(135deg, #6366f1, #4f46e5) !important;
    color: white !important;
    font-weight: 700 !important;
    font-size: 1rem !important;
    border: 1px solid rgba(255, 255, 255, 0.15) !important;
    border-radius: 12px !important;
    padding: 12px 20px !important;
    box-shadow: 0 4px 16px rgba(99, 102, 241, 0.35) !important;
    transition: all 0.2s ease !important;
}

.btn-primary-action:hover {
    transform: translateY(-1px) !important;
    box-shadow: 0 6px 20px rgba(99, 102, 241, 0.5) !important;
}

/* Live Navigation Toggle Button */
#live_nav_toggle_btn, .live-stream-btn {
    width: 100% !important;
    padding: 14px 20px !important;
    border: 1px solid rgba(255, 255, 255, 0.15) !important;
    border-radius: 12px !important;
    background: linear-gradient(135deg, #10b981, #059669) !important;
    color: white !important;
    font-weight: 700 !important;
    font-size: 1.05rem !important;
    cursor: pointer !important;
    box-shadow: 0 4px 16px rgba(16, 185, 129, 0.35) !important;
    transition: all 0.2s ease !important;
}

#live_nav_toggle_btn:hover {
    filter: brightness(1.1) !important;
    transform: translateY(-1px) !important;
}

#live_nav_toggle_btn.active {
    background: linear-gradient(135deg, #ef4444, #dc2626) !important;
    box-shadow: 0 4px 20px rgba(239, 68, 68, 0.5) !important;
    animation: livePulse 2s infinite;
}

@keyframes livePulse {
    0%, 100% { box-shadow: 0 0 15px rgba(239, 68, 68, 0.5); }
    50% { box-shadow: 0 0 25px rgba(239, 68, 68, 0.8); }
}

/* Spatial Radar HUD */
.hud-zones-bar {
    display: grid;
    grid-template-columns: repeat(3, 1fr);
    gap: 10px;
    margin-bottom: 14px;
}

.hud-zone {
    background: rgba(10, 15, 26, 0.7);
    border: 1px solid var(--iris-border);
    border-radius: 12px;
    padding: 10px 12px;
    transition: all 0.2s ease;
}

.hud-zone-header {
    display: flex;
    align-items: center;
    gap: 6px;
    margin-bottom: 4px;
}

.hud-zone-dot {
    width: 7px;
    height: 7px;
    border-radius: 50%;
}

.hud-zone-name {
    font-size: 0.7rem;
    font-weight: 700;
    letter-spacing: 0.05em;
    color: #94a3b8;
}

.hud-zone-val {
    font-family: 'JetBrains Mono', monospace;
    font-size: 0.88rem;
    font-weight: 700;
}

.val-clear { color: #10b981; }
.val-warn  { color: #f59e0b; }
.val-alert { color: #ef4444; }

/* Detection Chips */
.hud-breakdown {
    background: rgba(10, 15, 26, 0.5);
    border: 1px solid var(--iris-border);
    border-radius: 14px;
    padding: 12px 14px;
}

.hud-subhead {
    font-size: 0.7rem;
    font-weight: 700;
    letter-spacing: 0.08em;
    color: #64748b;
    margin-bottom: 8px;
}

.chips-flex {
    display: flex;
    flex-wrap: wrap;
    gap: 8px;
}

.det-chip {
    display: inline-flex;
    align-items: center;
    gap: 8px;
    background: rgba(255, 255, 255, 0.04);
    border: 1px solid var(--iris-border);
    border-radius: 8px;
    padding: 6px 10px;
    font-size: 0.8rem;
}

.det-indicator {
    width: 6px;
    height: 6px;
    border-radius: 50%;
}

.det-title {
    font-weight: 600;
    color: #f1f5f9;
}

.det-dir {
    font-family: 'JetBrains Mono', monospace;
    font-size: 0.7rem;
    font-weight: 700;
}

.det-score {
    font-family: 'JetBrains Mono', monospace;
    font-size: 0.72rem;
    color: #94a3b8;
}

.all-clear-banner {
    display: flex;
    align-items: center;
    gap: 8px;
    color: #10b981;
    font-size: 0.85rem;
    font-weight: 600;
    padding: 6px 0;
}

.check-icon {
    display: inline-flex;
    align-items: center;
    justify-content: center;
    width: 18px;
    height: 18px;
    border-radius: 50%;
    background: rgba(16, 185, 129, 0.2);
    font-size: 0.75rem;
}
</style>
"""

# ── Build Gradio Interface ───────────────────────────────────────────────────
with gr.Blocks(title="IRIS — Intelligent Visual Copilot") as demo:

    gr.HTML(custom_css)

    # Top Brand Bar
    gr.HTML("""
    <div class="iris-brand-hero">
        <div class="brand-left">
            <div class="brand-logo-hex">
                <svg width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="white" stroke-width="2.5">
                    <circle cx="12" cy="12" r="3"></circle>
                    <path d="M2.05 12a10 10 0 0 1 19.9 0 10 10 0 0 1-19.9 0Z"></path>
                </svg>
            </div>
            <div class="brand-text">
                <h1>IRIS VISUAL COPILOT</h1>
                <p>Real-Time Spatial Obstacle Detection & Voice Guidance</p>
            </div>
        </div>
        <div class="brand-badges">
            <div class="tech-badge">
                <span class="badge-pulse"></span>
                <span>SYSTEM ONLINE</span>
            </div>
            <div class="tech-badge">
                <span>YOLO11-NANO</span>
            </div>
            <div class="tech-badge">
                <span>3-ZONE RADAR</span>
            </div>
        </div>
    </div>
    """)

    with gr.Tabs():

        # ── TAB 1: Real-Time Live Vision ─────────────────────────────────────
        with gr.TabItem("📷 Live Visual Navigation", id="tab_live"):
            with gr.Row():
                with gr.Column(scale=5):
                    webcam_input = gr.Image(
                        sources=["webcam"],
                        type="numpy",
                        label="Live Optical Sensor",
                        elem_id="webcam_viewport",
                    )
                    
                    webcam_b64 = gr.Textbox(visible=False, elem_id="webcam_b64_buffer")

                    # Native Gradio button for live navigation toggle
                    live_nav_btn = gr.Button(
                        "▶️ Start Real-Time Navigation (Sequential Voice)",
                        elem_id="live_nav_toggle_btn",
                        elem_classes=["live-stream-btn"],
                        variant="primary",
                    )

                    gr.HTML("""
                    <p id="live_status_indicator" style="text-align: center; font-size: 0.82rem; color: #94a3b8; margin: 6px 0 12px 0;">
                        <span style="color:#64748b;">● STANDBY</span> — Click above to begin continuous scanning.
                    </p>
                    """)

                    webcam_snap_btn = gr.Button(
                        "⚡ Instant Scan (Current Frame)",
                        elem_id="scan_frame_btn",
                        elem_classes=["btn-primary-action"],
                        variant="secondary",
                    )

                with gr.Column(scale=6):
                    webcam_instruction = gr.Textbox(
                        label="📢 Priority Voice Navigation Alert",
                        interactive=False,
                        elem_id="webcam_instruction",
                        elem_classes=["instruction-box"],
                        value="STANDBY — Waiting for visual input.",
                    )
                    
                    webcam_audio = gr.Audio(
                        label="🔊 Audio Recording (Optional)",
                        autoplay=False,
                    )
                    
                    webcam_hud = gr.HTML(
                        label="Spatial Radar Breakdown",
                        value="""
                        <div class="hud-zones-bar">
                            <div class="hud-zone"><div class="hud-zone-val val-clear">STANDBY</div></div>
                        </div>
                        """,
                    )
                    
                    webcam_output = gr.Image(
                        label="Annotated Spatial View",
                    )

            # Wire up the instant scan button (extracts live frame via JS)
            # When inference completes, .then() triggers sequential voice & next scan
            webcam_snap_btn.click(
                fn=analyze_webcam_input,
                inputs=[webcam_b64, webcam_input],
                outputs=[webcam_output, webcam_instruction, webcam_audio, webcam_hud],
                js=js_extract_video,
            ).then(
                fn=None,
                inputs=[webcam_instruction],
                js="""(inst) => { if (window.onFrameOutputReceived) { window.onFrameOutputReceived(inst); } }""",
            )

            # Wire up the live navigation button to toggle the continuous loop
            live_nav_btn.click(
                fn=None,
                js=live_nav_js,
            )

        # ── TAB 2: Image Diagnostic ──────────────────────────────────────────
        with gr.TabItem("🖼️ Photo Inspection", id="tab_photo"):
            with gr.Row():
                with gr.Column(scale=5):
                    upload_input = gr.Image(
                        sources=["upload"],
                        type="numpy",
                        label="Upload Scene Photograph",
                    )
                    upload_btn = gr.Button(
                        "🔍 Analyze Scene Obstacles",
                        elem_classes=["btn-primary-action"],
                        variant="primary",
                    )

                with gr.Column(scale=6):
                    upload_instruction = gr.Textbox(
                        label="📢 Spoken Navigation Instruction",
                        interactive=False,
                        elem_classes=["instruction-box"],
                    )
                    upload_audio = gr.Audio(
                        label="🔊 Audio Instruction",
                        autoplay=True,
                    )
                    upload_hud = gr.HTML(
                        label="Spatial Radar Breakdown",
                    )
                    upload_output = gr.Image(
                        label="Annotated Spatial View",
                    )

            upload_btn.click(
                fn=analyze_frame,
                inputs=[upload_input],
                outputs=[upload_output, upload_instruction, upload_audio, upload_hud],
            )

        # ── TAB 3: Video Walkthrough ─────────────────────────────────────────
        with gr.TabItem("🎥 Video Walkthrough", id="tab_video"):
            with gr.Row():
                with gr.Column(scale=5):
                    video_input = gr.Video(
                        sources=["upload"],
                        label="Upload Navigation Footage (.mp4, .mov)",
                    )
                    video_btn = gr.Button(
                        "🎬 Analyze Navigation Corridor",
                        elem_classes=["btn-primary-action"],
                        variant="primary",
                    )

                with gr.Column(scale=6):
                    video_audio = gr.Audio(
                        label="🔊 Spoken Overview",
                        autoplay=True,
                    )
                    video_timeline = gr.Markdown(
                        label="Chronological Timeline Guidance",
                    )
                    video_gallery = gr.Gallery(
                        label="Corridor Keyframe Snapshots",
                        columns=2,
                    )

            video_btn.click(
                fn=analyze_video,
                inputs=[video_input],
                outputs=[video_gallery, video_timeline, video_audio],
            )

        # ── TAB 4: System Architecture ───────────────────────────────────────
        with gr.TabItem("ℹ️ System Architecture", id="tab_about"):
            gr.HTML("""
            <div style="background: rgba(15, 20, 32, 0.6); border: 1px solid rgba(255, 255, 255, 0.08); border-radius: 16px; padding: 24px;">
                <h2 style="color: #f1f5f9; font-size: 1.3rem; margin-bottom: 12px;">🧭 IRIS Navigation Engine Principles</h2>
                
                <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(280px, 1fr)); gap: 16px; margin-top: 16px;">
                    <div style="background: rgba(255, 255, 255, 0.03); border: 1px solid rgba(255, 255, 255, 0.06); border-radius: 12px; padding: 16px;">
                        <h3 style="color: #38bdf8; font-size: 1rem; margin-bottom: 6px;">1. 3-Zone Spatial Radar</h3>
                        <p style="color: #94a3b8; font-size: 0.85rem; line-height: 1.5; margin: 0;">
                            The sensor field is dynamically partitioned into <b>LEFT (33%)</b>, <b>CENTER PATH (33%)</b>, and <b>RIGHT (33%)</b>. Obstacles located directly in the center path receive critical priority.
                        </p>
                    </div>

                    <div style="background: rgba(255, 255, 255, 0.03); border: 1px solid rgba(255, 255, 255, 0.06); border-radius: 12px; padding: 16px;">
                        <h3 style="color: #10b981; font-size: 1rem; margin-bottom: 6px;">2. Zero-Cognitive-Overload Priority</h3>
                        <p style="color: #94a3b8; font-size: 0.85rem; line-height: 1.5; margin: 0;">
                            Instead of overwhelming a visually impaired person with a list of 10 items, IRIS selects <b>exactly ONE</b> urgent, actionable instruction (e.g. <i>"Caution! Person directly ahead"</i> or <i>"Watch out — chair ahead. Step around it"</i>).
                        </p>
                    </div>

                    <div style="background: rgba(255, 255, 255, 0.03); border: 1px solid rgba(255, 255, 255, 0.06); border-radius: 12px; padding: 16px;">
                        <h3 style="color: #a855f7; font-size: 1rem; margin-bottom: 6px;">3. Dual-Stream Voice Guidance</h3>
                        <p style="color: #94a3b8; font-size: 0.85rem; line-height: 1.5; margin: 0;">
                            Low-latency browser speech synthesis provides instant voice cues on device, complemented by server-side synthesized audio with automatic playback.
                        </p>
                    </div>
                </div>
            </div>
            """)

if __name__ == "__main__":
    demo.launch()