"""
app.py — ClearPath: Real-Time Scene Description for Visually-Impaired People
Pipeline: Upload Image → ViT-GPT2 Caption → Regex Safety Classifier → SAFE / DANGEROUS
"""

import gradio as gr
import numpy as np
import logging
import time
import cv2

from PIL import Image
from scene_captioner   import SceneCaptioner
from safety_classifier import SafetyClassifier, ClassificationResult

logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
logger = logging.getLogger(__name__)

# ── Load pipeline once at startup ─────────────────────────────────────────────
logger.info("🚀 Starting ClearPath — loading captioner …")
captioner  = SceneCaptioner()
classifier = SafetyClassifier()
logger.info(f"✅ Pipeline ready — captioner backend: {captioner._backend}")

history_log: list[dict] = []

# ── Core pipeline function ────────────────────────────────────────────────────

def analyse(image: np.ndarray):
    """
    Main pipeline:
      1. Convert numpy array → PIL Image
      2. SceneCaptioner.describe() → caption string
      3. SafetyClassifier.classify() → SAFE / DANGEROUS
      4. Return results to Gradio UI
    """
    if image is None:
        return (
            _info_html("⬆️ Please upload an image first.", "#6366f1"),
            "",
            _build_history_md(),
        )

    t0  = time.time()
    pil = Image.fromarray(image).convert("RGB")

    # ── Step 2: Caption ───────────────────────────────────────────────────────
    try:
        caption = captioner.describe(pil)
    except Exception as exc:
        logger.error(f"Caption error: {exc}")
        caption = "Unable to generate caption for this image."

    # ── Step 3: Classify ──────────────────────────────────────────────────────
    result  = classifier.classify(caption)
    elapsed = round(time.time() - t0, 2)

    # ── Build banner HTML ─────────────────────────────────────────────────────
    if result.label == "DANGEROUS":
        hazard_str  = "  |  ".join(result.hazards)
        token_str   = ", ".join(result.matches[:8])
        banner_html = f"""
        <div style="
            background:rgba(239,68,68,0.12);
            border:2px solid rgba(239,68,68,0.45);
            border-radius:14px; padding:1.1rem 1.4rem;
            display:flex; align-items:flex-start; gap:1rem;
            animation: fadeIn .3s ease;
        ">
          <span style="font-size:2.5rem; line-height:1;">⚠️</span>
          <div>
            <div style="font-weight:800; font-size:1.15rem; color:#fca5a5;
                        letter-spacing:.04em; margin-bottom:.3rem;">
              DANGER DETECTED
            </div>
            <div style="font-size:.85rem; color:#f87171; margin-bottom:.25rem;">
              <strong>Categories:</strong> {hazard_str}
            </div>
            <div style="font-size:.75rem; color:#94a3b8; font-family:monospace;">
              <strong>Matched tokens:</strong> {token_str}
            </div>
            <div style="font-size:.7rem; color:#64748b; margin-top:.3rem;">
              ⏱ Analysed in {elapsed}s &nbsp;|&nbsp; Backend: {captioner._backend}
            </div>
          </div>
        </div>"""
    else:
        banner_html = f"""
        <div style="
            background:rgba(34,197,94,0.1);
            border:2px solid rgba(34,197,94,0.4);
            border-radius:14px; padding:1.1rem 1.4rem;
            display:flex; align-items:flex-start; gap:1rem;
        ">
          <span style="font-size:2.5rem; line-height:1;">✅</span>
          <div>
            <div style="font-weight:800; font-size:1.15rem; color:#86efac;
                        letter-spacing:.04em; margin-bottom:.3rem;">
              SAFE ENVIRONMENT
            </div>
            <div style="font-size:.85rem; color:#4ade80;">
              No hazards detected by the 16-category regex engine.
            </div>
            <div style="font-size:.7rem; color:#64748b; margin-top:.3rem;">
              ⏱ Analysed in {elapsed}s &nbsp;|&nbsp; Backend: {captioner._backend}
            </div>
          </div>
        </div>"""

    # ── Log to history ────────────────────────────────────────────────────────
    history_log.insert(0, {
        "time"   : time.strftime("%H:%M:%S"),
        "label"  : result.label,
        "hazards": ", ".join(result.hazards) if result.hazards else "—",
        "caption": caption,
    })

    return banner_html, caption, _build_history_md()


def _info_html(msg: str, color: str) -> str:
    return (
        f'<div style="background:rgba(99,102,241,.08);border:1px solid {color}33;'
        f'border-radius:12px;padding:1rem 1.25rem;color:#94a3b8;font-size:.9rem;">'
        f'{msg}</div>'
    )


def _build_history_md() -> str:
    if not history_log:
        return "_No analyses yet — upload an image above._"
    rows = ["| Time | Result | Hazards | Caption |",
            "|------|--------|---------|---------|"]
    for h in history_log[:10]:
        short = (h["caption"][:70] + "…") if len(h["caption"]) > 70 else h["caption"]
        icon  = "⚠️" if h["label"] == "DANGEROUS" else "✅"
        rows.append(f"| `{h['time']}` | {icon} **{h['label']}** | {h['hazards']} | {short} |")
    return "\n".join(rows)


# ── Custom CSS ────────────────────────────────────────────────────────────────
CSS = """
@import url('https://fonts.googleapis.com/css2?family=DM+Sans:wght@400;600;800&family=JetBrains+Mono:wght@500;700&display=swap');

body, .gradio-container {
    background: #0a0a10 !important;
    color: #e2e8f0 !important;
    font-family: 'DM Sans', sans-serif !important;
}
.gradio-container { max-width: 1100px !important; margin: 0 auto !important; }

gradio-app { background: #0a0a10 !important; }

/* Header */
.app-header {
    text-align: center;
    padding: 2rem 1rem 1.25rem;
    border-bottom: 1px solid rgba(99,102,241,.2);
    margin-bottom: 1.25rem;
    background: linear-gradient(180deg,rgba(99,102,241,.07) 0%,transparent 100%);
}
.app-title {
    font-size: 2.5rem; font-weight: 800; letter-spacing: -.03em; margin: 0;
    background: linear-gradient(135deg,#a5b4fc,#e879f9);
    -webkit-background-clip: text; -webkit-text-fill-color: transparent;
}
.app-sub { color: #64748b; font-size: .9rem; margin-top: .4rem; }

/* Pipeline bar */
.pipe-bar {
    display: flex; align-items: center; justify-content: center;
    flex-wrap: wrap; gap: .4rem;
    padding: .75rem; margin-bottom: 1.25rem;
    background: rgba(99,102,241,.04);
    border: 1px solid rgba(99,102,241,.15); border-radius: 12px;
    font-family: 'JetBrains Mono', monospace; font-size: .75rem;
}
.pipe-node {
    background: rgba(99,102,241,.14); border: 1px solid rgba(99,102,241,.3);
    color: #a5b4fc; padding: .25rem .75rem; border-radius: 7px; font-weight: 700;
}
.pipe-arrow { color: #334155; font-size: .9rem; }

/* Panels */
.gr-block, .gr-box, .panel {
    background: #13131e !important;
    border: 1px solid rgba(99,102,241,.2) !important;
    border-radius: 14px !important;
}

/* Upload widget */
.gr-image { border-radius: 12px !important; }

/* Caption textbox */
.gr-textbox textarea {
    background: rgba(255,255,255,.03) !important;
    border: 1px solid rgba(99,102,241,.2) !important;
    border-radius: 10px !important;
    color: #e2e8f0 !important;
    font-family: 'DM Sans', sans-serif !important;
    font-size: .95rem !important;
    line-height: 1.75 !important;
}

/* Buttons */
.gr-button-primary, button[variant=primary] {
    background: linear-gradient(135deg,#6366f1,#8b5cf6) !important;
    border: none !important; border-radius: 10px !important;
    color: white !important; font-weight: 700 !important;
    font-family: 'DM Sans', sans-serif !important;
    font-size: .95rem !important;
    transition: opacity .2s !important;
}
.gr-button-primary:hover { opacity: .85 !important; }

/* History table */
.history-box table { width: 100%; border-collapse: collapse; font-size: .8rem; }
.history-box th {
    background: rgba(99,102,241,.1); color: #a5b4fc;
    padding: .4rem .65rem; text-align: left;
    border-bottom: 1px solid rgba(99,102,241,.2);
}
.history-box td {
    padding: .4rem .65rem; color: #64748b;
    border-bottom: 1px solid rgba(255,255,255,.04);
    vertical-align: top;
}

/* Tabs */
.tab-nav button {
    font-family: 'DM Sans', sans-serif !important;
    font-weight: 600 !important; color: #64748b !important;
}
.tab-nav button.selected { color: #a5b4fc !important; }

@keyframes fadeIn { from {opacity:0;transform:translateY(-6px)} to {opacity:1;transform:translateY(0)} }
"""

# ── Build Gradio UI ───────────────────────────────────────────────────────────

def build_ui():
    with gr.Blocks(css=CSS, title="ClearPath — Scene Description") as demo:

        # ── Header ────────────────────────────────────────────────────────────
        gr.HTML("""
        <div class="app-header">
          <h1 class="app-title">👁 ClearPath</h1>
          <p class="app-sub">Real-Time Scene Description for Visually-Impaired People</p>
        </div>
        <div class="pipe-bar">
          <span class="pipe-node">📥 Image Input</span>
          <span class="pipe-arrow">→</span>
          <span class="pipe-node">🧠 ViT-GPT2 / BLIP Captioning</span>
          <span class="pipe-arrow">→</span>
          <span class="pipe-node">🔍 Regex Safety Classifier</span>
          <span class="pipe-arrow">→</span>
          <span class="pipe-node">🏷️ SAFE / DANGEROUS</span>
        </div>
        """)

        with gr.Tabs():

            # ── Tab 1: Image Upload ───────────────────────────────────────────
            with gr.TabItem("📁 Upload Image"):
                with gr.Row():
                    with gr.Column(scale=1):
                        img_input = gr.Image(
                            label="Upload or drag an image",
                            type="numpy",
                            height=300,
                        )
                        analyse_btn = gr.Button(
                            "🔍  Analyse Scene",
                            variant="primary",
                            size="lg",
                        )

                    with gr.Column(scale=1):
                        result_banner = gr.HTML(
                            value='<div style="background:rgba(99,102,241,.06);border:1px solid rgba(99,102,241,.2);'
                                  'border-radius:12px;padding:1.25rem;color:#475569;text-align:center;">'
                                  '⬆️ Upload an image and click <strong>Analyse Scene</strong></div>'
                        )
                        caption_out = gr.Textbox(
                            label="🔊 Scene Description (generated caption)",
                            lines=5,
                            interactive=False,
                            placeholder="The AI-generated scene description will appear here…",
                        )

                analyse_btn.click(
                    fn=analyse,
                    inputs=[img_input],
                    outputs=[result_banner, caption_out, gr.State()],
                )

            # ── Tab 2: Webcam ─────────────────────────────────────────────────
            with gr.TabItem("📷 Webcam"):
                with gr.Row():
                    with gr.Column(scale=1):
                        cam_input = gr.Image(
                            label="Webcam — capture a snapshot",
                            sources=["webcam"],
                            type="numpy",
                            height=300,
                        )
                        cam_btn = gr.Button(
                            "📸  Capture & Analyse",
                            variant="primary",
                            size="lg",
                        )
                    with gr.Column(scale=1):
                        cam_banner  = gr.HTML(
                            value='<div style="background:rgba(99,102,241,.06);border:1px solid rgba(99,102,241,.2);'
                                  'border-radius:12px;padding:1.25rem;color:#475569;text-align:center;">'
                                  '📷 Point your camera and click <strong>Capture & Analyse</strong></div>'
                        )
                        cam_caption = gr.Textbox(
                            label="🔊 Scene Description",
                            lines=5,
                            interactive=False,
                        )

                cam_btn.click(
                    fn=analyse,
                    inputs=[cam_input],
                    outputs=[cam_banner, cam_caption, gr.State()],
                )

            # ── Tab 3: Video ──────────────────────────────────────────────────
            with gr.TabItem("🎬 Video"):
                gr.Markdown("Upload a video — ClearPath samples one frame every N seconds.")
                with gr.Row():
                    vid_input = gr.Video(label="Upload Video")
                    interval  = gr.Slider(1, 10, value=3, step=1, label="Interval (seconds)")
                vid_btn = gr.Button("▶  Analyse Video", variant="primary")
                vid_out = gr.Dataframe(
                    headers=["Frame", "Time (s)", "Label", "Hazards", "Caption"],
                    datatype=["number", "number", "str", "str", "str"],
                    visible=False,
                )

                def analyse_video(path, secs):
                    if path is None:
                        return gr.update(visible=False)
                    cap   = cv2.VideoCapture(path)
                    fps   = cap.get(cv2.CAP_PROP_FPS) or 25
                    step  = max(1, int(fps * secs))
                    rows, idx, n = [], 0, 0
                    while True:
                        ret, frame = cap.read()
                        if not ret:
                            break
                        if idx % step == 0:
                            pil = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
                            try:
                                cap_text = captioner.describe(pil)
                                res      = classifier.classify(cap_text)
                            except Exception as e:
                                cap_text, res = str(e), ClassificationResult("ERROR", [], [])
                            rows.append([n + 1, round(idx / fps, 1),
                                         res.label, ", ".join(res.hazards) or "—", cap_text])
                            n += 1
                        idx += 1
                    cap.release()
                    return gr.update(value=rows, visible=True)

                vid_btn.click(fn=analyse_video, inputs=[vid_input, interval], outputs=[vid_out])

        # ── History ───────────────────────────────────────────────────────────
        with gr.Accordion("📋 Analysis History", open=False):
            history_out = gr.Markdown(
                "_No analyses yet._",
                elem_classes=["history-box"],
            )

        # Wire history refresh on every analyse
        def analyse_with_history(image):
            banner, caption, _ = analyse(image)
            return banner, caption, _build_history_md()

        analyse_btn.click(
            fn=analyse_with_history,
            inputs=[img_input],
            outputs=[result_banner, caption_out, history_out],
        )
        cam_btn.click(
            fn=analyse_with_history,
            inputs=[cam_input],
            outputs=[cam_banner, cam_caption, history_out],
        )

    return demo


if __name__ == "__main__":
    demo = build_ui()
    demo.launch(server_name="0.0.0.0", server_port=7860)