Update app.py

app.py

@@ -1,397 +1,5 @@
-"""
-SanitizeAI - Watermark & Metadata Sanitizer Web Application
-Built with Streamlit and the SanitizeAI core engine.
+import uvicorn
+from api import app
 
-Tabs:
-1. Image Cleaner: Removes EXIF/PNG AI metadata (SD, ComfyUI, Midjourney, C2PA) and invisible AI watermarks.
-2. Video Cleaner: Uses FFmpeg to extract frames, cleans each frame, and stitches back together with audio.
-3. Text Cleaner: Purges hidden zero-width spaces, Unicode tracking markers, homoglyphs, and AI conversational markers.
-"""
-
-import os
-import re
-import sys
-import time
-import tempfile
-import shutil
-import subprocess
-from pathlib import Path
-from io import BytesIO
-
-import streamlit as st
-from PIL import Image
-
-# Import modular Core Engine functions
-from engine import (
-    CleanMode,
-    clean_image,
-    analyze_image_metadata,
-    clean_text_content,
-    get_ffmpeg_path,
-    get_video_fps,
-    FFMPEG_PATH,
-    HAS_DIFFUSION_REMOVER as HAS_REMOVER,
-    sanitize_video_pipeline,
-    calculate_logo_bbox,
-    probe_video_metadata,
-    inpaint_frame_logo,
-)
-
-# Page configuration
-st.set_page_config(
-    page_title="SanitizeAI - Watermark & Metadata Sanitizer",
-    page_icon="🛡️",
-    layout="wide",
-    initial_sidebar_state="expanded"
-)
-
-# Custom Styling
-st.markdown("""
-<style>
-    .main-header {
-        font-size: 2.2rem;
-        font-weight: 700;
-        background: linear-gradient(135deg, #1e293b 0%, #3b82f6 100%);
-        -webkit-background-clip: text;
-        -webkit-text-fill-color: transparent;
-        margin-bottom: 0.2rem;
-    }
-    .sub-header {
-        color: #64748b;
-        font-size: 1.05rem;
-        margin-bottom: 1.5rem;
-    }
-    .card-box {
-        background-color: #f8fafc;
-        border: 1px solid #e2e8f0;
-        border-radius: 10px;
-        padding: 1.2rem;
-        margin-bottom: 1rem;
-    }
-    .badge-success {
-        background-color: #dcfce7;
-        color: #15803d;
-        padding: 4px 10px;
-        border-radius: 12px;
-        font-size: 0.85rem;
-        font-weight: 600;
-    }
-    .badge-warning {
-        background-color: #fef3c7;
-        color: #b45309;
-        padding: 4px 10px;
-        border-radius: 12px;
-        font-size: 0.85rem;
-        font-weight: 600;
-    }
-    .stTabs [data-baseweb="tab-list"] {
-        gap: 8px;
-    }
-    .stTabs [data-baseweb="tab"] {
-        height: 50px;
-        white-space: pre-wrap;
-        border-radius: 8px 8px 0px 0px;
-        gap: 8px;
-        padding-top: 10px;
-        padding-bottom: 10px;
-        font-weight: 600;
-    }
-</style>
-""", unsafe_allow_html=True)
-
-
-# -----------------------------------------------------------------------------
-# SIDEBAR
-# -----------------------------------------------------------------------------
-
-with st.sidebar:
-    st.image("https://img.icons8.com/color/96/shield-clean.png", width=70)
-    st.title("SanitizeAI")
-    st.caption("AI Metadata & Watermark Purger v0.1.29")
-    
-    st.subheader("⚙️ System Status")
-    if FFMPEG_PATH:
-        st.markdown("<span class='badge-success'>✓ FFmpeg Available</span>", unsafe_allow_html=True)
-    else:
-        st.markdown("<span class='badge-warning'>⚠️ FFmpeg Not Found</span>", unsafe_allow_html=True)
-        st.caption("Video frame stitching requires FFmpeg.")
-        
-    if HAS_REMOVER:
-        st.markdown("<span class='badge-success'>✓ Diffusion Engine Ready</span>", unsafe_allow_html=True)
-    else:
-        st.markdown("<span class='badge-warning'>⚠️ PyTorch/Diffusers Limited</span>", unsafe_allow_html=True)
-
-    st.divider()
-    st.subheader("Global Options")
-    keep_standard_meta = st.checkbox("Keep Standard Metadata (Author/Title/DPI)", value=True, help="Retains camera/DPI info while stripping AI tags.")
-    
-    st.info("💡 **Tip**: Use 'Metadata Stripping' for instant lossless cleaning, or 'Deep Watermark Removal' to erase invisible AI watermarks like SynthID.")
-    
-    st.divider()
-    st.subheader("🔌 API Access")
-    st.caption("FastAPI backend is ready for n8n/Make workflows at `/v1/clean` with `X-API-Key`.")
-
-
-# -----------------------------------------------------------------------------
-# MAIN APP BODY
-# -----------------------------------------------------------------------------
-
-st.markdown("<div class='main-header'>🛡️ SanitizeAI - Watermark & Metadata Sanitizer</div>", unsafe_allow_html=True)
-st.markdown("<div class='sub-header'>Remove invisible AI watermarks (SynthID, StableSignature, TreeRing), strip AI metadata (SD, ComfyUI, C2PA), and sanitize AI-generated text and video frames.</div>", unsafe_allow_html=True)
-
-tab1, tab2, tab3 = st.tabs(["🖼️ Image Cleaner", "🎬 Video Cleaner", "📝 Text Cleaner"])
-
-
-# =============================================================================
-# TAB 1: IMAGE CLEANER
-# =============================================================================
-with tab1:
-    st.subheader("Image Metadata & AI Watermark Cleaner")
-    st.markdown("Upload images to inspect and remove AI generation prompts, C2PA digital signatures, and invisible watermarks.")
-    
-    uploaded_files = st.file_uploader("Upload Image(s)", type=["png", "jpg", "jpeg", "webp"], accept_multiple_files=True, key="img_uploader")
-    
-    col_mode, col_opt = st.columns([1, 1])
-    with col_mode:
-        clean_mode = st.radio(
-            "Cleaning Engine Mode",
-            [
-                "⭐ SynthID & Pixel Watermark Destructor (Recommended)",
-                "⭐ Visual Logo Purge + SynthID Destructor (Google Star / Corner Logo)",
-                "Metadata Stripping Only (Fast & Lossless)",
-                "Deep Diffusion Regeneration Attack"
-            ],
-            help="Pixel Destructor alters spatial/frequency noise to break invisible SynthID watermarks. Visual Logo Purge inpaints corner watermarks."
-        )
-    with col_opt:
-        logo_loc_img = "bottom_right"
-        synth_strength_img = 0.5
-        device_choice = "auto"
-        
-        if "Logo" in clean_mode:
-            logo_loc_img = st.selectbox("Visual Logo Location", ["bottom_right", "bottom_left", "all_corners", "top_right", "top_left"], key="img_logo_loc")
-        if "SynthID" in clean_mode or "Logo" in clean_mode:
-            synth_strength_img = st.slider("SynthID Destruction Intensity", min_value=0.1, max_value=1.0, value=0.5, step=0.1, key="img_strength")
-        elif "Deep" in clean_mode:
-            synth_strength_img = st.slider("Denoising Strength", min_value=0.01, max_value=0.20, value=0.04, step=0.01, help="Higher values erase stronger watermarks but slightly modify fine textures.")
-            device_choice = st.selectbox("Inference Device", ["auto", "cpu", "cuda"], index=0)
-        else:
-            st.caption("⚡ Header Metadata mode strips EXIF/PNGinfo/C2PA tags without altering pixels.")
-            
-    if uploaded_files:
-        st.divider()
-        for idx, file_obj in enumerate(uploaded_files):
-            with st.container():
-                st.markdown(f"#### 📄 `{file_obj.name}` ({file_obj.size / 1024:.1f} KB)")
-                
-                with tempfile.TemporaryDirectory() as tmp_dir:
-                    tmp_path = Path(tmp_dir) / file_obj.name
-                    out_path = Path(tmp_dir) / f"sanitized_{file_obj.name}"
-                    
-                    file_bytes = file_obj.getvalue()
-                    with open(tmp_path, "wb") as f:
-                        f.write(file_bytes)
-                        
-                    # Analysis
-                    analysis = analyze_image_metadata(tmp_path)
-                    
-                    c1, c2 = st.columns([1, 1])
-                    with c1:
-                        st.image(str(tmp_path), caption="Original Input", use_container_width=True)
-                        if analysis["detected_keys"]:
-                            st.warning(f"⚠️ Found {analysis['total_fields']} AI Metadata field(s): `{', '.join(analysis['detected_keys'])}`")
-                        else:
-                            st.info("ℹ️ No explicit AI metadata keys detected in file headers.")
-                        if analysis["has_c2pa"]:
-                            st.error("🔒 C2PA Content Credentials signature detected.")
-                            
-                    with c2:
-                        if st.button(f"✨ Clean Image #{idx+1}", key=f"btn_clean_{idx}"):
-                            with st.spinner("Processing image..."):
-                                start_t = time.time()
-                                
-                                # Use modular engine.py
-                                cleaned_bytes, mime_type = clean_image(
-                                    image_input=file_bytes,
-                                    mode=clean_mode,
-                                    intensity=synth_strength_img,
-                                    logo_location=logo_loc_img,
-                                    keep_standard_metadata=keep_standard_meta,
-                                    device=device_choice,
-                                )
-                                out_path.write_bytes(cleaned_bytes)
-                                    
-                                duration = time.time() - start_t
-                                out_size = len(cleaned_bytes)
-                                
-                                st.image(cleaned_bytes, caption="Sanitized Output", use_container_width=True)
-                                st.success(f"✓ Cleaned in {duration:.2f}s | Output size: {out_size / 1024:.1f} KB")
-                                
-                                st.download_button(
-                                    label=f"⬇️ Download Cleaned {file_obj.name}",
-                                    data=cleaned_bytes,
-                                    file_name=f"sanitized_{file_obj.name}",
-                                    mime=mime_type,
-                                    key=f"dl_{idx}"
-                                )
-
-
-# =============================================================================
-# TAB 2: VIDEO CLEANER
-# =============================================================================
-with tab2:
-    st.subheader("Video Frame & Audio Sanitizer")
-    st.markdown("Combines **Visible Watermark Logo Inpainting / Delogo**, **Invisible SynthID Watermark Destruction**, and **EXIF/C2PA Metadata Stripping** into one unified video pipeline.")
-    
-    if not FFMPEG_PATH:
-        st.error("❌ FFmpeg binary was not found on the system. Please ensure `ffmpeg` or `imageio-ffmpeg` is installed.")
-    else:
-        video_file = st.file_uploader("Upload Video File", type=["mp4", "mov", "avi", "mkv", "webm"], key="vid_uploader")
-        
-        st.markdown("### ⚙️ Sanitization Pipeline Settings")
-        
-        c_pipe1, c_pipe2, c_pipe3 = st.columns(3)
-        with c_pipe1:
-            opt_visible_logo = st.checkbox("🎨 Remove Visible Watermark Logo", value=True, help="Inpaints or delogos visual on-screen logos (e.g. Google Veo/Imagen 4-pointed star logo).")
-        with c_pipe2:
-            opt_synthid = st.checkbox("🔮 Remove Invisible SynthID Watermarks", value=True, help="Alters frequency/spatial latent noise to break invisible SynthID signatures.")
-        with c_pipe3:
-            opt_metadata = st.checkbox("🏷️ Remove AI Metadata Headers", value=True, help="Strips EXIF/PNG parameters and C2PA credentials.")
-            
-        v_col1, v_col2 = st.columns([1, 1])
-        with v_col1:
-            logo_region = st.selectbox(
-                "Visible Logo Region Target",
-                [
-                    "Bottom Right Corner (Google Veo / Imagen / Sora)",
-                    "Bottom Left Corner",
-                    "Top Right Corner",
-                    "Top Left Corner",
-                    "Custom Bounding Box (x, y, w, h)"
-                ],
-                index=0,
-                disabled=not opt_visible_logo,
-                help="Targets the region where the visible watermark logo is located."
-            )
-            
-            logo_loc_key = "bottom_right"
-            if "Bottom Left" in logo_region:
-                logo_loc_key = "bottom_left"
-            elif "Top Right" in logo_region:
-                logo_loc_key = "top_right"
-            elif "Top Left" in logo_region:
-                logo_loc_key = "top_left"
-            elif "Custom" in logo_region:
-                logo_loc_key = "custom"
-                
-        with v_col2:
-            preset_quality = st.select_slider("Video Quality (CRF)", options=[18, 20, 23, 26, 28], value=18, help="Lower CRF = higher video quality (18 = visually lossless).")
-            engine_backend = st.radio("Delogo Filter Engine", ["OpenCV Local Inpainter (High Quality)", "FFmpeg Delogo Filter (Ultra Fast)"], index=0, disabled=not opt_visible_logo)
-            backend_key = "ffmpeg_delogo" if "FFmpeg" in engine_backend else "opencv"
-
-        # Custom Bounding Box Controls
-        custom_bbox_val = None
-        if opt_visible_logo:
-            with st.expander("📐 Fine-Tune Bounding Box Coordinates (x, y, width, height)", expanded=(logo_loc_key == "custom")):
-                st.caption("Enter custom pixel coordinates or leave as -1 to use automatic corner calculation.")
-                col_x, col_y, col_w, col_h = st.columns(4)
-                cx = col_x.number_input("X Position (px)", value=-1, step=10)
-                cy = col_y.number_input("Y Position (px)", value=-1, step=10)
-                cw = col_w.number_input("Width (px)", value=-1, step=10)
-                ch = col_h.number_input("Height (px)", value=-1, step=10)
-                if cx >= 0 and cy >= 0 and cw > 0 and ch > 0:
-                    custom_bbox_val = (int(cx), int(cy), int(cw), int(ch))
-            
-        if video_file:
-            st.divider()
-            with tempfile.TemporaryDirectory() as tmp_dir:
-                tmp_dir_p = Path(tmp_dir)
-                in_vid = tmp_dir_p / video_file.name
-                out_vid = tmp_dir_p / f"sanitized_{video_file.name}"
-                
-                with open(in_vid, "wb") as f:
-                    f.write(video_file.getbuffer())
-                    
-                st.video(str(in_vid))
-                
-                if st.button("🚀 Process & Sanitize Video"):
-                    progress_bar = st.progress(0, text="Initializing Video Pipeline...")
-                    
-                    meta_info = probe_video_metadata(in_vid, FFMPEG_PATH)
-                    calc_bbox = calculate_logo_bbox(meta_info["width"], meta_info["height"], location=logo_loc_key, custom_bbox=custom_bbox_val)
-                    
-                    st.caption(f"📹 Video Specs: **{meta_info['width']}x{meta_info['height']} @ {meta_info['fps']:.2f} FPS** | Target Logo Bounding Box `(x={calc_bbox[0]}, y={calc_bbox[1]}, w={calc_bbox[2]}, h={calc_bbox[3]})`")
-                    
-                    def on_progress(pct: int, msg: str):
-                        progress_bar.progress(pct, text=msg)
-
-                    try:
-                        sanitize_video_pipeline(
-                            input_video=in_vid,
-                            output_video=out_vid,
-                            ffmpeg_path=FFMPEG_PATH,
-                            remove_visible_logo=opt_visible_logo,
-                            remove_synthid=opt_synthid,
-                            remove_metadata=opt_metadata,
-                            logo_location=logo_loc_key,
-                            custom_bbox=custom_bbox_val,
-                            engine_backend=backend_key,
-                            crf_quality=preset_quality,
-                            progress_callback=on_progress
-                        )
-                        
-                        progress_bar.progress(100, text="Sanitization Complete!")
-                        
-                        if out_vid.exists():
-                            st.success("🎉 Video successfully sanitized! Clean MP4 container rebuilt & watermarks purged.")
-                            st.video(str(out_vid))
-                            
-                            with open(out_vid, "rb") as vf:
-                                st.download_button(
-                                    label="⬇️ Download Sanitized Video",
-                                    data=vf.read(),
-                                    file_name=f"sanitized_{video_file.name}",
-                                    mime="video/mp4"
-                                )
-                        else:
-                            st.error("Video assembly failed. Check FFmpeg output.")
-                    except Exception as err:
-                        st.error(f"Pipeline error during video processing: {err}")
-
-
-# =============================================================================
-# TAB 3: TEXT CLEANER
-# =============================================================================
-with tab3:
-    st.subheader("AI Text Sanitizer & Hidden Marker Purger")
-    st.markdown("Removes zero-width spaces (`\\u200B`), invisible Unicode tracking markers, homoglyphs, and AI conversational artifacts.")
-    
-    text_input = st.text_area("Paste AI-Generated Text Here", height=200, placeholder="Paste text here to detect and purge hidden zero-width spaces and tracking markers...")
-    
-    t_c1, t_c2 = st.columns([1, 1])
-    with t_c1:
-        opt_ai_phrases = st.checkbox("Remove AI Conversational Introductions/Outros", value=True)
-    with t_c2:
-        st.caption("Detects zero-width spaces, soft hyphens, directionality markers, and Cyrillic homoglyphs.")
-        
-    if text_input.strip():
-        cleaned_text, stats = clean_text_content(text_input, remove_ai_phrases=opt_ai_phrases)
-        
-        st.divider()
-        st.markdown("### 📊 Sanitization Results")
-        
-        m1, m2, m3, m4 = st.columns(4)
-        m1.metric("Characters Before", stats["initial_len"])
-        m2.metric("Characters After", stats["final_len"])
-        m3.metric("Zero-Width Purged", stats["zero_width"], delta=f"-{stats['zero_width']}" if stats["zero_width"] > 0 else None)
-        m4.metric("Homoglyphs / Markers", stats["homoglyphs"] + stats["control_chars"], delta=f"-{stats['homoglyphs']}" if stats["homoglyphs"] > 0 else None)
-        
-        st.markdown("### 📝 Cleaned Text")
-        st.text_area("Sanitized Output", value=cleaned_text, height=200, key="cleaned_txt_output")
-        
-        st.download_button(
-            label="⬇️ Download Sanitized Text (.txt)",
-            data=cleaned_text,
-            file_name="sanitized_text.txt",
-            mime="text/plain"
-        )
+if __name__ == "__main__":
+    uvicorn.run(app, host="0.0.0.0", port=7860)