still initial commit

app.py

@@ -3,7 +3,7 @@ import cv2
 import numpy as np
 import gradio as gr
 import modal
-from fastrtc import WebRTC, get_cloudflare_turn_credentials
+from fastrtc import Stream, get_cloudflare_turn_credentials
 
 # ── Modal Backend ───────────────────────────────────────────────────────────
 try:
@@ -18,17 +18,28 @@ except Exception as e:
         print(f"❌ Modal backend offline: {ex}")
         voxel_backend = None
 
+status_color = "🟢" if voxel_backend is not None else "🔴"
+status_text  = "Connected" if voxel_backend is not None else "Offline"
+
 
 # ── Helpers ─────────────────────────────────────────────────────────────────
-def _run_voxel_backend(frame: np.ndarray, prompt: str, strength: float) -> np.ndarray:
+def _offline_frame(frame: np.ndarray, message: str) -> np.ndarray:
+    """Generates a styled placeholder frame when backend is offline."""
+    h, w = (frame.shape[:2] if frame is not None else (480, 640))
+    out = np.zeros((h, w, 3), dtype=np.uint8)
+    cv2.putText(out, message, (max(10, w // 8), h // 2),
+                cv2.FONT_HERSHEY_DUPLEX, 0.8, (0, 0, 220), 2)
+    return out
+
+
+def _run_voxel_backend(frame: np.ndarray) -> np.ndarray:
+    """Encodes and ships raw image bytes to the Modal worker without extra parameters."""
     success, encoded = cv2.imencode(".jpg", frame, [int(cv2.IMWRITE_JPEG_QUALITY), 85])
     if not success:
         return frame
     try:
-        try:
-            processed_bytes = voxel_backend.remote(encoded.tobytes(), prompt, strength)
-        except TypeError:
-            processed_bytes = voxel_backend.remote(encoded.tobytes())
+        # Prompt and strength arguments have been fully stripped here
+        processed_bytes = voxel_backend.remote(encoded.tobytes())
         result = cv2.imdecode(np.frombuffer(processed_bytes, dtype=np.uint8), cv2.IMREAD_COLOR)
         return result if result is not None else frame
     except Exception as err:
@@ -38,134 +49,68 @@ def _run_voxel_backend(frame: np.ndarray, prompt: str, strength: float) -> np.nd
         return out
 
 
-def _offline_frame(frame: np.ndarray, message: str) -> np.ndarray:
-    h, w = (frame.shape[:2] if frame is not None else (480, 640))
-    out = np.zeros((h, w, 3), dtype=np.uint8)
-    cv2.putText(out, message, (max(10, w // 8), h // 2),
-                cv2.FONT_HERSHEY_DUPLEX, 0.8, (0, 0, 220), 2)
-    return out
-
-
-# ── Stream Handlers ─────────────────────────────────────────────────────────
-def process_video_stream(frame: np.ndarray, prompt: str, strength: float) -> np.ndarray:
+# ── Unified Core Stream Handler ─────────────────────────────────────────────
+def process_video_stream(frame: np.ndarray, mode: str) -> np.ndarray:
+    """
+    Unified real-time handler mapping directly to FastRTC's stream pipeline.
+    Accepts incoming frame along with the UI's selected operating mode.
+    """
     if frame is None:
         return None
-    if voxel_backend is None:
-        return _offline_frame(frame, "Modal Backend Offline")
-    return _run_voxel_backend(frame, prompt, strength)
-
 
-def process_demo_stream(frame: np.ndarray) -> np.ndarray:
-    if frame is None:
-        return None
     if voxel_backend is None:
-        err = _offline_frame(frame, "Backend Offline")
-        return np.hstack([frame, err])
-    processed = _run_voxel_backend(
-        frame,
-        prompt="vanilla minecraft voxel landscape, 3d blocky style, cube aesthetic",
-        strength=0.55,
-    )
-    if processed.shape[0] != frame.shape[0]:
-        processed = cv2.resize(processed, (frame.shape[1], frame.shape[0]))
-    raw_labeled = frame.copy()
-    cv2.putText(raw_labeled, "RAW",       (10, 28), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (255, 255, 255), 2)
-    cv2.putText(processed,   "MINECRAFT", (10, 28), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (255, 255, 255), 2)
-    return np.hstack([raw_labeled, processed])
-
-
-# ── CSS: official Gradio WebRTC pattern (from gradio.app/guides) ───────────
-# Target gr.Group and gr.Column via elem_classes — NOT the WebRTC component
-css = """
-.webrtc-group  { max-width: 600px  !important; max-height: 600px !important; }
-.webrtc-col    { display: flex !important; justify-content: center !important; align-items: center !important; }
-.demo-group    { max-width: 1000px !important; max-height: 600px !important; }
-"""
-
-status_color = "🟢" if voxel_backend is not None else "🔴"
-status_text  = "Connected" if voxel_backend is not None else "Offline"
-
-# ── UI ───────────────────────────────────────────────────────────────────────
-with gr.Blocks(title="Minecraft Spatial Voxel Filter", css=css) as demo:
-
-    gr.HTML("""
-    <div style='text-align:center; padding: 16px 0'>
-        <h1 style='margin:0'>⛏️ MINECRAFT SPATIAL VOXEL FILTER ⛏️</h1>
-        <p style='margin:6px 0 0 0'>Transform your environment into a real-time blocky landscape</p>
-    </div>
-    """)
-
-    # ── Shared controls (always visible) ─────────────────────────────────────
-    with gr.Row():
-        mode_dropdown = gr.Dropdown(
+        err_frame = _offline_frame(frame, "Modal Backend Offline")
+        if mode == "Streaming Demo":
+            return np.hstack([frame, err_frame])
+        return err_frame
+
+    # Process via the lightweight single-image pipeline
+    processed = _run_voxel_backend(frame)
+
+    # Mode A: Full view rendering
+    if mode == "Minecraft Filter":
+        return processed
+
+    # Mode B: Side-by-side split rendering
+    elif mode == "Streaming Demo":
+        if processed.shape[0] != frame.shape[0]:
+            processed = cv2.resize(processed, (frame.shape[1], frame.shape[0]))
+            
+        raw_labeled = frame.copy()
+        cv2.putText(raw_labeled, "RAW", (10, 28), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (255, 255, 255), 2)
+        cv2.putText(processed, "MINECRAFT", (10, 28), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (255, 255, 255), 2)
+        return np.hstack([raw_labeled, processed])
+
+    return processed
+
+
+# ── Setup Stream and Auto-UI Generation ──────────────────────────────────────
+stream = Stream(
+    handler=process_video_stream,
+    modality="video",
+    mode="send-receive",
+    rtc_configuration=get_cloudflare_turn_credentials(),
+    additional_inputs=[
+        gr.Dropdown(
             choices=["Minecraft Filter", "Streaming Demo"],
             value="Minecraft Filter",
             label="🎯 Pipeline Mode",
             interactive=True,
+        ),
+        gr.Markdown(
+            f"### ⚡ Backend Connection Status\n"
+            f"Status: {status_color} **{status_text}**\n\n"
+            f"Switching between pipeline modes instantly updates your active viewport feed below."
         )
-        prompt_input = gr.Textbox(
-            value="vanilla minecraft voxel landscape, 3d blocky style, retro game cube aesthetic",
-            label="Biome Blueprint",
-            lines=1,
-        )
-        denoise_strength = gr.Slider(
-            minimum=0.1, maximum=1.0, step=0.05, value=0.55,
-            label="Denoising Strength",
-        )
-        gr.Markdown(f"**Backend:** {status_color} `{status_text}`")
-
-    # ── Mode 1: Minecraft Filter ──────────────────────────────────────────────
-    # OFFICIAL PATTERN: gr.Column(elem_classes) → gr.Group(elem_classes) → WebRTC
-    with gr.Column(elem_classes=["webrtc-col"], visible=True) as minecraft_layout:
-        gr.Markdown("### 📺 Spatial Render Pipeline")
-        with gr.Group(elem_classes=["webrtc-group"]):
-            webrtc_single = WebRTC(
-                label="Live Voxel Viewport",
-                modality="video",
-                mode="send-receive",
-                rtc_configuration=get_cloudflare_turn_credentials,
-            )
-
-    # ── Mode 2: Streaming Demo ────────────────────────────────────────────────
-    with gr.Column(elem_classes=["webrtc-col"], visible=False) as demo_layout:
-        gr.Markdown("### 📺 Dual-Feed · _Left: Raw · Right: Minecraft_")
-        with gr.Group(elem_classes=["demo-group"]):
-            webrtc_demo = WebRTC(
-                label="Live Side-by-Side Feed",
-                modality="video",
-                mode="send-receive",
-                rtc_configuration=get_cloudflare_turn_credentials,
-            )
-
-    # ── Mode Switch ───────────────────────────────────────────────────────────
-    def switch_layout(selected_mode):
-        return (
-            gr.update(visible=selected_mode == "Minecraft Filter"),
-            gr.update(visible=selected_mode == "Streaming Demo"),
-        )
-
-    mode_dropdown.change(
-        fn=switch_layout,
-        inputs=[mode_dropdown],
-        outputs=[minecraft_layout, demo_layout],
-    )
-
-    # ── Stream Bindings ───────────────────────────────────────────────────────
-    webrtc_single.stream(
-        fn=process_video_stream,
-        inputs=[webrtc_single, prompt_input, denoise_strength],
-        outputs=[webrtc_single],
-        time_limit=150,
-        concurrency_limit=4,
-    )
-
-    webrtc_demo.stream(
-        fn=process_demo_stream,
-        inputs=[webrtc_demo],
-        outputs=[webrtc_demo],
-        time_limit=150,
-        concurrency_limit=4,
-    )
+    ],
+    ui_args={
+        "title": "⛏️ Minecraft Spatial Voxel Filter",
+        "pulse_color": "rgb(40, 167, 69)" if voxel_backend else "rgb(220, 53, 69)",
+        "icon_button_color": "rgb(40, 167, 69)" if voxel_backend else "rgb(220, 53, 69)",
+    },
+    time_limit=150,
+    concurrency_limit=4,
+)
 
 if __name__ == "__main__":
-    demo.launch()
+    stream.ui.launch()