app.py

import io, uuid, base64, requests, random
from PIL import Image
import gradio as gr

# ====== CONFIG ======
# Your Modal endpoint (POST with multipart body; prompt/seed in querystring)
MODAL_URL = "https://moonmath-ai--moonmath-i2v-backend-moonmathinference-run.modal.run"
REQUEST_TIMEOUT_SEC = 600  # adjust if your backend needs longer

# ====== BACKEND CALLER ======
def call_modal_backend(prompt: str, image: Image.Image | None, seed: int | None):
    """
    Sends prompt + optional image to the Modal backend.
    Accepts:
      - raw MP4 bytes response
      - JSON with video_url or base64 video
    Returns a path or URL usable by gr.Video.
    """
    if not prompt and image is None:
        raise gr.Error("Please provide a prompt or upload an image.")

    # Build multipart body if image provided
    files = None
    if image is not None:
        buf = io.BytesIO()
        image.save(buf, format="PNG")  # change to JPEG if your backend expects it
        buf.seek(0)
        files = {"image_bytes": ("input.png", buf, "image/png")}

    # Query string params
    params = {}
    if prompt:
        params["prompt"] = prompt
    if seed is not None:
        params["seed"] = str(seed)

    # Perform request
    res = requests.post(
        MODAL_URL,
        params=params,
        files=files,
        headers={"accept": "application/json"},
        timeout=REQUEST_TIMEOUT_SEC,
    )
    res.raise_for_status()

    ctype = (res.headers.get("content-type") or "").lower()

    # 1) Raw MP4 bytes directly
    if "video/mp4" in ctype or ctype.startswith("application/octet-stream"):
        mp4_path = f"out_{uuid.uuid4().hex[:8]}.mp4"
        with open(mp4_path, "wb") as f:
            f.write(res.content)
        return mp4_path

    # 2) JSON (URL or base64)
    if "application/json" in ctype:
        data = res.json()
        url = data.get("video_url") or data.get("url") or data.get("result", {}).get("video_url")
        if url:
            return url  # gr.Video can stream a URL

        b64 = (
            data.get("video_b64")
            or data.get("video_bytes")
            or data.get("result", {}).get("video_b64")
        )
        if b64:
            if "," in b64:  # strip data: header if present
                b64 = b64.split(",", 1)[1]
            blob = base64.b64decode(b64)
            mp4_path = f"out_{uuid.uuid4().hex[:8]}.mp4"
            with open(mp4_path, "wb") as f:
                f.write(blob)
            return mp4_path

        raise gr.Error(f"Backend JSON did not contain a video field. Keys: {list(data.keys())}")

    # 3) Fallback: write bytes as mp4
    mp4_path = f"out_{uuid.uuid4().hex[:8]}.mp4"
    with open(mp4_path, "wb") as f:
        f.write(res.content)
    return mp4_path

# ====== UI CALLBACK ======
def on_generate(prompt, image, seed, lock_longshot):
    """
    lock_longshot is included so you can later inject constraints server-side if needed.
    For now it simply forwards prompt & image to your Modal backend.
    """
    # If user left seed blank, generate one
    if seed is None or str(seed).strip() == "":
        seed_val = random.randint(0, 2**31 - 1)
    else:
        # Gradio Number returns float; cast safely
        try:
            seed_val = int(seed)
        except Exception:
            seed_val = random.randint(0, 2**31 - 1)

    # (Optional) reinforce long-shot constraints in prompt (safe no-op if you don’t need it)
    if lock_longshot and prompt:
        musts = [
            "single continuous long shot",
            "no cuts, no new shot, no angle switch",
            "smooth camera motion (pan/tilt/zoom only)",
            "unbroken continuity"
        ]
        prompt = prompt.strip() + ". " + "; ".join(musts)

    video_path_or_url = call_modal_backend(prompt, image, seed_val)
    info = f"Seed: {seed_val}"
    return video_path_or_url, info

# ====== STYLE ======
CUSTOM_CSS = """
.gradio-container { padding: 24px; }

/* Big rounded prompt box */
#prompt-box textarea {
  border-radius: 28px !important;
  min-height: 180px;
  font-size: 18px;
  line-height: 1.45;
  padding: 18px 22px;
}

/* Rounded square image card */
#add-image .wrap, 
#add-image .input-image,
#add-image .empty {
  border-radius: 28px !important;
  min-width: 240px;
  min-height: 240px;
}

/* Pill generate button */
#gen-btn button {
  border-radius: 999px !important;
  padding: 12px 24px;
  font-size: 18px;
}
"""

# ====== APP ======
with gr.Blocks(css=CUSTOM_CSS, title="Stitch UI – Modal Hook") as demo:
    gr.Markdown("### Stitch – turn prompt/image into a generated video (Modal backend)")

    # Row 1: Big rounded prompt input
    prompt_tb = gr.Textbox(
        label=None,
        placeholder="Prompt input",
        lines=8,
        elem_id="prompt-box"
    )

    # Row 2: Left image card, right controls (seed + generate)
    with gr.Row():
        with gr.Column(scale=1, min_width=300):
            img_in = gr.Image(label="Add Image", type="pil", elem_id="add-image")
        with gr.Column(scale=3, min_width=300):
            with gr.Row():
                seed_in = gr.Number(value=None, label="Seed (optional)")
                lock_long = gr.Checkbox(value=True, label="Lock camera (long shot, no cuts)")
            gen_btn = gr.Button("Generate", elem_id="gen-btn")

    # Output
    with gr.Row():
        video_out = gr.Video(label="Output Video", interactive=False, autoplay=True)
        info_out = gr.Markdown("")

    gen_btn.click(
        fn=on_generate,
        inputs=[prompt_tb, img_in, seed_in, lock_long],
        outputs=[video_out, info_out]
    )

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