Update app.py

app.py

@@ -1,223 +1,274 @@
-import gradio as gr
-import requests
+# app.py
+import os
+import time
 import base64
+import json
 from pathlib import Path
+
+import gradio as gr
+import requests
 import jwt
-import time
-import logging
-import os
+from PIL import Image
 
-# Set up logging for debugging
-logging.basicConfig(level=logging.DEBUG)
-logger = logging.getLogger(__name__)
+# ──────────────────────────────────────────────────────────────────────────────
+# CONFIG — set your keys as HF Space secrets or env vars for safety.
+# (Falls back to the keys you shared.)
+# ──────────────────────────────────────────────────────────────────────────────
+ACCESS_KEY_ID = os.getenv("KLING_ACCESS_KEY_ID", "AGBGmadNd9hakFYfahytyQQJtN8CJmDJ")
+ACCESS_KEY_SECRET = os.getenv("KLING_ACCESS_KEY_SECRET", "dp3pAe4PpdmnAHCAPgEd3PyLmBQrkMde")
 
-# Kling AI API configuration (keys hardcoded as requested)
-ACCESS_KEY_ID = "AGBGmadNd9hakFYfahytyQQJtN8CJmDJ"
-ACCESS_KEY_SECRET = "dp3pAe4PpdmnAHCAPgEd3PyLmBQrkMde"
-API_URL = "https://api.klingai.com/v1/images/image2image"  # Correct endpoint
+API_BASE = "https://api.klingai.com"
+ENDPOINT_KOLORS = f"{API_BASE}/v1/images/kolors"           # face/subject reference modes (image-to-image)
+ENDPOINT_GENERATIONS = f"{API_BASE}/v1/images/generations"  # listing (used as a fallback poller)
+ENDPOINT_TASK = lambda tid: f"{API_BASE}/v1/tasks/{tid}"    # primary poller
 
-def generate_jwt_token():
-    """Generate JWT token for Kling AI API authentication."""
-    headers = {
-        "alg": "HS256",
-        "typ": "JWT"
-    }
+# ──────────────────────────────────────────────────────────────────────────────
+# AUTH — Kling uses JWT: iss / exp / nbf with HS256 (no "access_key" field)
+# ──────────────────────────────────────────────────────────────────────────────
+def make_jwt() -> str:
+    headers = {"alg": "HS256", "typ": "JWT"}
+    now = int(time.time())
     payload = {
         "iss": ACCESS_KEY_ID,
-        "exp": int(time.time()) + 1800,  # Token expires in 30 minutes
-        "nbf": int(time.time()) - 5  # Effective 5 seconds before current time
+        "exp": now + 1800,  # 30 minutes
+        "nbf": now - 5,     # start now (minus small skew)
     }
-    token = jwt.encode(payload, ACCESS_KEY_SECRET, headers=headers)
-    logger.debug(f"Generated JWT token: {token}")
-    return token
+    return jwt.encode(payload, ACCESS_KEY_SECRET, algorithm="HS256", headers=headers)
 
-def generate_image(image, prompt=""):
+# ──────────────────────────────────────────────────────────────────────────────
+# HELPERS
+# ──────────────────────────────────────────────────────────────────────────────
+def ensure_image_ok(img_path: str):
+    with Image.open(img_path) as im:
+        im.verify()  # quick integrity check
+
+def b64_data_uri(img_path: str) -> str:
+    mime = "image/png" if img_path.lower().endswith(".png") else "image/jpeg"
+    with open(img_path, "rb") as f:
+        b = base64.b64encode(f.read()).decode("utf-8")
+    return f"data:{mime};base64,{b}"
+
+def extract_task_id(resp_json: dict) -> str | None:
+    # Common shapes seen in the wild
+    if not resp_json:
+        return None
+    for key in ("task_id", "taskId", "id"):
+        if key in resp_json:
+            return str(resp_json[key])
+    data = resp_json.get("data") or {}
+    for key in ("task_id", "taskId", "id"):
+        if key in data:
+            return str(data[key])
+    # Sometimes nested deeper (e.g., {"task": {"id": ...}})
+    task = resp_json.get("task") or data.get("task") or {}
+    if "id" in task:
+        return str(task["id"])
+    return None
+
+def extract_image_urls(resp_json: dict) -> list[str]:
+    if not resp_json:
+        return []
+    data = resp_json.get("data") or {}
+    # Typical: data.task_result.images = [{url: "..."}]
+    task_result = data.get("task_result") or {}
+    images = task_result.get("images") or []
+    urls = [img.get("url") for img in images if isinstance(img, dict) and img.get("url")]
+    if urls:
+        return urls
+    # Some variants: output, image_url, result.image_url
+    for k in ("output", "image_url"):
+        if k in resp_json and isinstance(resp_json[k], str):
+            return [resp_json[k]]
+    result = resp_json.get("result") or {}
+    if isinstance(result, dict) and result.get("image_url"):
+        return [result["image_url"]]
+    # Works array pattern
+    works = resp_json.get("works") or data.get("works") or []
+    urls = []
+    for w in works:
+        if isinstance(w, dict):
+            u = w.get("url") or w.get("imageUrl")
+            if u:
+                urls.append(u)
+    return urls
+
+def download_to_file(url: str, out_path: Path) -> Path:
+    r = requests.get(url, timeout=60)
+    r.raise_for_status()
+    out_path.parent.mkdir(parents=True, exist_ok=True)
+    with open(out_path, "wb") as f:
+        f.write(r.content)
+    return out_path
+
+def poll_for_result(task_id: str, headers: dict, timeout_s: int = 300, interval_s: float = 3.0) -> dict:
+    """Poll task endpoint first; fallback to listing."""
+    deadline = time.time() + timeout_s
+    last_error = None
+
+    while time.time() < deadline:
+        try:
+            # Preferred: direct task status
+            r = requests.get(ENDPOINT_TASK(task_id), headers=headers, timeout=30)
+            if r.status_code == 200:
+                j = r.json()
+                # Either "status_name":"succeed" or "data.task_status":"succeed"
+                status_name = (j.get("status_name")
+                               or (j.get("data") or {}).get("task_status")
+                               or (j.get("task") or {}).get("status_name"))
+                if isinstance(status_name, dict):
+                    # Some SDKs wrap status as enum-like
+                    status_name = status_name.get("value")
+                if status_name in ("succeed", "succeeded", "success",  "SUCCEED"):
+                    return j
+                if status_name in ("failed", "FAIL", "failed_with_error"):
+                    return j
+            elif r.status_code in (401, 403, 404):
+                last_error = r.text
+            # Fallback: scan generations list
+            r2 = requests.get(ENDPOINT_GENERATIONS, headers=headers, params={"pageSize": 200}, timeout=30)
+            if r2.status_code == 200:
+                j2 = r2.json()
+                for item in (j2.get("data") or []):
+                    if str(item.get("task_id")) == str(task_id):
+                        status = item.get("task_status")
+                        if status in ("succeed", "succeeded", "success"):
+                            return item
+                        if status in ("failed",):
+                            return item
+        except requests.RequestException as e:
+            last_error = str(e)
+        time.sleep(interval_s)
+    raise TimeoutError(f"Polling timed out for task_id {task_id}. Last error: {last_error or 'n/a'}")
+
+# ──────────────────────────────────────────────────────────────────────────────
+# CORE CALL — Kolors face reference (single reference, faceStrength=97)
+# ──────────────────────────────────────────────────────────────────────────────
+def kling_face_reference(image_path: str, prompt: str, face_strength: int = 97, aspect_ratio: str = "1:1") -> tuple[str, str]:
     """
-    Call Kling AI API for single-reference face generation.
-    
-    Args:
-        image: Uploaded image file (from Gradio, face image)
-        prompt (str): Optional text prompt to guide transformation
-    
-    Returns:
-        tuple: (Path to generated image or None, error message or None)
+    Returns (display_image_path, download_file_path)
     """
-    if not image:
-        logger.error("No image uploaded")
-        return None, "Error: Please upload a valid face image (PNG/JPEG, <10 MB, ≥512x512 pixels)."
-    
-    # Convert image to base64
-    try:
-        with open(image, "rb") as img_file:
-            image_base64 = base64.b64encode(img_file.read()).decode("utf-8")
-        logger.debug("Image converted to base64 successfully")
-    except Exception as e:
-        logger.error(f"Failed to process image: {str(e)}")
-        return None, f"Error: Failed to process image. Ensure it’s a valid PNG/JPEG. Details: {str(e)}"
-    
-    headers = {
-        "Authorization": f"Bearer {generate_jwt_token()}",
-        "Content-Type": "application/json"
+    if not image_path:
+        raise gr.Error("Please upload a face/reference image.")
+    ensure_image_ok(image_path)
+
+    token = make_jwt()
+    headers_json = {
+        "Authorization": f"Bearer {token}",
+        "Content-Type": "application/json",
     }
-    
-    payload = {
-        "model_name": "kling-v2-1",  # V2.1 model for image-to-image
-        "image": image_base64,
-        "prompt": prompt or "Transform the face into a cartoon style while preserving identity",
-        "strength": 0.97,  # High reference strength for face preservation
-        "output_format": "png"
+    headers_multipart = {
+        "Authorization": f"Bearer {token}",
     }
-    
+
+    # First try: multipart/form-data (send file as `imageReference`)
+    data_multipart = {
+        "prompt": (None, prompt),
+        "reference": (None, "face"),
+        "faceStrength": (None, str(max(1, min(100, int(face_strength))))),
+        "faceNo": (None, "1"),           # single face reference
+        "imageCount": (None, "1"),
+        "aspect_ratio": (None, aspect_ratio),
+    }
+    files = {
+        "imageReference": (os.path.basename(image_path), open(image_path, "rb"),
+                           "image/png" if image_path.lower().endswith(".png") else "image/jpeg")
+    }
+
+    # Attempt 1 — multipart
     try:
-        logger.debug(f"Sending POST request to {API_URL} with payload: {payload}")
-        response = requests.post(API_URL, json=payload, headers=headers, timeout=30)
-        response.raise_for_status()
-        data = response.json()
-        logger.debug(f"API response: {data}")
-        
-        task_id = data.get("task_id") or data.get("taskId") or data.get("id")
-        if not task_id:
-            logger.error("No task ID returned in API response")
-            return None, "Error: No task ID returned. Check API endpoint or keys at https://app.klingai.com/global/dev."
-        
-        # Poll for task completion
-        status_url = f"https://api.klingai.com/v1/tasks/{task_id}"  # Updated status endpoint
-        for _ in range(60):  # Poll for up to 5 minutes
-            logger.debug(f"Polling status at {status_url}")
-            status_response = requests.get(status_url, headers=headers, timeout=30)
-            status_response.raise_for_status()
-            status_data = status_response.json()
-            logger.debug(f"Status response: {status_data}")
-            status = status_data.get("status")
-            if status == "succeeded":
-                image_url = status_data.get("image_url") or status_data.get("result", {}).get("image_url") or status_data.get("output") or status_data.get("data", {}).get("image")
-                if not image_url:
-                    logger.error("No image URL in API response")
-                    return None, "Error: No image URL in API response. Check API documentation at https://app.klingai.com/global/dev."
-                # Download the image
-                logger.debug(f"Downloading image from {image_url}")
-                image_response = requests.get(image_url, timeout=30)
-                image_response.raise_for_status()
-                output_path = Path("/tmp/output_image.png")  # Use /tmp for Spaces compatibility
-                with open(output_path, "wb") as f:
-                    f.write(image_response.content)
-                if not output_path.exists():
-                    logger.error("Output image file not created")
-                    return None, "Error: Failed to save output image."
-                logger.debug(f"Image saved to {output_path}")
-                return str(output_path), None
-            elif status == "failed":
-                logger.error("Image generation failed")
-                return None, "Error: Image generation failed. Ensure the image contains a clear face and avoid NSFW or sensitive content."
-            elif status == "processing" and status_data.get("progress", 0) >= 0.99:
-                logger.error("Generation stuck at 99%")
-                return None, "Error: Generation stuck at 99%. Check account credits or upgrade to a paid plan at https://app.klingai.com."
-            time.sleep(5)
-        
-        logger.error("Image generation timed out")
-        return None, "Error: Image generation timed out. Try during off-peak hours (e.g., 2 AM IST) or check account status."
-    
-    except requests.exceptions.HTTPError as e:
-        status_code = e.response.status_code if e.response else None
-        error_data = e.response.json() if e.response and e.response.text else {}
-        service_code = error_data.get("code", 0)
-        logger.error(f"HTTP error: {status_code}, Service Code: {service_code}, Details: {str(e)}")
-        
-        if status_code == 404:
-            if service_code == 1202:
-                return None, "Error: Invalid request method (1202). Check https://app.klingai.com/global/dev for correct endpoint."
-            elif service_code == 1203:
-                return None, "Error: Resource does not exist (1203). Verify model or endpoint."
-            return None, f"Error: Endpoint not found (404). Verify https://api.klingai.com/v1/images/image2image at https://app.klingai.com/global/dev."
-        elif status_code == 401:
-            if service_code == 1000:
-                return None, "Error: Authentication failed (1000). Verify API keys are correct."
-            elif service_code == 1001:
-                return None, "Error: Authorization empty (1001). Ensure JWT token is included."
-            elif service_code == 1002:
-                return None, "Error: Authorization invalid (1002). Check key format."
-            elif service_code == 1003:
-                return None, "Error: Authorization not yet valid (1003). Adjust token start time (nbf)."
-            elif service_code == 1004:
-                return None, "Error: Authorization expired (1004). Generate a new token."
-        elif status_code == 429:
-            if service_code == 1100:
-                return None, "Error: Account exception (1100). Verify account configuration at https://app.klingai.com."
-            elif service_code == 1101:
-                return None, "Error: Account in arrears (1101). Recharge your account."
-            elif service_code == 1102:
-                return None, "Error: Resource pack depleted or expired (1102). Purchase additional resources."
-            elif service_code in (1302, 1303):
-                return None, "Error: Rate limit exceeded (1302/1303). Reduce request frequency or contact support."
-            elif service_code == 1304:
-                return None, "Error: IP whitelisting issue (1304). Contact Kling AI support."
-        elif status_code == 403 and service_code == 1103:
-            return None, "Error: Unauthorized access (1103). Verify account permissions."
-        elif status_code == 400:
-            if service_code == 1200:
-                return None, "Error: Invalid request parameters (1200). Check payload format."
-            elif service_code == 1201:
-                return None, "Error: Invalid parameters (1201). Use PNG/JPEG (<10 MB, ≥512x512) with a clear face."
-            elif service_code == 1300:
-                return None, "Error: Platform policy triggered (1300). Check input content."
-            elif service_code == 1301:
-                return None, "Error: Content security policy triggered (1301). Ensure the image contains a clear face and avoid NSFW content."
-        elif status_code in (500, 503, 504):
-            if service_code == 5000:
-                return None, "Error: Server internal error (5000). Try again later."
-            elif service_code == 5001:
-                return None, "Error: Server unavailable due to maintenance (5001). Try again later."
-            elif service_code == 5002:
-                return None, "Error: Server timeout (5002). Try during off-peak hours."
-        return None, f"Error: API request failed. HTTP {status_code}, Service Code {service_code}. Details: {str(e)}"
-    except requests.exceptions.RequestException as e:
-        logger.error(f"Network error: {str(e)}")
-        return None, f"Error: Network issue. Ensure stable internet, disable VPN, and try again. Details: {str(e)}"
-
-def chatbot_interface(image, prompt):
-    """
-    Gradio interface for single-reference face generation.
-    
-    Args:
-        image: Uploaded image file (containing a face)
-        prompt (str): Optional text prompt
-    
-    Returns:
-        tuple: (Output image path for display, output image path for download, error message)
-    """
-    output_path, error = generate_image(image, prompt)
-    if error:
-        logger.debug(f"Error returned: {error}")
-        return None, None, error
-    logger.debug(f"Returning image path for display and download: {output_path}")
-    return output_path, output_path, None
-
-# Define Gradio interface
-with gr.Blocks() as iface:
-    gr.Markdown(
-        """
-        # Kling AI Single-Reference Face Generator
-        Upload a PNG/JPEG image (<10 MB, ≥512x512 pixels) with a clear face to generate a transformed image using Kling AI V2.1 API (reference strength 0.97). Avoid NSFW or sensitive content.
-        If errors persist, check the correct API endpoint at https://app.klingai.com/global/dev.
-        """
-    )
+        resp = requests.post(ENDPOINT_KOLORS, headers=headers_multipart, files=files, data=data_multipart, timeout=60)
+        if resp.status_code == 200:
+            j = resp.json()
+        else:
+            # Read JSON anyway if possible
+            try:
+                j = resp.json()
+            except Exception:
+                j = {"code": resp.status_code, "message": resp.text}
+    finally:
+        # Close file handle if opened
+        try:
+            files["imageReference"][1].close()
+        except Exception:
+            pass
+
+    task_id = extract_task_id(j)
+
+    # If Kolors rejected multipart or no task_id, try JSON with data URI
+    if not task_id:
+        payload = {
+            "prompt": prompt,
+            "reference": "face",
+            "faceStrength": max(1, min(100, int(face_strength))),
+            "faceNo": 1,
+            "imageCount": 1,
+            "aspect_ratio": aspect_ratio,
+            "imageReference": b64_data_uri(image_path),
+        }
+        resp2 = requests.post(ENDPOINT_KOLORS, headers=headers_json, json=payload, timeout=60)
+        try:
+            j = resp2.json()
+        except Exception:
+            j = {"code": resp2.status_code, "message": resp2.text}
+        task_id = extract_task_id(j)
+
+    if not task_id:
+        code = j.get("code") or j.get("service_code") or "?"
+        msg = j.get("message") or j.get("error") or f"HTTP {resp.status_code if 'resp' in locals() else '?'}"
+        raise gr.Error(f"Create task failed. Code: {code}. Message: {msg}")
+
+    # Poll
+    result_json = poll_for_result(task_id, headers=headers_json, timeout_s=420, interval_s=3.0)
+
+    # Gather image URLs
+    urls = extract_image_urls(result_json)
+    if not urls:
+        # Some APIs return the latest object on /v1/images/generations with same task_id
+        try:
+            listing = requests.get(ENDPOINT_GENERATIONS, headers=headers_json, params={"pageSize": 200}, timeout=30).json()
+            for item in (listing.get("data") or []):
+                if str(item.get("task_id")) == str(task_id):
+                    urls = extract_image_urls(item)
+                    if urls:
+                        break
+        except Exception:
+            pass
+
+    if not urls:
+        raise gr.Error(f"Task {task_id} succeeded but no image URL found in response.")
+
+    # Download first image
+    out_dir = Path("outputs")
+    out_dir.mkdir(parents=True, exist_ok=True)
+    out_path = out_dir / f"kling_face_{task_id}.png"
+    download_to_file(urls[0], out_path)
+
+    # Return same path for preview and download
+    return str(out_path), str(out_path)
+
+# ──────────────────────────────────────────────────────────────────────────────
+# GRADIO UI
+# ──────────────────────────────────────────────────────────────────────────────
+with gr.Blocks(title="Kling AI — Image to Image (Face Reference)") as demo:
+    gr.Markdown("### Kling AI — Image-to-Image (Single Face Reference)\nUpload a face image and a prompt. Strength defaults to 97.")
     with gr.Row():
-        with gr.Column():
-            image_input = gr.Image(type="filepath", label="Upload Face Image (PNG/JPEG, <10 MB, ≥512x512)")
-            prompt_input = gr.Textbox(lines=2, placeholder="Enter an optional prompt (e.g., 'Turn this face into a cartoon')", label="Prompt")
-            generate_button = gr.Button("Generate")
-        with gr.Column():
-            output_image = gr.Image(label="Generated Face Image")
-            output_file = gr.File(label="Download Generated Image")
-            error_message = gr.Textbox(label="Status/Error Message", interactive=False)
-    
-    generate_button.click(
-        fn=chatbot_interface,
-        inputs=[image_input, prompt_input],
-        outputs=[output_image, output_file, error_message]
-    )
-
-# Launch the interface
+        in_image = gr.Image(type="filepath", label="Reference Face Image (PNG/JPG)")
+        in_prompt = gr.Textbox(label="Prompt", placeholder="e.g., Ultra-detailed portrait, soft light, studio background", lines=2)
+    with gr.Row():
+        in_strength = gr.Slider(1, 100, value=97, step=1, label="Face Reference Strength")
+        in_aspect = gr.Dropdown(choices=["1:1", "3:4", "4:3", "2:3", "3:2", "16:9", "9:16", "21:9"], value="1:1", label="Aspect Ratio")
+    btn = gr.Button("Generate", variant="primary")
+
+    out_img = gr.Image(label="Generated Image", show_download_button=False)
+    out_file = gr.File(label="Download Image")
+
+    def run(image, prompt, strength, aspect):
+        if not prompt or not prompt.strip():
+            raise gr.Error("Please enter a prompt.")
+        return kling_face_reference(image, prompt.strip(), int(strength), aspect)
+
+    btn.click(fn=run, inputs=[in_image, in_prompt, in_strength, in_aspect], outputs=[out_img, out_file])
+
 if __name__ == "__main__":
-    logger.debug("Launching Gradio interface")
-    iface.launch(server_name="0.0.0.0", server_port=7860)
+    # On HF Spaces, just `python app.py` is enough — no need to set host/port.
+    demo.launch()