fix: switch to Docker SDK with direct FastAPI, remove Gradio wrapper

Dockerfile

@@ -59,7 +59,7 @@ EXPOSE 7860
 
 # Health check for container monitoring
 HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
-    CMD python -c "import requests; requests.get('http://localhost:7860/health', timeout=5)"
+    CMD python -c "import requests; requests.get('http://localhost:7860/', timeout=5)"
 
-# Run Gradio wrapper (which includes FastAPI)
-CMD ["python", "app.py"]
+# Run FastAPI directly on port 7860 (HF Spaces standard)
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -3,9 +3,8 @@ title: VIBE_LINK Server
 emoji: 🎨
 colorFrom: gray
 colorTo: green
-sdk: gradio
-sdk_version: 6.3.0
-app_file: app.py
+sdk: docker
+app_port: 7860
 pinned: false
 ---
 

app.py

@@ -1,13 +0,0 @@
-"""
-Gradio wrapper for HF Spaces compatibility
-Mounts FastAPI app to Gradio for proper routing
-"""
-import gradio as gr
-from app.main import app
-
-# Mount FastAPI to Gradio
-gradio_app = gr.mount_gradio_app(
-    gr.Blocks(title="VIBE_LINK API"),
-    app,
-    path="/"
-)

main.py

@@ -1,485 +0,0 @@
-"""
-VIBE_LINK Backend API
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-Serverless-style FastAPI backend for converting websites into Vibe Posters.
-Optimized for Hugging Face Spaces (Docker SDK) with 16GB RAM.
-
-Architecture:
-    1. Screenshot Capture → pyppeteer (Headless Chrome)
-    2. AI Analysis → Google Gemini 2.5 Flash
-    3. Image Generation → Hugging Face Flux.1-dev
-    4. Image Hosting → ImgBB API
-
-Author: S-Grade Developer | Production-Ready Code
-━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-"""
-
-import os
-import json
-import base64
-import asyncio
-import tempfile
-from pathlib import Path
-from typing import Optional, Dict
-from contextlib import asynccontextmanager
-
-import requests
-from fastapi import FastAPI, HTTPException, Request
-from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import JSONResponse
-from pydantic import BaseModel, HttpUrl, Field
-from pyppeteer import launch
-from huggingface_hub import InferenceClient
-import google.generativeai as genai
-from PIL import Image
-from dotenv import load_dotenv
-
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-# CONFIGURATION & ENVIRONMENT
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-load_dotenv()
-
-# API Keys (Secured via Environment Variables)
-HF_TOKEN = os.getenv("HF_TOKEN")
-GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
-IMGBB_KEY = os.getenv("IMGBB_KEY")
-
-# Validate API Keys
-if not all([HF_TOKEN, GEMINI_API_KEY, IMGBB_KEY]):
-    raise RuntimeError(
-        "Missing API Keys! Set HF_TOKEN, GEMINI_API_KEY, IMGBB_KEY in environment."
-    )
-
-# Initialize AI Services
-genai.configure(api_key=GEMINI_API_KEY)
-hf_client = InferenceClient(token=HF_TOKEN)
-
-# Prevent pyppeteer from downloading Chromium (use system-installed one)
-os.environ['PUPPETEER_SKIP_CHROMIUM_DOWNLOAD'] = 'true'
-os.environ['PUPPETEER_EXECUTABLE_PATH'] = '/usr/bin/chromium'
-
-# Constants
-SCREENSHOT_TIMEOUT = 20000  # 20 seconds
-VIEWPORT_WIDTH = 1280
-VIEWPORT_HEIGHT = 1200
-FLUX_MODEL = "black-forest-labs/FLUX.1-schnell"  # Faster model, same quality (4 steps vs 30)
-TEMP_DIR = Path(tempfile.gettempdir()) / "vibe_link"
-TEMP_DIR.mkdir(exist_ok=True)
-
-
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-# PYDANTIC MODELS
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-class CreateRequest(BaseModel):
-    url: HttpUrl = Field(..., description="Target website URL to analyze")
-
-
-class CreateResponse(BaseModel):
-    status: str
-    poster_url: str
-    vibe: Optional[str] = None
-    summary: Optional[str] = None
-
-
-class ErrorResponse(BaseModel):
-    status: str = "error"
-    message: str
-    detail: Optional[str] = None
-
-
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-# FASTAPI LIFECYCLE MANAGEMENT
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-@asynccontextmanager
-async def lifespan(app: FastAPI):
-    """Manage browser lifecycle to minimize resource usage"""
-    print("🚀 VIBE_LINK Backend Starting...")
-    yield
-    print("🛑 Cleaning up resources...")
-    # Cleanup temp files on shutdown
-    for file in TEMP_DIR.glob("*"):
-        try:
-            file.unlink()
-        except Exception:
-            pass
-
-
-app = FastAPI(
-    title="VIBE_LINK API",
-    description="AI-powered Website to Vibe Poster Generator",
-    version="1.0.0",
-    lifespan=lifespan,
-)
-
-# CORS Middleware (Allow all origins for public API)
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-# CORE PIPELINE FUNCTIONS
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-async def capture_screenshot(url: str) -> Path:
-    """
-    STEP 1: Screenshot Capture
-    Uses Headless Chrome to capture website screenshot.
-    Optimized for Docker environment with sandbox disabled.
-    """
-    screenshot_path = TEMP_DIR / f"screenshot_{id(url)}.jpg"
-    
-    browser = None
-    try:
-        # Launch Chromium with Docker-safe arguments
-        browser = await launch(
-            headless=True,
-            executablePath="/usr/bin/chromium",
-            args=[
-                "--no-sandbox",
-                "--disable-setuid-sandbox",
-                "--disable-dev-shm-usage",
-                "--disable-gpu",
-                "--disable-software-rasterizer",
-                "--disable-extensions",
-                "--single-process",  # Memory optimization
-            ],
-        )
-        
-        page = await browser.newPage()
-        await page.setViewport({"width": VIEWPORT_WIDTH, "height": VIEWPORT_HEIGHT})
-        
-        # Navigate with timeout
-        await page.goto(
-            url,
-            {"waitUntil": "domcontentloaded", "timeout": SCREENSHOT_TIMEOUT}
-        )
-        
-        # Capture screenshot
-        await page.screenshot({"path": str(screenshot_path), "type": "jpeg", "quality": 85})
-        
-        return screenshot_path
-        
-    except Exception as e:
-        raise HTTPException(
-            status_code=500,
-            detail=f"Screenshot capture failed: {str(e)}"
-        )
-    finally:
-        if browser:
-            await browser.close()
-
-
-async def analyze_with_gemini(screenshot_path: Path) -> Dict[str, str]:
-    """
-    STEP 2: AI Analysis
-    Uses Google Gemini 2.5 Flash to analyze screenshot and generate
-    vibe description + image generation prompt.
-    """
-    try:
-        # Upload image to Gemini
-        uploaded_file = genai.upload_file(str(screenshot_path))
-        
-        # AI Analysis Prompt
-        analysis_prompt = """
-        You are a world-class Creative Director and Prompt Engineer.
-        Analyze this website screenshot and generate a high-end vertical brand poster concept for Flux.1.
-
-        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-        STEP 1: EXTRACT BRAND IDENTITY
-        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-        1. Find the main brand name/logo in the screenshot.
-        2. **CRITICAL: Convert Korean to English romanization.**
-           Examples:
-           - "무신사" → "MUSINSA"
-           - "떡볶이 천국" → "TTEOKBOKKI HEAVEN"
-           - "이창민" → "LEE CHANGMIN"
-           - "카페 서울" → "CAFE SEOUL"
-        3. If no brand name exists, use the domain name.
-
-        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-        STEP 2: ANALYZE BUSINESS TYPE & VISUAL DNA
-        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-        Identify the business category and extract visual elements:
-        
-        **Business Examples:**
-        - Fashion (무신사) → Clothes, fabrics, hangers, runway
-        - Food (떡볶이) → Ingredients, steam, spices, bowls
-        - Tech (코딩) → Code snippets, circuits, data streams
-        - Fitness (헬스장) → Dumbbells, protein, energy
-        - Corporate (법률) → Books, marble, gold, scales
-        
-        **Extract:**
-        - Dominant Color Palette (2-3 colors)
-        - Key Physical Objects (what fills the background)
-        - Material/Texture (fabric, metal, food, digital)
-        - Lighting Mood (neon, sunlight, studio, dark)
-
-        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-        STEP 3: CONSTRUCT FLUX.1 PROMPT
-        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-        Build a detailed scene description:
-
-        **BACKGROUND (70% of prompt):**
-        - Describe a 3D environment FILLED with objects related to the business.
-        - Example (Fashion): "floating designer clothes, silk fabrics, leather jackets, sneakers, accessories scattered in mid-air"
-        - Example (Food): "steaming bowls, fresh ingredients, spices exploding, sauce splash, vibrant colors"
-        - Use the dominant colors extracted in Step 2.
-
-        **CENTER TYPOGRAPHY (30% of prompt):**
-        - Place the ENGLISH brand name in the CENTER.
-        - Make it a 3D object matching the theme:
-          * Fashion: "metallic chrome letters with fabric texture"
-          * Food: "letters made of the food itself (cookies, cake, rice)"
-          * Tech: "neon holographic glowing text"
-        - Keywords: "big bold 3D typography", "magazine cover layout", "centered composition"
-
-        **MANDATORY RULES:**
-        ✅ NO Korean characters (Hangul) - Flux CANNOT render them
-        ✅ Convert all Korean to English romanization
-        ✅ Brand name must be in CENTER in big 3D letters
-        ✅ Background must be FILLED with thematic 3D objects
-        ✅ Use "vertical poster", "9:16 aspect ratio", "high quality 3D render"
-
-        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-        OUTPUT FORMAT (JSON)
-        ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-        Return ONLY valid JSON:
-        {
-            "vibe": "Business type in 1-2 English words (e.g., Fashion, Spicy Food, Tech)",
-            "summary": "One sentence description in Korean",
-            "flux_prompt": "Complete Flux.1 prompt in English following the rules above"
-        }
-
-        **Example flux_prompt structure:**
-        "A vertical 3D poster (9:16), [BACKGROUND: detailed scene with thematic objects], centered big bold 3D typography text '[BRAND NAME IN ENGLISH]' made of [material matching theme], [lighting description], magazine cover layout, high quality render, cinematic composition"
-        """
-        
-        # Call Gemini API
-        model = genai.GenerativeModel("gemini-2.0-flash-exp")
-        response = model.generate_content([uploaded_file, analysis_prompt])
-        
-        # Robust JSON parsing with multiple fallback strategies
-        response_text = response.text.strip()
-        
-        # Strategy 1: Remove markdown code blocks
-        if response_text.startswith("```json"):
-            response_text = response_text[7:-3].strip()
-        elif response_text.startswith("```"):
-            response_text = response_text[3:-3].strip()
-        
-        # Strategy 2: Find JSON object between curly braces
-        if not response_text.startswith("{"):
-            start = response_text.find("{")
-            end = response_text.rfind("}") + 1
-            if start != -1 and end > start:
-                response_text = response_text[start:end]
-        
-        try:
-            analysis = json.loads(response_text)
-        except json.JSONDecodeError as e:
-            raise ValueError(f"Failed to parse Gemini JSON response: {e}. Response: {response_text[:200]}")
-        
-        # Validate required fields
-        if not all(k in analysis for k in ["vibe", "summary", "flux_prompt"]):
-            raise ValueError(f"Missing required fields in Gemini response. Got: {list(analysis.keys())}")
-            
-        return analysis
-        
-    except Exception as e:
-        raise HTTPException(
-            status_code=500,
-            detail=f"AI analysis failed: {str(e)}"
-        )
-
-
-async def generate_poster(flux_prompt: str) -> Path:
-    """
-    STEP 3: Image Generation
-    Uses Hugging Face Inference API with Flux.1-dev model
-    to generate vibe poster.
-    """
-    poster_path = TEMP_DIR / f"poster_{id(flux_prompt)}.webp"
-    
-    try:
-        # Call Flux.1-schnell via HF Inference API
-        # Schnell is 10x faster than dev with similar quality
-        image = hf_client.text_to_image(
-            flux_prompt,
-            model=FLUX_MODEL,
-            # Parameters optimized for vertical poster
-            width=768,
-            height=1344,  # 9:16 aspect ratio
-            num_inference_steps=4,  # Schnell optimized for 1-4 steps
-        )
-        
-        # Save as WebP for optimal compression
-        image.save(str(poster_path), "WEBP", quality=90, method=6)
-        
-        return poster_path
-        
-    except Exception as e:
-        raise HTTPException(
-            status_code=500,
-            detail=f"Image generation failed: {str(e)}"
-        )
-
-
-async def upload_to_imgbb(image_path: Path) -> str:
-    """
-    STEP 4: Image Hosting
-    Uploads generated poster to ImgBB for permanent hosting.
-    Returns public URL.
-    """
-    try:
-        # Read image and encode as base64
-        with open(image_path, "rb") as f:
-            image_data = base64.b64encode(f.read()).decode("utf-8")
-        
-        # Upload to ImgBB
-        response = requests.post(
-            "https://api.imgbb.com/1/upload",
-            data={
-                "key": IMGBB_KEY,
-                "image": image_data,
-                "expiration": 0,  # Never expire
-            },
-            timeout=30,
-        )
-        
-        response.raise_for_status()
-        result = response.json()
-        
-        if not result.get("success"):
-            raise ValueError("ImgBB upload failed")
-            
-        return result["data"]["url"]
-        
-    except Exception as e:
-        raise HTTPException(
-            status_code=500,
-            detail=f"Image upload failed: {str(e)}"
-        )
-    finally:
-        # Clean up local file immediately after upload
-        try:
-            image_path.unlink()
-        except Exception:
-            pass
-
-
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-# API ENDPOINTS
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-@app.get("/")
-async def root():
-    """API Root - Health & Info"""
-    return {
-        "service": "VIBE_LINK API",
-        "status": "operational",
-        "version": "1.0.0",
-        "endpoints": {
-            "create": "POST /create - Generate vibe poster from URL",
-            "health": "GET /health - Service health check",
-        },
-    }
-
-
-@app.get("/health")
-async def health_check():
-    """Health Check Endpoint (for Docker HEALTHCHECK)"""
-    return {"status": "healthy", "service": "vibe-link-backend"}
-
-
-@app.post("/create", response_model=CreateResponse, responses={500: {"model": ErrorResponse}})
-async def create_vibe_poster(request: CreateRequest):
-    """
-    🎨 Main Endpoint: Generate Vibe Poster
-    
-    Pipeline:
-        1. Capture screenshot of website
-        2. Analyze with Gemini AI
-        3. Generate poster with Flux.1
-        4. Upload to ImgBB
-        
-    Returns:
-        CreateResponse with poster URL and metadata
-    """
-    screenshot_path = None
-    
-    try:
-        # STEP 1: Screenshot
-        print(f"📸 Capturing screenshot: {request.url}")
-        screenshot_path = await capture_screenshot(str(request.url))
-        
-        # STEP 2: AI Analysis
-        print("🧠 Analyzing with Gemini...")
-        analysis = await analyze_with_gemini(screenshot_path)
-        
-        # STEP 3: Generate Poster
-        print("🎨 Generating poster with Flux.1...")
-        poster_path = await generate_poster(analysis["flux_prompt"])
-        
-        # STEP 4: Upload to ImgBB
-        print("☁️ Uploading to ImgBB...")
-        poster_url = await upload_to_imgbb(poster_path)
-        
-        print(f"✅ Success! Poster URL: {poster_url}")
-        
-        return CreateResponse(
-            status="success",
-            poster_url=poster_url,
-            vibe=analysis["vibe"],
-            summary=analysis["summary"],
-        )
-        
-    except HTTPException:
-        raise
-    except Exception as e:
-        raise HTTPException(
-            status_code=500,
-            detail=f"Unexpected error: {str(e)}"
-        )
-    finally:
-        # Cleanup screenshot file
-        if screenshot_path and screenshot_path.exists():
-            try:
-                screenshot_path.unlink()
-            except Exception:
-                pass
-
-
-@app.exception_handler(Exception)
-async def global_exception_handler(request: Request, exc: Exception):
-    """Global error handler for unhandled exceptions"""
-    return JSONResponse(
-        status_code=500,
-        content={
-            "status": "error",
-            "message": "Internal server error",
-            "detail": str(exc),
-        },
-    )
-
-
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-# MAIN ENTRY POINT
-# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-
-if __name__ == "__main__":
-    import uvicorn
-    
-    uvicorn.run(
-        app,
-        host="0.0.0.0",
-        port=7860,
-        log_level="info",
-    )

pyproject.toml

@@ -1,22 +0,0 @@
-[project]
-name = "vibe-link-backend"
-version = "1.0.0"
-description = "AI-powered Website to Vibe Poster Generator"
-requires-python = ">=3.9"
-dependencies = [
-    "fastapi==0.109.0",
-    "uvicorn[standard]==0.27.0",
-    "python-multipart==0.0.6",
-    "pyppeteer==1.0.2",
-    "google-generativeai==0.3.2",
-    "huggingface-hub==0.20.3",
-    "Pillow==10.2.0",
-    "requests==2.31.0",
-    "httpx==0.26.0",
-    "python-dotenv==1.0.1",
-    "aiofiles==23.2.1",
-]
-
-[build-system]
-requires = ["hatchling"]
-build-backend = "hatchling.build"

requirements.txt

@@ -6,9 +6,6 @@ fastapi==0.115.0
 uvicorn[standard]==0.32.0
 python-multipart==0.0.12
 
-# Gradio (HF Spaces requirement)
-gradio==6.3.0
-
 # Headless Browser (Screenshot)
 pyppeteer==1.0.2