app.py

# app.py — Story Generator with Elegant UI
import os, json, re, pathlib, base64, time, uuid
from huggingface_hub import InferenceClient
from PIL import Image
import gradio as gr

# ---------- Config ----------
HF_TOKEN = os.environ.get("HF_TOKEN")
if not HF_TOKEN:
    raise RuntimeError("⚠️ Set HF_TOKEN environment variable (use Spaces Secrets).")

CHAT_MODEL = "meta-llama/Llama-3.1-8b-instruct"
IMAGE_MODEL = "black-forest-labs/FLUX.1-schnell"
SAFETY_MODEL = "meta-llama/Meta-Llama-Guard-2-8B" # Safety model for content moderation
OUT_DIR = pathlib.Path("/tmp/generated")
OUT_DIR.mkdir(parents=True, exist_ok=True)
client = InferenceClient(api_key=HF_TOKEN, provider="auto")


# ---------- Utility ----------
def try_chat_completion(model_id, messages, max_tokens=3000):
    r = client.chat.completions.create(model=model_id, messages=messages, max_tokens=max_tokens)
    return r.choices[0].message.get("content", "")

def extract_json_from_text(text):
    if not text:
        return None
    match = re.search(r'```json\s*(\{[\s\S]*\})\s*```', text)
    if match:
        json_str = match.group(1)
    else:
        match = re.search(r'\{[\s\S]*\}', text)
        if not match:
            return None
        json_str = match.group(0)
    try:
        return json.loads(json_str)
    except json.JSONDecodeError as e:
        print(f"--- JSON PARSING FAILED ---\nError: {e}\nContent: {json_str}\n--------------------------")
        return None

def hf_text_to_image(model_id, prompt, out_path):
    img = client.text_to_image(prompt=prompt, model=model_id)
    if isinstance(img, Image.Image):
        img.save(out_path)
        return out_path
    raise RuntimeError("Invalid image response")

def is_content_inappropriate(text_to_check):
    """Uses Llama Guard to check for inappropriate content."""
    try:
        r = client.chat.completions.create(
            model=SAFETY_MODEL,
            messages=[{"role": "user", "content": f"Please evaluate if the following content is safe or unsafe based on typical safety guidelines for an AI assistant. Output 'safe' or 'unsafe'.\n\nContent: \"{text_to_check}\""}],
            max_tokens=20,
            temperature=0.1
        )
        response = r.choices[0].message.get("content", "").lower()
        return "unsafe" in response
    except Exception as e:
        print(f"Error in safety check: {e}")
        return False

# ---------- Story logic ----------
def make_prompt(user_prompt, nscenes=6, nsent=5):
    # This prompt is excellent and requires no changes.
    return f"""
You are a creative story writer. Your task is to write a compelling story based on a user's prompt.
You MUST return the story in a single, valid JSON object. Do not write any text or explanations outside of the JSON structure.

Here is an example of the required JSON format:
{{
  "title": "A descriptive title for the entire story",
  "scenes": [
    {{
      "id": 1,
      "text": "The full story text for this scene. This should be a complete paragraph with around {nsent} sentences, describing the events and setting.",
      "visual_prompt": "A detailed, vivid description for an image generation model, capturing the key visual elements of this scene."
    }}
  ]
}}

Please use the following details for the story:
- Story Prompt: "{user_prompt}"
- Total number of scenes: {nscenes}

Now, generate the story in the specified JSON format.
"""

def generate_story_and_images(prompt, nscenes, nsent, img_model):
    start = time.time()
    logs = []

    logs.append("🎬 Generating story...")
    raw = try_chat_completion(CHAT_MODEL, [{"role": "user", "content": make_prompt(prompt, nscenes, nsent)}])
    story = extract_json_from_text(raw)
    
    if not story:
        story = {"title": "Untitled", "scenes": [{"id": i + 1, "text": f"Scene {i+1}. The AI failed to generate a proper story, or the JSON was malformed.", "visual_prompt": prompt} for i in range(nscenes)]}
        logs.append("⚠️ Failed to parse story JSON, using fallback.")
    else:
        logs.append("✅ Story JSON parsed.")

    image_paths = []
    for s in story["scenes"]:
        visual_prompt = s.get("visual_prompt", s.get("text", prompt))
        if is_content_inappropriate(visual_prompt):
            gr.Warning(f"Visual prompt for Scene {s['id']} was moderated for safety. Generating a default image.")
            visual_prompt = "A serene landscape with gentle colors."
        
        name = OUT_DIR / f"{uuid.uuid4().hex[:6]}_scene_{s['id']}.png"
        logs.append(f"🎨 Generating image for Scene {s['id']}...")
        hf_text_to_image(img_model, visual_prompt, str(name))
        image_paths.append(str(name))
    total = time.time() - start
    logs.append(f"✨ Done in {total:.2f}s")
    return story, image_paths, "\n".join(logs)

# ---------- UI ----------
def build_ui():
    css = """
    .main-container { max-width: 1400px; margin: 0 auto; }
    .prompt-section { background: linear-gradient(135deg, #667eea 0%, #764ba2 100%); padding: 32px; border-radius: 16px; margin-bottom: 24px; }
    .prompt-box textarea { font-size: 16px !important; }
    .story-panel { background: rgba(255,255,255,0.05); padding: 24px; border-radius: 12px; backdrop-filter: blur(10px); border: 1px solid rgba(255,255,255,0.1); max-height: 800px; overflow-y: auto; }
    .story-title { font-size: 32px; font-weight: 700; margin-bottom: 24px; background: linear-gradient(135deg, #667eea 0%, #764ba2 100%); -webkit-background-clip: text; -webkit-text-fill-color: transparent; }
    .story-content { line-height: 1.8; font-size: 16px; color: rgba(255,255,255,0.9); }
    .story-content p { margin-bottom: 16px; }
    """

    with gr.Blocks(css=css, theme=gr.themes.Soft(), title="Story Generator") as demo:
        gr.Markdown("# 📚 AI Story Generator", elem_classes="main-title")
        
        with gr.Column(elem_classes="main-container"):
            with gr.Column(elem_classes="prompt-section"):
                prompt_box = gr.Textbox(label="✨ Enter your story idea", placeholder="e.g. A pirate discovering a hidden island...", lines=3, elem_classes="prompt-box")
                with gr.Row():
                    generate_btn = gr.Button("🚀 Generate Story", variant="primary", size="lg", scale=3)
                    with gr.Column(scale=1):
                        with gr.Accordion("⚙️ Settings", open=False):
                            nscenes = gr.Slider(2, 12, value=6, step=1, label="📖 Number of scenes")
                            nsent = gr.Slider(2, 8, value=5, step=1, label="📝 Sentences per scene")
                            img_model = gr.Dropdown(choices=[IMAGE_MODEL], value=IMAGE_MODEL, label="🎨 Image model")
                            log_box = gr.Textbox(label="📋 Generation Logs", lines=6, interactive=False)

            with gr.Row():
                with gr.Column(scale=5, elem_classes="story-panel"):
                    story_html = gr.HTML("<div style='text-align:center;padding:40px;color:#888;'>Your story will appear here...<br><br>Click 'Generate Story' to begin! ✨</div>")
                
                with gr.Column(scale=7):
                    image_gallery = gr.Gallery(
                        label="📷 Scene Visuals", show_label=False, elem_id="gallery",
                        columns=2, object_fit="cover", height="auto", preview=True
                    )

        def on_generate(prompt, nscenes, nsent, img_model):
            if is_content_inappropriate(prompt):
                gr.Warning("Your prompt seems to violate the safety policy. Please try again.")
                return "<div style='text-align:center;padding:40px;color:#888;'>Prompt rejected due to safety policy.</div>", [], "Prompt rejected by safety filter."
            
            story, imgs, logs = generate_story_and_images(prompt, int(nscenes), int(nsent), img_model)
            
            story_output_html = f"<div class='story-title'>{story.get('title', 'Untitled')}</div>\n<div class='story-content'>\n"
            
            for s in story.get('scenes', []):
                scene_text = s.get('text', '')
                if is_content_inappropriate(scene_text):
                    scene_text = f"**[Scene {s['id']} was moderated for safety and replaced with a placeholder.]**"
                    gr.Info(f"Scene {s['id']} content was flagged and replaced.")
                story_output_html += f"<p>{scene_text}</p>\n\n"
            
            story_output_html += "</div>"
            
            return story_output_html, imgs, logs

        generate_btn.click(
            on_generate, 
            inputs=[prompt_box, nscenes, nsent, img_model], 
            outputs=[story_html, image_gallery, log_box]
        )

    return demo

app = build_ui()

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