from transformers import pipeline

# ✅ Use a lightweight instruction-tuned model (fast for Spaces CPU)
generator = pipeline("text2text-generation", model="google/flan-t5-small")

def generate_text_content(topic, content_format="Blog Post", tone="Casual", audience="General"):
    """
    Generate content (title, outline, and body) based on user input.
    """
    try:
        # Construct structured prompt
        prompt = (
            f"You are an expert content creator and copywriter.\n"
            f"Create a {tone.lower()} {content_format.lower()} about {topic} "
            f"for {audience if audience else 'a general audience'}.\n"
            "Include:\n"
            "1. Catchy title\n"
            "2. Detailed outline\n"
            "3. Full content\n"
            "4. Call-to-action"
        )

        # Generate response (fast + bounded length)
        response = generator(
            prompt,
            max_length=200,     # keep small for speed
            do_sample=True,
            temperature=0.7,
            top_p=0.9,
            num_return_sequences=1,
        )

        return response[0]["generated_text"]

    except Exception as e:
        return f"⚠️ Error generating content: {str(e)}"