from diffusers import StableDiffusionXLPipeline
import torch
from langchain.chains import LLMChain
from langchain.llms import HuggingFaceHub
from langchain.prompts import PromptTemplate
import streamlit as st
import json

# Load existing ideas from a file
def load_ideas():
    try:
        with open("ideas.json", "r") as file:
            ideas = json.load(file)
    except FileNotFoundError:
        ideas = []
    return ideas

# Save ideas to a file
def save_ideas(ideas):
    with open("ideas.json", "w") as file:
        json.dump(ideas, file)

# Function to generate content
def generate_content(topic):
    hub_llm = HuggingFaceHub(repo_id="HuggingFaceH4/zephyr-7b-beta")
    prompt = PromptTemplate(
        input_variables=['keyword'],
        template="""
        Write a comprehensive article about {keyword} covering the following aspects:
        Introduction, History and Background, Key Concepts and Terminology, Use Cases and Applications, Benefits and Drawbacks, Future Outlook, Conclusion
        Ensure that the article is well-structured, informative, and at least 1500 words long. Use SEO best practices for content optimization.
        """
    )
    hub_chain = LLMChain(prompt=prompt, llm=hub_llm, verbose=True)
    content = hub_chain.run(topic)

    subheadings = [
        "Introduction",
        "History and Background",
        "Key Concepts and Terminology",
        "Use Cases and Applications",
        "Benefits and Drawbacks",
        "Future Outlook",
        "Conclusion",
    ]

    for subheading in subheadings:
        if (subheading + ":") in content:
            content = content.replace(subheading + ":", "## " + subheading + "\n")
        elif subheading in content:
            content = content.replace(subheading, "## " + subheading + "\n")

    return content

def make_pipe():
  pipe = StableDiffusionXLPipeline.from_pretrained("segmind/SSD-1B", torch_dtype=torch.float16, use_safetensors=True, variant="fp16")
  pipe.to("cuda")
  return pipe

# generate image
def generate_image(pipe,topic):
  prompt = f"A banner for a blog about{topic}" # Your prompt here
  neg_prompt = "ugly, blurry, poor quality" # Negative prompt here
  image = pipe(prompt=prompt, negative_prompt=neg_prompt).images[0]
  return image



pipe = make_pipe()
with st.spinner('Please Wait for the models to train...'):
  if pipe:
    # Streamlit app
    st.title("Blog Generator")

    # Input and button
    topic = st.text_input("Enter Title for the blog")
    button_clicked = st.button("Create blog!")
    st.subheader(topic)
    # Load existing ideas
    existing_ideas = load_ideas()
    st.sidebar.header("Previous Ideas:")

    # Display existing ideas in the sidebar
    keys = list(set([key for idea in existing_ideas for key in idea.keys()]))
    if topic in keys:
      index = keys.index(topic)
      selected_idea = st.sidebar.selectbox("Select Idea", keys, key="selectbox", index=index)
      # Display content for the selected idea
      selected_idea_from_list = next((idea for idea in existing_ideas if selected_idea in idea), None)
      st.markdown(selected_idea_from_list[selected_idea])
    else:
      index = 0
  st.success('')


# Handle button click
if button_clicked:
    # Generate content and update existing ideas
    image = generate_image(pipe,topic)
    st.image(image)
    content = generate_content(topic)
    existing_ideas.append({topic: content})
    save_ideas(existing_ideas)
    st.experimental_rerun()
    # Update keys and selected idea in the sidebar
    keys = list(set([key for idea in existing_ideas for key in idea.keys()]))
    selected_idea = st.sidebar.selectbox("Select Idea", keys, key="selectbox", index=keys.index(topic))
    st.markdown(content)