import os
import gradio as gr
from huggingface_hub import InferenceClient
import markdown

client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

# Function to read and process Markdown files from the 'data' directory and its subfolders
def load_markdown_files(data_folder='data'):
    documents = []
    for root, dirs, files in os.walk(data_folder):
        for filename in files:
            if filename.endswith('.md'):
                file_path = os.path.join(root, filename)
                try:
                    with open(file_path, 'r', encoding='utf-8') as file:
                        content = file.read()
                        # Convert Markdown to plain text if needed
                        html_content = markdown.markdown(content)
                        documents.append(html_content)  # Store HTML content or plain text
                except Exception as e:
                    print(f"Error reading {file_path}: {e}")
    return documents

# Load documents at startup
documents = load_markdown_files()

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    messages = [{"role": "system", "content": system_message}]

    # Retrieve relevant context from loaded documents based on the user message
    relevant_contexts = retrieve_relevant_context(message, documents)

    # Add retrieved contexts to the messages for better responses
    messages.append({"role": "context", "content": " ".join(relevant_contexts)})

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    response = ""

    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content
        response += token
        yield response

def retrieve_relevant_context(query, documents):
    # Simple keyword matching to find relevant documents
    relevant_contexts = []
    
    for doc in documents:
        if query.lower() in doc.lower():  # Basic keyword search
            relevant_contexts.append(doc)

    return relevant_contexts[:3]  # Return top 3 relevant contexts

demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot. Only answer questions that you have knowledge of, in the language of Traditional Chinese.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)
#To create a public link, set `share=True` in `launch()`.
if __name__ == "__main__":
    demo.launch(share=True)