import os
import json
import re
import gradio as gr
import pandas as pd
import requests
import random
import urllib.parse
from tempfile import NamedTemporaryFile
from typing import List
from bs4 import BeautifulSoup
from langchain_core.prompts import ChatPromptTemplate
from langchain_community.vectorstores import FAISS
from langchain_community.document_loaders import PyPDFLoader
from langchain_core.output_parsers import StrOutputParser
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.llms import HuggingFaceHub
from langchain_core.runnables import RunnableParallel, RunnablePassthrough
from langchain_core.documents import Document
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity

huggingface_token = os.environ.get("HUGGINGFACE_TOKEN")

# Memory database to store question-answer pairs
memory_database = {}
conversation_history = []

def load_and_split_document_basic(file):
    """Loads and splits the document into pages."""
    loader = PyPDFLoader(file.name)
    data = loader.load_and_split()
    return data

def load_and_split_document_recursive(file: NamedTemporaryFile) -> List[Document]:
    """Loads and splits the document into chunks."""
    loader = PyPDFLoader(file.name)
    pages = loader.load()
    
    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=1000,
        chunk_overlap=200,
        length_function=len,
    )
    
    chunks = text_splitter.split_documents(pages)
    return chunks

def get_embeddings():
    return HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")

def create_or_update_database(data, embeddings):
    if os.path.exists("faiss_database"):
        db = FAISS.load_local("faiss_database", embeddings, allow_dangerous_deserialization=True)
        db.add_documents(data)
    else:
        db = FAISS.from_documents(data, embeddings)
    db.save_local("faiss_database")

def clear_cache():
    if os.path.exists("faiss_database"):
        os.remove("faiss_database")
        return "Cache cleared successfully."
    else:
        return "No cache to clear."

def get_similarity(text1, text2):
    vectorizer = TfidfVectorizer().fit_transform([text1, text2])
    return cosine_similarity(vectorizer[0:1], vectorizer[1:2])[0][0]

prompt = """
Answer the question based on the following information:

Conversation History:
{history}

Context from documents:
{context}

Current Question: {question}

If the question is referring to the conversation history, use that information to answer.
If the question is not related to the conversation history, use the context from documents to answer.
If you don't have enough information to answer, say so.

Provide a concise and direct answer to the question:
"""

def get_model(temperature, top_p, repetition_penalty):
    return HuggingFaceHub(
        repo_id="mistralai/Mistral-7B-Instruct-v0.3",
        model_kwargs={
            "temperature": temperature,
            "top_p": top_p,
            "repetition_penalty": repetition_penalty,
            "max_length": 1000
        },
        huggingfacehub_api_token=huggingface_token
    )

def generate_chunked_response(model, prompt, max_tokens=1000, max_chunks=5):
    full_response = ""
    for i in range(max_chunks):
        chunk = model(prompt + full_response, max_new_tokens=max_tokens)
        chunk = chunk.strip()
        if chunk.endswith((".", "!", "?")):
            full_response += chunk
            break
        full_response += chunk
    return full_response.strip()

def manage_conversation_history(question, answer, history, max_history=5):
    history.append({"question": question, "answer": answer})
    if len(history) > max_history:
        history.pop(0)
    return history

def is_related_to_history(question, history, threshold=0.3):
    if not history:
        return False
    history_text = " ".join([f"{h['question']} {h['answer']}" for h in history])
    similarity = get_similarity(question, history_text)
    return similarity > threshold

def extract_text_from_webpage(html):
    soup = BeautifulSoup(html, 'html.parser')
    for script in soup(["script", "style"]):
        script.extract()  # Remove scripts and styles
    text = soup.get_text()
    lines = (line.strip() for line in text.splitlines())
    chunks = (phrase.strip() for line in lines for phrase in line.split("  "))
    text = '\n'.join(chunk for chunk in chunks if chunk)
    return text

_useragent_list = [
    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36",
    "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36",
    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Edge/91.0.864.59 Safari/537.36",
    "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Edge/91.0.864.59 Safari/537.36",
    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Safari/537.36",
    "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Safari/537.36",
]

def google_search(term, num_results=5, lang="en", timeout=5, safe="active", ssl_verify=None):
    escaped_term = urllib.parse.quote_plus(term)
    start = 0
    all_results = []
    max_chars_per_page = 8000  # Limit the number of characters from each webpage to stay under the token limit

    print(f"Starting Google search for term: '{term}'")

    with requests.Session() as session:
        while start < num_results:
            try:
                user_agent = random.choice(_useragent_list)
                headers = {
                    'User-Agent': user_agent
                }
                resp = session.get(
                    url="https://www.google.com/search",
                    headers=headers,
                    params={
                        "q": term,
                        "num": num_results - start,
                        "hl": lang,
                        "start": start,
                        "safe": safe,
                    },
                    timeout=timeout,
                    verify=ssl_verify,
                )
                resp.raise_for_status()
                print(f"Successfully retrieved search results page (start={start})")
            except requests.exceptions.RequestException as e:
                print(f"Error retrieving search results: {e}")
                break

            soup = BeautifulSoup(resp.text, "html.parser")
            result_block = soup.find_all("div", attrs={"class": "g"})
            if not result_block:
                print("No results found on this page")
                break
            
            print(f"Found {len(result_block)} results on this page")
            for result in result_block:
                link = result.find("a", href=True)
                if link:
                    link = link["href"]
                    print(f"Processing link: {link}")
                    try:
                        webpage = session.get(link, headers=headers, timeout=timeout)
                        webpage.raise_for_status()
                        visible_text = extract_text_from_webpage(webpage.text)
                        if len(visible_text) > max_chars_per_page:
                            visible_text = visible_text[:max_chars_per_page] + "..."
                        all_results.append({"link": link, "text": visible_text})
                        print(f"Successfully extracted text from {link}")
                    except requests.exceptions.RequestException as e:
                        print(f"Error retrieving webpage content: {e}")
                        all_results.append({"link": link, "text": None})
                else:
                    print("No link found for this result")
                    all_results.append({"link": None, "text": None})
            start += len(result_block)

    print(f"Search completed. Total results: {len(all_results)}")
    print("Search results:")
    for i, result in enumerate(all_results, 1):
        print(f"Result {i}:")
        print(f"  Link: {result['link']}")
        if result['text']:
            print(f"  Text: {result['text'][:100]}...")  # Print first 100 characters
        else:
            print("  Text: None")
    print("End of search results")

    if not all_results:
        print("No search results found. Returning a default message.")
        return [{"link": None, "text": "No information found in the web search results."}]

    return all_results

def ask_question(question, temperature, top_p, repetition_penalty, web_search):
    global conversation_history
    
    if not question:
        return "Please enter a question."
    
    if question in memory_database and not web_search:
        answer = memory_database[question]
    else:
        model = get_model(temperature, top_p, repetition_penalty)
        embed = get_embeddings()
        
        if web_search:
            search_results = google_search(question)
            context_str = "\n".join([result["text"] for result in search_results if result["text"]])
            
            # Convert web search results to Document format
            web_docs = [Document(page_content=result["text"], metadata={"source": result["link"]}) for result in search_results if result["text"]]
            
            # Load or create the vector database
            if os.path.exists("faiss_database"):
                database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
                database.add_documents(web_docs)
            else:
                database = FAISS.from_documents(web_docs, embed)
                database.save_local("faiss_database")
            
            prompt_template = """
            Answer the question based on the following web search results:
            Web Search Results:
            {context}
            Current Question: {question}
            If the web search results don't contain relevant information, state that the information is not available in the search results.
            Provide a concise and direct answer to the question without mentioning the web search or these instructions:
            """
            prompt_val = ChatPromptTemplate.from_template(prompt_template)
            formatted_prompt = prompt_val.format(context=context_str, question=question)
        else:
            database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
            history_str = "\n".join([f"Q: {item['question']}\nA: {item['answer']}" for item in conversation_history])
            
            if is_related_to_history(question, conversation_history):
                context_str = "No additional context needed. Please refer to the conversation history."
            else:
                retriever = database.as_retriever()
                relevant_docs = retriever.get_relevant_documents(question)
                context_str = "\n".join([doc.page_content for doc in relevant_docs])
            
            prompt_val = ChatPromptTemplate.from_template(prompt)
            formatted_prompt = prompt_val.format(history=history_str, context=context_str, question=question)
        
        answer = generate_chunked_response(model, formatted_prompt)
        answer = re.split(r'Question:|Current Question:', answer)[-1].strip()
        
        # Remove any remaining prompt instructions from the answer
        answer_lines = answer.split('\n')
        answer = '\n'.join(line for line in answer_lines if not line.startswith('If') and not line.startswith('Provide'))
        
        if not web_search:
            memory_database[question] = answer
    
    if not web_search:
        conversation_history = manage_conversation_history(question, answer, conversation_history)
    
    return answer

def update_vectors(files, use_recursive_splitter):
    if not files:
        return "Please upload at least one PDF file."
    
    embed = get_embeddings()
    total_chunks = 0
    
    for file in files:
        if use_recursive_splitter:
            data = load_and_split_document_recursive(file)
        else:
            data = load_and_split_document_basic(file)
        create_or_update_database(data, embed)
        total_chunks += len(data)
    
    return f"Vector store updated successfully. Processed {total_chunks} chunks from {len(files)} files."

def extract_db_to_excel():
    embed = get_embeddings()
    database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
    
    documents = database.docstore._dict.values()
    data = [{"page_content": doc.page_content, "metadata": json.dumps(doc.metadata)} for doc in documents]
    df = pd.DataFrame(data)
    
    with NamedTemporaryFile(delete=False, suffix='.xlsx') as tmp:
        excel_path = tmp.name
        df.to_excel(excel_path, index=False)
    
    return excel_path

def export_memory_db_to_excel():
    data = [{"question": question, "answer": answer} for question, answer in memory_database.items()]
    df_memory = pd.DataFrame(data)
    
    data_history = [{"question": item["question"], "answer": item["answer"]} for item in conversation_history]
    df_history = pd.DataFrame(data_history)
    
    with NamedTemporaryFile(delete=False, suffix='.xlsx') as tmp:
        excel_path = tmp.name
        with pd.ExcelWriter(excel_path, engine='openpyxl') as writer:
            df_memory.to_excel(writer, sheet_name='Memory Database', index=False)
            df_history.to_excel(writer, sheet_name='Conversation History', index=False)
    
    return excel_path

# Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Chat with your PDF documents")
    
    with gr.Row():
        file_input = gr.Files(label="Upload your PDF documents", file_types=[".pdf"])
        update_button = gr.Button("Update Vector Store")
        use_recursive_splitter = gr.Checkbox(label="Use Recursive Text Splitter", value=False)
    
    update_output = gr.Textbox(label="Update Status")
    update_button.click(update_vectors, inputs=[file_input, use_recursive_splitter], outputs=update_output)
    
    with gr.Row():
        with gr.Column(scale=2):
            chatbot = gr.Chatbot(label="Conversation")
            question_input = gr.Textbox(label="Ask a question about your documents")
            submit_button = gr.Button("Submit")
        with gr.Column(scale=1):
            temperature_slider = gr.Slider(label="Temperature", minimum=0.0, maximum=1.0, value=0.5, step=0.1)
            top_p_slider = gr.Slider(label="Top P", minimum=0.0, maximum=1.0, value=0.9, step=0.1)
            repetition_penalty_slider = gr.Slider(label="Repetition Penalty", minimum=1.0, maximum=2.0, value=1.0, step=0.1)
            web_search_checkbox = gr.Checkbox(label="Enable Web Search", value=False)
    
    def chat(question, history):
        answer = ask_question(question, temperature_slider.value, top_p_slider.value, repetition_penalty_slider.value, web_search_checkbox.value)
        history.append((question, answer))
        return "", history
    
    submit_button.click(chat, inputs=[question_input, chatbot], outputs=[question_input, chatbot])
    
    extract_button = gr.Button("Extract Database to Excel")
    excel_output = gr.File(label="Download Excel File")
    extract_button.click(extract_db_to_excel, inputs=[], outputs=excel_output)
    
    export_memory_button = gr.Button("Export Memory Database to Excel")
    memory_excel_output = gr.File(label="Download Memory Excel File")
    export_memory_button.click(export_memory_db_to_excel, inputs=[], outputs=memory_excel_output)
    
    clear_button = gr.Button("Clear Cache")
    clear_output = gr.Textbox(label="Cache Status")
    clear_button.click(clear_cache, inputs=[], outputs=clear_output)

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