app.py

import os
import io
import re
import numpy as np
import pytesseract
from PIL import Image
from typing import List
from sentence_transformers import SentenceTransformer
from langchain_community.vectorstores import Chroma
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.embeddings import SentenceTransformerEmbeddings
from groq import Groq
import gradio as gr
import requests

# Ensure the Tesseract OCR path is set correctly
pytesseract.pytesseract.tesseract_cmd = r'/usr/bin/tesseract'

GROQ_API_KEY = "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"

def preprocess_text(text):
    try:
        text = text.replace('\n', ' ').replace('\r', ' ')
        text = re.sub(r'[^\x00-\x7F]+', ' ', text)
        text = text.lower()
        text = re.sub(r'[^\w\s]', '', text)
        text = re.sub(r'\s+', ' ', text).strip()
        return text
    except Exception as e:
        print("Failed to preprocess text:", e)
        return ""

def fetch_text_file_from_huggingface_space():
    url = "https://huggingface.co/spaces/Luciferalive/goosev9/blob/main/extracted_text.txt"
    try:
        response = requests.get(url)
        response.raise_for_status()
        text_content = response.text
        print("Successfully downloaded the text file")
        return text_content
    except Exception as e:
        print(f"Failed to download the text file: {e}")
        return ""

def create_vector_store(text_content):
    embeddings = SentenceTransformerEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
    texts = text_splitter.split_text(text_content)
    if not texts:
        print("No text chunks created.")
        return None

    vector_store = Chroma.from_texts(texts, embeddings, collection_name="insurance_cosine")
    print("Vector DB Successfully Created!")
    return vector_store

def load_vector_store():
    embeddings = SentenceTransformerEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
    try:
        db = Chroma(embedding_function=embeddings, collection_name="insurance_cosine")
        print("Vector DB Successfully Loaded!")
        return db
    except Exception as e:
        print("Failed to load Vector DB:", e)
        return None

def answer_query(query):
    try:
        vector_store = load_vector_store()
        if not vector_store:
            return None

        docs = vector_store.similarity_search(query)
        print(f"\n\nDocuments retrieved: {len(docs)}")

        if not docs:
            print("No documents match the query.")
            return None

        docs_content = [doc.page_content for doc in docs]
        all_docs_content = " ".join(docs_content)

        client = Groq(api_key=GROQ_API_KEY)
        template = """
                ### [INST] Instruction:
                You are an AI assistant named Goose. Your purpose is to provide accurate, relevant, and helpful information to users in a friendly, warm, and supportive manner, similar to ChatGPT. When responding to queries, please keep the following guidelines in mind:
                - When someone says hi, or small talk, only respond in a sentence.
                - Retrieve relevant information from your knowledge base to formulate accurate and informative responses.
                - Always maintain a positive, friendly, and encouraging tone in your interactions with users.
                - Strictly write crisp and clear answers, don't write unnecessary stuff.
                - Only answer the asked question, don't hallucinate or print any pre-information.
                - After providing the answer, always ask for any other help needed in the next paragraph.
                - Writing in bullet format is our top preference.
                Remember, your goal is to be a reliable, friendly, and supportive AI assistant that provides accurate information while creating a positive user experience, just like ChatGPT. Adapt your communication style to best suit each user's needs and preferences.
                ### Docs: {docs}
                ### Question: {question}
                """

        chat_completion = client.chat.completions.create(
            messages=[
                {
                    "role": "system", 
                    "content": template.format(docs=all_docs_content, question=query)
                },
                {
                    "role": "user",
                    "content": query
                }
            ],
            model="llama3-8b-8192",
        )

        answer = chat_completion.choices[0].message.content.strip()
        return answer
    except Exception as e:
        print("An error occurred while getting the answer: ", str(e))
        return None

def process_query(query):
    try:
        response = answer_query(query)
        if response:
            return "Answer: " + response
        else:
            return "No answer found."
    except Exception as e:
        print("An error occurred while getting the answer: ", str(e))
        return "An error occurred: " + str(e)


# Set up the Gradio interface
def launch_assistant():
    text_content = fetch_text_file_from_huggingface_space()
    if not text_content.strip():
        print("No text content fetched.")
        return

    vector_store = create_vector_store(text_content)
    if not vector_store:
        print("Failed to create Vector DB.")
        return

    iface = gr.Interface(
        fn=process_query,
        inputs=gr.Textbox(lines=7, label="Enter your question"),
        outputs="text",
        title="Goose AI Assistant",
        description="Ask a question and get an answer from the AI assistant."
    )

    iface.launch()

launch_assistant()