openai_tts_tool.py

import os
import openai
import PyPDF2
from deep_translator import GoogleTranslator
from dotenv import load_dotenv
import tiktoken
import pytesseract
import fitz  # PyMuPDF for PDF processing
import docx  # For processing DOCX files
from PIL import Image

# Load environment variables
load_dotenv()

# Initialize OpenAI client
openai_api_key = os.getenv("OPENAI_API_KEY")
client = openai.OpenAI(api_key=openai_api_key)

# Define model specifications
MODEL_SPECS = {
    'gpt-4o': {
        'max_context_tokens': 128000,
        'max_output_tokens': 4096,
    },
    'gpt-4o-mini': {
        'max_context_tokens': 128000,
        'max_output_tokens': 16384,
    },
    'gpt-4': {
        'max_context_tokens': 8192,
        'max_output_tokens': 8192,
    },
    # Add other models as needed
}

# Set the path for Tesseract OCR (only needed on Windows)
pytesseract.pytesseract.tesseract_cmd = r'C:\\Program Files\\Tesseract-OCR\\tesseract.exe'  # Adjust path accordingly

# Function to extract text from PDF, using OCR for scanned documents
def extract_text_from_pdf(pdf_path):
    doc = fitz.open(pdf_path)
    text = ""
    for page_num in range(doc.page_count):
        page = doc[page_num]
        page_text = page.get_text()

        # If no text (i.e., scanned PDF), use OCR
        if not page_text.strip():
            pix = page.get_pixmap()
            img = Image.frombytes("RGB", [pix.width, pix.height], pix.samples)
            page_text = pytesseract.image_to_string(img)

        text += page_text
    return text

# Function to handle .docx files
def load_docx_file(docx_path):
    doc = docx.Document(docx_path)
    full_text = []
    for para in doc.paragraphs:
        full_text.append(para.text)
    return '\n'.join(full_text)

# Function to handle .txt files
def load_txt_file(txt_path):
    with open(txt_path, 'r', encoding='utf-8') as f:
        return f.read()

# Function to handle file based on its extension
def load_file_based_on_extension(file_path):
    if file_path.endswith('.pdf'):
        return extract_text_from_pdf(file_path)
    elif file_path.endswith('.docx'):
        return load_docx_file(file_path)
    elif file_path.endswith('.txt'):
        return load_txt_file(file_path)
    else:
        raise ValueError(f"Unsupported file format: {file_path}")

# Function to process a folder and index all files within it
def process_folder(folder_path):
    documents = []
    for filename in os.listdir(folder_path):
        file_path = os.path.join(folder_path, filename)
        if os.path.isfile(file_path):
            try:
                text = load_file_based_on_extension(file_path)
                documents.append(text)
            except ValueError as e:
                print(f"Skipping unsupported file: {file_path} ({e})")
    return ' '.join(documents)  # Combine all documents text

# Function to count tokens
def count_tokens(text, model_name):
    encoding = tiktoken.encoding_for_model(model_name)
    num_tokens = len(encoding.encode(text))
    return num_tokens

# Function to split text into chunks
def split_text_into_chunks(text, max_tokens, model_name):
    encoding = tiktoken.encoding_for_model(model_name)
    tokens = encoding.encode(text)
    chunks = []
    start = 0
    text_length = len(tokens)

    while start < text_length:
        end = start + max_tokens
        chunk_tokens = tokens[start:end]
        chunk_text = encoding.decode(chunk_tokens)
        chunks.append(chunk_text)
        start = end

    return chunks

# Modified summarize_text function
def summarize_text(text, length, model_name, additional_prompt):
    model_specs = MODEL_SPECS.get(model_name)
    if not model_specs:
        raise ValueError(f"Model specifications not found for model {model_name}")

    max_output_tokens = model_specs['max_output_tokens']
    max_context_tokens = model_specs['max_context_tokens']

    if length > max_output_tokens:
        length = max_output_tokens

    input_token_count = count_tokens(text, model_name)
    buffer_tokens = 500

    if input_token_count + buffer_tokens + length > max_context_tokens:
        max_chunk_tokens = max_context_tokens - buffer_tokens - length
        chunks = split_text_into_chunks(text, max_chunk_tokens, model_name)
        summaries = [summarize_text(chunk, length, model_name, additional_prompt) for chunk in chunks]
        combined_summary = ' '.join(summaries)
        final_summary = summarize_text(combined_summary, length, model_name, additional_prompt)
        return final_summary
    else:
        prompt = (
            f"Please provide a clear and concise summary of the following text in approximately {length} words. "
            "Ensure that the summary does not include any special characters, symbols, or markdown formatting. "
            "Use plain language and proper punctuation."
        )
        if additional_prompt:
            prompt += f"\n\nAdditional instructions: {additional_prompt}"
        prompt += f"\n\nText to summarize:\n{text}"

        # Use the chat completion as per your snippet
        completion = client.chat.completions.create(
            model=model_name,
            messages=[
                {"role": "system", "content": "You are a helpful assistant"},
                {"role": "user", "content": prompt}
            ],
            max_tokens=length
        )
        return completion.choices[0].message.content.strip()

# Function to calculate summary length based on desired audio duration
def calculate_summary_length_by_duration(duration_minutes, voice_speed):
    words_per_minute = 150 if voice_speed == 'normal' else 120
    summary_length = int(duration_minutes * words_per_minute)
    return summary_length

# Function to translate the summarized text using deep-translator
def translate_text(text, target_language):
    translated = GoogleTranslator(source='auto', target=target_language).translate(text)
    return translated

# Function to estimate audio duration
def estimate_audio_duration(text, voice_speed):
    word_count = len(text.split())
    words_per_minute = 150 if voice_speed == 'normal' else 120
    duration_minutes = word_count / words_per_minute
    duration_seconds = duration_minutes * 60
    return duration_seconds

# Function to convert text to audio using OpenAI TTS-1
def text_to_speech_openai(text, audio_path, voice, speed):
    response = client.audio.speech.create(
        model="tts-1-hd",
        voice=voice,
        input=text
    )
    response.stream_to_file(audio_path)

def process_input(pdf_path=None, input_text=None, summary_length=None, voice=None, language=None, voice_speed=None, model_name=None, additional_prompt=None, generate_audio=True, folder_path=None):
    if folder_path:
        extracted_text = process_folder(folder_path)
    elif pdf_path:
        extracted_text = load_file_based_on_extension(pdf_path)
    elif input_text:
        extracted_text = input_text
    else:
        raise ValueError("No input provided for processing.")

    summary_text = summarize_text(extracted_text, summary_length, model_name, additional_prompt)
    translated_summary = translate_text(summary_text, language)
    estimated_audio_duration = estimate_audio_duration(translated_summary, voice_speed)

    base_filename = os.path.splitext(os.path.basename(pdf_path or 'document'))[0]
    audio_file_path = os.path.join('uploads', f"{base_filename}_audio_{language}.mp3")
    summary_file_path = os.path.join('uploads', f"{base_filename}_summary_{language}.txt")

    with open(summary_file_path, "w", encoding="utf-8") as summary_file:
        summary_file.write(translated_summary)

    if generate_audio:
        text_to_speech_openai(translated_summary, audio_file_path, voice, voice_speed)

    return translated_summary, audio_file_path if generate_audio else None, summary_file_path, estimated_audio_duration