src/processor.py

import pixeltable as pxt
from pixeltable.iterators import DocumentSplitter
from pixeltable.functions import openai
import os
import requests
import tempfile
import gradio as gr

def process_document(pdf_file, api_key, voice_choice, style_choice, chunk_size, temperature, max_tokens, system_prompt, progress=gr.Progress()):
    try:
        os.environ['OPENAI_API_KEY'] = api_key
        
        progress(0.1, desc="Initializing...")
        pxt.drop_dir('document_audio', force=True)
        pxt.create_dir('document_audio')
        
        docs = pxt.create_table(
            'document_audio.documents',
            {
                'document': pxt.Document,
                'voice': pxt.String,
                'style': pxt.String,
                'mode_prompt': pxt.String
            }
        )
        
        progress(0.2, desc="Processing document...")
        docs.insert([{
            'document': pdf_file.name,
            'voice': voice_choice,
            'style': style_choice,
            'mode_prompt': system_prompt
        }])
        
        chunks = pxt.create_view(
            'document_audio.chunks',
            docs,
            iterator=DocumentSplitter.create(
                document=docs.document,
                separators='token_limit',
                limit=chunk_size
            )
        )
        
        progress(0.4, desc="Text processing...")
        chunks['content_response'] = openai.chat_completions(
            messages=[
                {
                    'role': 'system',
                    'content': docs.mode_prompt  # Use the mode-specific prompt
                },
                {'role': 'user', 'content': chunks.text}
            ],
            model='gpt-4o-mini-2024-07-18',
            max_tokens=max_tokens,
            temperature=temperature
        )
        
        chunks['content'] = chunks.content_response['choices'][0]['message']['content']

        progress(0.6, desc="Script generation...")
        chunks['script_response'] = openai.chat_completions(
            messages=[
                {
                    'role': 'system',
                    'content': f"""Convert content to audio script.
                    Style: {docs.style}
                    Format:
                    - Clear sentence structures
                    - Natural pauses (...)
                    - Term definitions when needed
                    - Proper transitions"""
                },
                {'role': 'user', 'content': chunks.content}
            ],
            model='gpt-4o-mini-2024-07-18',
            max_tokens=max_tokens,
            temperature=temperature
        )
        chunks['script'] = chunks.script_response['choices'][0]['message']['content']

        progress(0.8, desc="Audio synthesis...")
        @pxt.udf(return_type=pxt.Audio)
        def generate_audio(script: str, voice: str):
            if not script or not voice:
                return None
            try:
                response = requests.post(
                    "https://api.openai.com/v1/audio/speech",
                    headers={"Authorization": f"Bearer {api_key}"},
                    json={"model": "tts-1", "input": script, "voice": voice}
                )
                if response.status_code == 200:
                    temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.mp3')
                    temp_file.write(response.content)
                    temp_file.close()
                    return temp_file.name
            except Exception as e:
                print(f"Error in audio synthesis: {e}")
            return None

        chunks['audio'] = generate_audio(chunks.script, docs.voice)
        
        audio_path = chunks.select(chunks.audio).tail(1)['audio'][0]
        
        results = chunks.select(
            chunks.content,
            chunks.script
        ).collect()
        
        display_data = [
            [f"Segment {idx + 1}", row['content'], row['script']]
            for idx, row in enumerate(results)
        ]
        
        progress(1.0, desc="Complete")
        return display_data, audio_path, "Processing complete"
        
    except Exception as e:
        return None, None, f"Error: {str(e)}"