Offline PDF Voice GPT Assistant

I need a standalone Windows application that bundles a local GPT-style language model with a voice layer powered by Google Assistant. The workflow is simple: I open any PDF, the model ingests it locally, and then—through voice alone—I can ask questions about the document and get spoken answers back.

Key points I must see working:
• Runs entirely on Windows without needing command-line setup.
• Uses the Google Assistant voice interface for both input and spoken output.
• Processes PDFs offline: the text is extracted, fed to the model, and retained only on my machine.
• Voice queries such as “What is the author’s main argument on page 5?” return an accurate, concise spoken answer within a reasonable response time.

I’m open to whichever open-source LLM, PDF parsing library, or local vector store you prefer, as long as everything stays offline after the initial Google Assistant handshake. Package any runtimes you rely on, provide a clear installer, and include brief setup instructions so I can start asking questions the moment I launch the app.