GPT-3.5 PDF Assistant
Ask questions from any PDF and get AI-powered voice answers.
Note:
This application is currently available in French.
PDF must be a text PDF if app doesn't work try a PDF with less pictures/images (no computer vision embeeded at this point).
It is still under improvement : large PDF files may take time to process, and a loading indicator will be added soon for better UX.
The app is hosted on Render's free tier, which can result in longer cold start times when launching the service.
Key Features
| Feature | Description |
|---|---|
| PDF Indexing | Extract and split PDF content, generate OpenAI embeddings, and store them in a FAISS index. |
| AI-Powered Q&A | Use semantic search to find relevant document content and generate contextual answers using GPT-3.5. |
| Text-to-Speech | Convert generated answers into French audio using IBM Watson Text-to-Speech, downloadable by users. |
Tech Stack
Integration of GPT-3.5 language model to understand user queries and generate contextual answers.
Efficient similarity search library used to index and retrieve the most relevant document chunks.
Lightweight backend framework used to handle API routes and application logic.
Converts the generated text answers into natural French speech, downloadable as audio files.
Python library used to extract raw text from PDF documents before indexing.
Cloud deployment platform used to host the app and manage serverless execution.