Question Answering from PDFs

Question Answering from PDFs

A Gradio app that answers questions from PDF files using a retrieval model and a SQuAD2 reading model, with two sample PDFs included. Suits people who want to try document question answering in a browser.

Made by ThePixOne on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Question Answering from PDFs is a Gradio app by ThePixOne. It names a large roberta model trained on SQuAD2 and a multi-qa MPNet sentence model. The repository holds two sample PDFs, one on China and one on London, plus two text files. By its title, you ask a question and it answers from the PDFs. The page has no description, so the exact flow is not confirmed.

Who it suits: People who want to see question answering over documents working, and developers who want a compact example that combines two models in a Gradio page.

Industry | Who uses it | How they use it | Example
Education | Students of language models | Run the demo and read app.py to see how the two models cooperate | Ask a question about the London sample
Software | Developers | Use the project as a pattern for a Gradio document question tool | Swap the sample PDFs for their own
Research | Language researchers | Try an open question answering setup on a few documents | Ask the same question of both sample PDFs

Skip it if you need answers you can trust without checking the source page, because the models can pick the wrong passage.

Where to use it:

• A web browser, on the Space's page on Hugging Face
• Python on Windows, macOS or Linux, using a downloaded copy

Use case | Who | What they get back
Ask questions of sample PDFs online | Anyone curious | An answer drawn from the bundled documents
Study a two model question answering demo | Developers | A short Python project that pairs two named models

How to set it up:

• Nothing to install for the online version
• Python on your computer for a local copy
• The packages listed in requirements.txt

1. Install the Hugging Face hub package.
2. Download the Space's files to a folder.
3. Move into that folder.
4. Install gradio, then the packages from requirements.txt.
5. Start the app with app.py.

pip install -U huggingface_hub
hf download ThePixOne/open_domain_qa --repo-type space --local-dir open_domain_qa
cd open_domain_qa
pip install gradio
pip install -r requirements.txt
python app.py

How to use it: Open the Space page in a browser and type a question into the page. For a local copy, run the app and open the page Gradio opens for you.

python app.py

Tool | What it does
python app.py | Starts the Gradio app on your computer
hf download | Copies the Space's files into a local folder

Example: Open the Space and ask a plain question about London, such as its size or history. The app should answer from the London sample PDF. The page does not say how answers are shown, so use what the page displays.

For newcomers and experts: If you are new to this, ask one short factual question about a topic in the sample PDFs and check the answer against the PDF. Then try a question the PDFs do not cover and see what happens.

If you know the field, the pairing suggests a SQuAD2 reader, roberta-large-squad2, behind a multi-qa MPNet retriever. A dot-product sentence embedder would pick passages first, then the reader would pull out an answer. The page does not confirm this.

You type a question
The MPNet model finds passages
The roberta model reads them
Gradio shows the answer

Good to know:

• The page gives no description, accuracy figures or limits on PDF size.
• The online version needs nothing installed, and a local copy needs Python.
• License: wtfpl.
• Cost to run: none named, only the packages in requirements.txt.
• Data it reaches: the sample PDFs in its repository and the question you type. A local copy downloads the two models from Hugging Face.
• Not for: the source names none. Check each answer against the PDF page it came from.

On Hugging Face, 13 November 2026

• Likes: 31
• Built with: gradio
• License: wtfpl

Links

• Space: https://huggingface.co/spaces/ThePixOne/open_domain_qa
• Files: https://huggingface.co/spaces/ThePixOne/open_domain_qa/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.