Feature
PDF to Q&A dataset
Upload a PDF and FAQai generates canonical question-answer pairs grounded in that file, plus the other dataset types your plan includes. You export the result into your own RAG stack.
What you can put in
Text-based PDF, DOCX, TXT, and Markdown. Paid plans also OCR scanned PDF pages that have little extractable text. You can ingest a public page or a same-domain crawl instead of a file. The crawler reads static HTML, honours robots.txt, and stops at the plan's crawl cap.
What comes out
- RAG Chunks, the retrieval units.
- Canonical Q&A grounded in those chunks.
- Query Variants, alternative phrasings of the canonical questions.
- Evaluation and Adversarial datasets on paid plans.
Quality Insights is a separate review of coverage and weaknesses. It is not a sixth dataset type. FAQai does not store embedding vectors and does not run your production chatbot.
Free and paid
The Free plan includes JSON export and does not include OCR, evaluation datasets, or adversarial datasets. Basic, Starter, and Pro unlock the other export formats, OCR within monthly caps, and the evaluation and adversarial sets. Limits are on the pricing page.
Illustrative pair
This sample is synthetic. It is not a customer document.
{
"question": "What is the return window?",
"answer": "Items can be returned within 30 days of delivery.",
"context": "Returns: items can be returned within 30 days of delivery."
}The next step is the PDF evaluation tutorial or the user guide.