FAQ and troubleshooting

General

Is it really free? Yes. ReadAnyDoc is Apache-2.0. You pay only your model provider for the calls it makes; every result reports what they cost.

Is it as good as Reducto? Ahead on reading text and multi-column pages, level on tables and forms, behind on page layout, formulas, removing headers and footers, and speed. See Benchmarks.

Which models can I use? Any Claude model, OpenAI models, or a model you host behind an OpenAI-compatible server. See Choosing models. Only Claude Haiku 5.5 is benchmarked.

Do I need a server? No. The library runs in your process. Run the server for Reducto's API, other languages, or a shared service.

Where do my documents go? To your machine and the model provider you configure, under your account. Nothing goes to the ReadAnyDoc project.

Is the confidence a probability? No. It is a tier (high, low) from how a value was found. Use it to decide what to check first.

Troubleshooting

Symptom	Cause and fix
The first parse takes minutes	The layout and OCR weights (100 to 200 MB) download once. Later runs are normal speed
warnings says the layout model is not installed	pip install "readanydoc[engine]"
401 or authentication_error from the provider	ANTHROPIC_API_KEY / OPENAI_API_KEY missing or wrong where the code runs
A page is empty and warnings mentions content filtering	The provider declined to return that page. Try another model for that document
415 for a .doc, .xls, .ppt or .heic file	Install LibreOffice or a HEIC decoder; the error names the command
442	The PDF is password protected: send settings.document_password
504 timeout	Raise READANYDOC_DOCUMENT_TIMEOUT_SECONDS, or split the document
Rate-limit errors	Lower READANYDOC_API_CONCURRENCY or raise your provider tier
Reducto's SDK says an API key is required	Pass any string; it is only checked when READANYDOC_API_KEY is set

Still stuck? Open an issue with the version, mode, model and the command you ran.