Questions,
answered straight.
The things people ask before the first install. Where the answer is no, it says no.
What is AskDocs, in one sentence?
A self-hosted system that indexes your documents so you can ask questions in plain language and get answers with citations back to the source — running entirely on your own infrastructure.
Why not just use ChatGPT, Claude or Gemini?
Three reasons, and only the third is about the model.
- Privacy. Plenty of teams — legal, healthcare, finance, government — simply cannot upload their documents to a third-party service.
- Coverage. A public model knows the public internet. It does not know your contracts, your minutes or your handover notes.
- Citations. AskDocs always shows which document an answer came from, so you can check it.
What do I need to run it?
Docker 24 or newer, 8 GB of RAM as a floor (16 GB is comfortable), and roughly 8 GB of disk for the stack plus room for your own documents. Full detail on the self-hosting page.
Do I need a GPU?
No. Embeddings run on CPU by default, with no API key and no accelerator. A GPU speeds up indexing if you have one, but nothing in the install assumes it.
Which file types does it read?
PDF, DOCX, PPTX, XLSX, Markdown, plain text, and images. Scanned PDFs and images go through OCR automatically — you do not have to convert anything first.
Does anything leave my server?
With a local chat model configured: nothing. No telemetry, no analytics, no licence check.
If you configure a hosted model instead, your question and the retrieved passages are sent to that provider. That is how hosted models work, and you should know it before choosing one.
The choice is a single environment variable, and it is yours.
Do you train on my documents?
No, and it is not architecturally possible for us to: we do not operate your instance and never receive your data. AskDocs is a retrieval system — the model reads retrieved passages at the moment you ask and retains nothing afterwards.
Can a team share one instance without seeing each other's files?
Yes. Workspaces are the isolation boundary, and inside one you can set per-document permissions for a role, a person or a group. Permissions are applied before retrieval runs, so a document you cannot open also cannot influence an answer you receive.
Is there an audit trail?
There is an append-only log of every upload, deletion, permission change and membership change, with who did it and when.
To be precise about the limit: the current release does not log which user asked which question. If query-level auditing is a compliance requirement for you, AskDocs does not meet it today.
What licence is it under, and what does that cost?
The application is open source under the MIT licence. There is no fee, no per-seat pricing and no paid tier — the self-hosted edition is the entire product. See the licence note for how that relates to this website's own content.
How large an archive can it handle?
It is tuned for small and mid-sized teams. Past roughly a million documents you want a dedicated enterprise search product instead — that is a different class of tool and we would rather say so than let you find out at scale.
Who supports it if something breaks?
The self-hosted edition is community-supported. There is no paid support contract, no SLA, and no on-call rota behind it. For questions you can use the contact form, and for bugs, the issue tracker in the repository.