Terms of use
SecureRAG is a free information tool you run in your browser. These terms describe what the service does, what you may use it for, who owns the documents and the code, and what happens when a local model gets something wrong.
Last updated:
Using this site means you accept these terms. They are short because the service is simple: a free, local tool for searching your own documents, provided as it is, with no promise that it will always be available.
1. What the service is
SecureRAG is a browser-based document question-answering tool. Documents are parsed, indexed and searched on your device; the site itself is a set of static files. There is no account, no server-side processing and no fee. It is free to use, including for work and commercial purposes, and there is nothing to subscribe to or cancel.
- No account, no subscription, no payment details, at any point.
- No service level agreement. We do not promise uptime, response times, support hours or that a particular browser will keep working with it.
- No guarantee of continuity. Features may change, and the project may be discontinued, without notice beyond the changelog.
- Data is yours to keep locally and yours to lose locally: we hold no backup and cannot restore an index you cleared.
2. Acceptable use
Use the tool for documents you are entitled to use. That is the whole rule, and the rest of this section is what it means in practice.
- Do not use it to process files you have no right to access, such as confidential material belonging to someone else, or documents obtained without permission.
- Do not rely on the tool to keep someone else’s confidential material safe on a device that is not yours. Everything runs locally, but local storage on a shared machine is still shared.
- Do not attempt to make the site serve content that misrepresents what it does, including framing it inside another product in a way that hides the local-only design.
- Do not use it to build a service that claims to upload your documents, since that would misdescribe how this software behaves.
- Do not use it in a way that breaks the law where you are, including laws about the material you choose to index.
3. What you own, and what we own
Your documents, your questions and the answers you generate remain yours. We claim no licence over them, and we cannot read them. Anything you index or export is your responsibility, including the duty to keep it accurate and lawfully handled.
The site itself — its code, interface, text, page copy, illustrations and design — belongs to the project and its contributors. Read it, learn from it, quote it with attribution, and use the tool to do your work. Do not republish the site wholesale as your own product, and do not strip attribution from code that carries it.
The models downloaded to your browser are third-party works with their own licences, listed with each model on the models page. The licence on the model page governs your use of that model, and it is the upstream project’s terms, not ours.
4. This is an information tool, not professional advice
Answers produced here are a summary of documents you supplied. They are not legal, medical, financial, tax or safety advice, and they do not create a professional relationship of any kind. A local embedding model and a small language model cannot weigh facts, jurisdiction, timing or consequences the way a qualified professional can.
- A contract summary can miss the clause that decides the matter, and the retrieval tier will faithfully return the wrong paragraph if the document is indexed badly.
- A medical or dosage question is not answered safely by pattern matching over a PDF you happened to have.
- A financial or tax figure taken from a stale document is stale, and the tool cannot know which documents you left out.
- Anything with legal, medical, financial or safety consequences needs a human professional to review it. Check the cited passage against the source document itself, not against our summary of it.
5. Local models get things wrong
Small models running on a laptop are useful and imperfect. They may misread a table, merge two clauses, answer partially, or produce text that reads confidently and is wrong. The optional generation tier writes fluent prose from retrieved passages, which makes an error harder to notice, not less likely. Every claim in the interface is tied to a citation so that verification is one click away, and the retrieval-only tier answers strictly from the passages it found.
Where the interface reports a failure — an unreadable scan, an oversized file, a collection over its limit, “not found in your documents” — that report is part of the tool working correctly. Treat it as information, and do not treat a fluent answer as confirmation.
6. Limits of liability
The service is provided as it is, without warranties of any kind, express or implied, including fitness for a particular purpose and accuracy of results. To the fullest extent permitted by law, we are not liable for indirect, incidental or consequential losses arising from use of the site: lost profits, lost data, decisions taken on the basis of an answer, or the consequences of a document being indexed on a shared device.
Since the service is free and we never receive your documents, the arrangement is that you carry the risk of using it and we carry the cost of running it. Where the law does not allow a limitation of this kind, this section applies only so far as it can. Nothing here excludes rights you have that cannot be excluded.
7. Advertising
The site is supported by Google AdSense. Ad slots are labelled, up to six per page, and kept out of the chat workspace. Ad scripts load personalized only after you agree, and the privacy policy describes what those cookies do and how to refuse them. If you paid for the site in any other way, no advertising would be necessary, and there is no way to pay, which is why it is there.
8. Changes to these terms
These terms may change as the service changes. The last-updated date at the top of this page always reflects the current version, and a change that affects your rights is written into the section it concerns rather than buried in a summary. Continuing to use the site after a change means you accept the updated terms; if you disagree with them, stop using the site and clear its site data.
9. Contact
Questions about these terms, correction requests and reports of misuse go to guweiicy@gmail.com. The site is run by an individual and a few contributors, not a company: there is no legal entity, registered address or support desk behind it, just an address that reaches the people who build it.