Privacy Policy

What ResearchLens is

ResearchLens is a small, single-purpose tool: you enter a machine learning topic, it searches public paper metadata, ranks and scores the results, and returns a Markdown report. There is no account system, no login, and no payment. This policy describes exactly what happens to your data given that scope — nothing more.

What we collect

• The topic you search for. Submitted to our backend (/api/run) to run the search/rank/score pipeline, and used to generate a report file named from a slugified version of your topic. We do not attach your search to a name, email, or account, because none exists.
• Your IP address. Seen by our server and reverse proxy as part of normal HTTP handling, and used transiently, in memory only, to enforce a per-client rate limit on the expensive search endpoint (see "Rate limiting" below). It is not written to a database and is not linked to the topics you search.
• Standard web server logs. Our reverse proxy (nginx) keeps ordinary access logs (IP, requested path, timestamp, status code) for operating the service — the same thing essentially every web server logs by default.

We do not use cookies, browser local storage, or any client-side tracking identifier, and we do not run session recording or advertising scripts. We do run cookie-free analytics and error tracking, described below — neither sets a cookie or builds a cross-site profile of you, which is why there is nothing here that needs a cookie consent banner.

Analytics and error tracking

• Analytics (Plausible). We use Plausible, a cookie-free analytics service, to see aggregate page views and a small number of named actions: starting a search, a report finishing successfully, opening a paper link, and downloading a report. These events carry only small structured details (e.g. how many papers were requested, a paper's source and reproducibility score) — never the text of your search topic, a paper's title, or anything else you typed.
• Error tracking (Sentry). If the app crashes or hits an unexpected error, technical details (the error message, a stack trace, and basic browser/server information) are sent to Sentry so we can find and fix the bug. This is only enabled once we've configured it for this deployment; it exists to catch crashes, not to monitor your usage.

Where generated reports go

Each report is written to a file on our server, named from your topic, and served back to you at a predictable URL. Report filenames are derived only from the topic text (lowercased, non-alphanumeric characters stripped) — they are not tied to a user identity, session, or IP. Because the URL is predictable, do not submit a topic that itself contains information you consider sensitive; treat generated reports as publicly retrievable by URL, not as private storage.

Third parties we talk to

• arXiv, OpenReview, and Semantic Scholar. Our server queries these public paper databases with your topic text to find candidate papers. These are the publishers/maintainers of the paper metadata we display — ResearchLens does not own or control that content, and their own terms and availability apply to it.
• Google Fonts. Your browser loads two font families directly from Google's servers (fonts.googleapis.com / fonts.gstatic.com), which — like any third-party resource — sees the requesting IP address and browser details as part of that request. No ResearchLens-specific identifier is attached to it.
• Plausible and Sentry. See "Analytics and error tracking" above.

We do not sell or share your data with anyone else, because we do not have an advertising or data-brokerage relationship with any third party.

Rate limiting

The search endpoint enforces a short per-IP rate limit and a small concurrency cap to keep the service usable for everyone and to avoid abuse of the underlying provider APIs. This check happens in server memory, is not persisted to disk, and is cleared automatically over time.

Reliability of results

Reports are generated from metadata returned by third-party academic databases, ranked and scored by our own heuristics. They are a research-assistance aid, not an authoritative or verified academic conclusion — always confirm anything important against the original paper.

Changes to this policy

If what ResearchLens collects or how it's used changes materially, this page will be updated and the "Last updated" date above will change.

Contact

Questions about this policy can be raised via the project's GitHub.