[ PRIVACY ]
Your request, handled with care.
Here is what happens to the information you share when you ask us about a software project.
Development preview. The effective date is set at the production release.
Who is responsible
PUNI is operated by FOP Fedorov Danylo Dmytrovych (ФОП Федоров Данило Дмитрович), an individual entrepreneur registered in Ukraine. Danylo Fedorov is the controller of the personal data described here. Contact dany@puni.dev about your information, access, corrections or deletion. You do not need to start a sales request to write to us.
Which law applies
We process personal data under the Law of Ukraine On Personal Data Protection. When you contact us from the European Economic Area or the United Kingdom, the GDPR or the UK GDPR also applies, and this page states the legal basis for each use. Where this page and the law differ, the law wins.
What we collect and why
We save what you give us and what we need to run the service.
Your request. The description you enter, selected choices and explanations, optional references, draft and suggested briefs, and the final brief and email address you explicitly submit. We use these to understand your project, prepare your request and respond to it. Legal basis: steps at your request before a possible contract (GDPR Article 6(1)(b)).
AI-guided questions. The chat messages you send and the AI replies you receive, kept with your request. Legal basis: the same pre-contract steps; you choose whether to start the chat.
Operational records. A daily keyed source code derived from your network address, request counts, costs and security events. We use them to run the service and limit abuse. Legal basis: our legitimate interest in keeping the service available and affordable (Article 6(1)(f)).
Analytics, where enabled and as described below. Legal basis: consent for analytics cookies where required (Article 6(1)(a)); legitimate interest for the regional cookie default elsewhere and cookieless measurement, which you can stop at any time.
References are text you choose to share; we do not fetch their links. Do not include passwords, credentials or confidential access links. You can review and edit the final request before submitting it. Saving a request does not confirm email delivery or promise a response deadline.
Who and what reads it
Danylo reads submitted requests. So do machines. PUNI works with AI assistants and agents throughout its work, so your description, choices, references, brief, chat and contact details can be read, summarised, planned around and acted on by AI systems we run, on the manual route as well as after AI-guided questions. Those systems use third-party AI providers, currently Anthropic, OpenAI and Google, so your text can reach those providers while we work on your request. What they retain follows their terms, which we do not control, and we cannot recall text that has reached them. We do not train models on your request. Working notes and plans derived from your request live in our own tools on the same terms.
Decisions about your request, whether we take it on, what we propose and what it costs, are made by Danylo. The AI systems inform and draft; a person is answerable for what we send you.
AI-guided questions
Starting AI-guided questions sends your description, choices and conversation context to OpenRouter, Inc. (United States), which routes the request, and Microsoft Azure OpenAI Service, which runs the model. References and brief text can be included as context. The email entered in the contact form is not sent as model context; anything you type into chat itself can be.
The configured route requests the Azure EU endpoint, zero data retention and no provider data collection, without fallback to another provider. An EU endpoint label does not prove all processing stays in the EU. These routing requirements do not verify account-level logging or training settings, and they do not mean PUNI stores no conversation.
You can use the manual request route without starting AI chat; that keeps your text away from the chat processors named here, not from the AI systems described under Who and what reads it. Switching to manual review after chatting does not undo information already sent to AI providers. When AI is unavailable, manual review remains available. AI replies are not prices, delivery dates or contracts.
Where it is kept
Your request lives in a database on servers rented from Hetzner Online GmbH in Finland. Copies of that database are written every hour to Hetzner object storage in the European Union so that we can restore it after a failure. Danylo administers these systems from Ukraine. When an abuse limit trips, an alert goes to the notification service ntfy.sh; alerts carry only counts and amounts, never your text, email or address. Replies to you are sent from dany@puni.dev, a Gmail mailbox, so Google processes that correspondence.
Transfers outside the EU
If you write to us from the EEA or the UK, your information leaves that area in three ways: Danylo accesses it from Ukraine; the AI chat processors and the AI systems we work with are operated from the United States; Google processes analytics data and our email in the United States. For Ukraine we rely on the fact that you asked us for the service (GDPR Article 49(1)(b)) and on this notice. For the US providers we rely on the data protection terms they publish, including EU standard contractual clauses where they offer them; we have not audited those terms independently.
Necessary browser storage
Functional cookies hold a browser claim for 24 hours and protect requests against forgery. A separate necessary abuse-prevention cookie lasts up to 31 days and connects usage across new drafts in the same browser. It is not a hardware device identifier. Necessary browser recovery storage preserves a pending submission so a lost response can be checked without silently submitting again. These are needed for the request journey and are separate from optional analytics consent. No visitor account is needed. The site loads fonts and media from its own domain only.
Operational records
We process network and request information to run the service and limit abuse. Daily keyed source codes are pseudonymous abuse identifiers: your network address is combined with a key that changes every day, and only the resulting code is stored. Usage, costs and security records are distinct from conversation text. We do not claim that these records are anonymous or impossible to link. Our web servers keep no access logs of page views.
Approximate country for analytics preferences
Our server estimates your country from your network address using a local copy of DB-IP Country Lite, November 2026, used unchanged under CC BY 4.0. There is no per-visitor request to DB-IP. This feature does not add storage of visitor addresses or countries. Country is approximate, including VPN exits, and does not establish residence or certify every country's legal requirements.
In the EEA, UK, Switzerland and corresponding EU territories, and when country is unknown or lookup fails, analytics cookies need explicit permission. In known other countries, analytics cookies are allowed automatically unless you have saved a choice. Your saved acceptance or rejection always wins. Privacy choices stays available everywhere to allow analytics cookies or disable measurement.
Optional Google Analytics
Where explicitly enabled, basic cookieless Google Analytics measurement starts on arrival with analytics storage denied. It sends approved page, service, article and journey events, including fixed starting categories and limited journey metadata. It processes device, browser and network information, including network addresses. Cookieless does not mean no collection or guaranteed anonymity. Allow analytics cookies permits pseudonymous browser and session identifiers in analytics cookies for future measurement. We do not send your description, raw answer text, selected suggested-answer text, free explanations, chat, references, brief, email or request identifiers to analytics. No detailed pre-consent activity history is recorded or replayed after cookie permission. Google account retention and logging settings are not independently verified here.
Disable measurement in Privacy choices stops further Google measurement and removes this environment's analytics cookies and permitted lead markers. A saved rejection prevents Google from loading. Functional claim, forgery protection and request recovery storage remain separate, and requests and human review remain usable. Development and production have separate preferences. The standard local demo and builds keep analytics disabled; production measurement remains held and requires its own configured account. Local collector acknowledgements do not prove Google account ingestion or processed reports.
How long we keep it
Anonymous draft access expires after 24 hours. Access expiry does not prove that stored content or backups have been deleted. Our policy is to keep submitted request content and contact details for 12 months unless the request becomes a client project or a legal or security hold applies. Project records are agreed separately.
The current retention process is manual and report-only: reports identify records due for review, and a person handles deletion requests. Automatic erasure is not activated. Backup removal and aging have not been verified; we do not promise that deleted text has already left every backup. When content is erased, the date, cost and abuse-limit code of each AI reply remain as accounting and security facts, without message text. Anything an AI provider received while we worked on your request is kept under that provider's terms, not ours.
Security
Connections to the site, the app and the API use TLS. Request data is stored on servers that only Danylo administers, in directories readable only by the service. Cookies that grant access to your request or the operator inbox are HttpOnly and restricted to the host that issued them. Preference and analytics cookies are managed by browser scripts. Backups are stored with versioning in object storage. No system is perfectly secure; if we learn of a breach affecting your data, we will tell you and the competent authority as the law requires.
Your rights
You can ask for access to the personal data we hold about you, for corrections, for deletion, for a restriction of processing, for a copy in a portable format, and you can object to processing based on our legitimate interest. Where processing rests on consent, you can withdraw it at any time; Disable measurement in Privacy choices does this for analytics. Email dany@puni.dev with enough context to identify your request, without sending passwords. A person handles these requests.
You can complain to a supervisory authority: in Ukraine, the Ukrainian Parliament Commissioner for Human Rights; in the EEA or the UK, the data protection authority of your country.
Automated decisions
No decision with legal or similarly significant effect on you is made by automated means. AI replies and AI-assisted drafts inform Danylo; they do not decide whether we work with you or on what terms.
Children
This service is for businesses and adults. We do not knowingly collect personal data from anyone under 16. If you think a child has sent us a request, email dany@puni.dev and we will delete it.
Changes
We update this page when our practice changes and show the effective date at the top. Material changes to how we use a request you already sent will be announced to the email address you gave.

