Privacy

This site is run by Rainbow After Storm Inc., a one-person AI/LLM engineering practice based in Toronto, Canada. This note explains, in plain language, what the site collects and who helps process it. It is written to match what the site actually does, nothing more. It reflects the principles of Canada's Personal Information Protection and Electronic Documents Act (PIPEDA); it isn't legal advice.

What I collect, and why

This site only takes in your information when you choose to give it — by writing to me, or by using the assistant:

• The contact form. When you send a message from the contact page or the FinMRI page, I receive the name, email address, and message you enter, plus a note of which page you sent it from. It's used for one thing: to read and reply to what you wrote. It's stored in a Postgres database (hosted by Neon) so the message isn't lost before I get to it.
• The site assistant. If you use the chat assistant, the question you type is sent to Google's Gemini API to generate an answer from this site's own content. I also save the conversation to my database: your questions, the answers, which pages the answer was drawn from, and timing details like how long the answer took. I read these to see what this site is failing to explain, and to reply properly if you go on to leave your details.
• What Google may do with an assistant question. The assistant currently runs on Gemini's free tier, and Google's terms for that tier let them use what is sent to it to improve their products, including for machine learning, and allow their reviewers to read it. That is Google's policy rather than mine, and it is the honest reason not to type anything confidential into a chat box on a stranger's website — this one included. If you want to tell me something private, the contact form goes to my database and my inbox, and nowhere near a model.
• What the assistant records alongside a conversation. So that the turns of one conversation hang together, your browser generates a random identifier and keeps it for that visit. Stored with it: the page you started the conversation on, the site that referred you if there was one, and your country. That is all. No IP address, no browser or device fingerprint, nothing finer-grained than the country, and nothing that would identify you across other visits or other sites. Your IP address is used for a moment to rate-limit abuse and is never written down. Please don't type anything personal into the assistant; if you want to tell me something about yourself, use the contact form, which is meant for it.
• Leaving your details from the assistant. If the assistant can't answer, it offers a short form. If you fill it in, I keep the conversation identifier with your message, so that when I reply I can see what you were actually trying to find out.

Cookies, browser storage, and analytics

This site sets no cookies. The one thing it does keep in your browser is the assistant's conversation identifier and the transcript you can see in the panel, held in session storage — and only from the moment you first open the assistant. Your browser erases it when you close the tab, it is not sent to any other site, and if you never open the assistant nothing is stored at all.

The site does count page views, using Vercel Web Analytics, so I can see which pages people actually read. That measurement is aggregate: it records the page, the referrer and coarse details like country and device type, and it does not set a cookie, does not assign you an identifier, and cannot follow you to any other site. The script is served from this domain rather than fetched from a third-party host.

There is still no advertising, no pixels, and no third-party profiling. Web fonts are served from this site itself rather than fetched from Google. If any of this changes, this note will be updated to say so.

Who else processes it

Running the site relies on a few service providers. Your information may pass through or be stored by them, including on servers in the United States:

• Vercel: hosting; handles standard server request logs, and the aggregate page-view counts described above.
• Neon: the Postgres database that stores contact messages and assistant conversations.
• Google: the Gemini API that powers the assistant, for the questions you send it.
• ImprovMX: forwards email sent to the address on the contact page to my private mailbox.

FinMRI, the live demo at agents.rainbowafterstorm.com, is a separate application with its own data handling. This note covers only this website.

How long it's kept, and your choices

I keep contact messages only as long as they're useful for replying to you and for any working relationship that follows. Assistant conversations are deleted after 12 months.

You can ask me at any time to see what I hold about you, correct it, or delete it — including a conversation with the assistant. Just email hello@rainbowafterstorm.com and I'll take care of it. For a conversation, it helps to say roughly when it was and what you asked about, since there is nothing on my side that links it to your name.

Children

This is a professional services site and isn't directed at children.

Changes

If this note changes, the “last updated” date above will change with it. Questions about any of this are welcome at hello@rainbowafterstorm.com.