← Back to The Analogy Architect

Privacy Policy

Last updated: September 2026

The Analogy Architect is a small toolkit of simple web apps — Analogize and Ghar Ka GPT, with more on the way. The two apps handle your data very differently, so here is exactly what happens, app by app, followed by what applies to the whole site.

Analogize

What's public: the "Recently decoded" wall

Analogize has a shared, public wall. When you generate an analogy, the topic you typed and the analogy you got — along with the region, language and script you picked — can appear in the "Recently decoded" section that every visitor sees. The wall shows the latest 10 entries. It is not linked to your name, an account, or your IP address.

To keep personal details off the wall, only short, concept-style topics are published: up to 60 characters and six words, and never a topic containing a question mark, an @ sign, a web address, or a long number. Anything else still gets you an analogy — it just isn't published. Even so, please don't type personal information into Analogize.

We also keep one site-wide number: how many analogies have been generated in total.

What we send to our servers

When you generate an analogy, your browser sends your topic and the region, language and script you picked to our API. Your IP address reaches the server too, as with any web request; we use it only for rate limiting — to prevent abuse — held in memory and discarded when the server instance restarts. Our code doesn't write it to a database or a log.

The only things we store on our side are the public wall and the total count described above.

Third-party AI services

Your topic and settings are sent, from our server, to Google's Vertex AI (Gemini, or Anthropic's Claude hosted on Vertex AI) and/or Anthropic's Claude API to generate the analogy. Your IP address is not passed on to them. Please review Google's privacy policy and Anthropic's privacy policy for how they handle API inputs.

Location detection

If you haven't picked a region yet (and the link you opened doesn't name one), the app asks our server once per visit for your approximate city, to pre-select the regional flavour that fits you best (e.g., Mumbai, Bengaluru, Chennai). The city and country come from Cloudflare's built-in IP geolocation; nothing about the lookup is stored on our side. If your city matches one of our regions, that region is saved in your browser and the lookup doesn't happen again. If it doesn't match (for example, you're outside India), the lookup may repeat on later visits.

What's saved on your device

Analogize keeps a few small items in your browser's localStorage — on your device, never sent to us:

You can clear all of this at any time by clearing your browser's site data for this domain (Browser Settings → Privacy → Clear site data).

Ghar Ka GPT

A different privacy model entirely

Ghar Ka GPT works nothing like Analogize, and it's worth explaining exactly how, since the whole point of the app is that it doesn't need our servers.

Your conversation never reaches a server

The AI model runs entirely inside your browser tab using WebGPU — your own device's graphics hardware does the thinking, not ours. We do not see, receive, log, or store anything you type into Ghar Ka GPT. There is no backend endpoint for chat messages to be sent to. You can verify this yourself: open your browser's network inspector while chatting, and you'll see zero requests during a reply.

What gets downloaded, and from where

The first time you "hire a cook," your browser downloads that model's weights (roughly 300 MB to about 2 GB, depending on which model you pick) directly from Hugging Face's CDN, plus a small WebAssembly runtime from GitHub. These are one-way downloads — files arriving in your browser, not data leaving it. Your browser caches them afterward, so later visits load from your device instead of re-downloading. Review Hugging Face's privacy policy for how they handle file-hosting requests.

What's saved on your device

Ghar Ka GPT saves your conversation and your settings (which cook, temperature, max tokens, system prompt — plus which cook last loaded successfully, and whether you dismissed the save-for-offline tip) to your browser's localStorage, purely so refreshing the page doesn't wipe your chat and your cook is ready next time. This never leaves your device and is never visible to us. Clear it anytime with the "Clear conversation" button in the app, or by clearing your browser's site data for this domain. Downloaded cooks stay in your browser's storage until you clear site data.

We built Ghar Ka GPT so you'd have somewhere to ask things you wouldn't want to type into someone else's server. That promise only means something if we're precise about what does and doesn't happen — and nothing happens beyond what's described here.

The whole site

Hosting

The site runs on Cloudflare, which delivers every page and API request and protects the site from abuse. Like any host, Cloudflare processes your IP address and basic request details to do that, and it may set strictly necessary security cookies (for example, after a bot check). See Cloudflare's privacy policy.

Fonts

Our pages load their typefaces from Google Fonts, so your browser requests font files from Google's servers, which receive your IP address and browser details in the process. See Google's privacy policy.

Offline support

A small service worker saves copies of our pages on your device, so the site (and Ghar Ka GPT, once you've used it) can open without a connection. These copies live only in your browser.

No ads, no analytics, no trackers

We don't run advertising, analytics platforms, tracking pixels, or cookies of our own.

Children

This service is not directed at children under 13. We do not knowingly collect any information from children.

Changes

If we make material changes to this policy, we'll update the "last updated" date above.

Contact

Questions? Reach out at hello@theanalogyarchitect.com.


© 2026 The Analogy Architect.