Tindahan

Guide

Using your own AI key.

Optional, and only needed if you want the assistant to answer questions your own shop data cannot. If you are happy with the built-in answers, skip this page entirely.

First, do you need a key

Probably not. The built-in assistant answers price, sizes, stock, shipping and returns from the data already in your shop. It costs nothing, needs no key, and never invents an answer it does not have.

A key is for the questions beyond that, the ones a shopper asks in their own words, such as what suits a wedding, or what goes with a linen shirt. That is where a language model earns its place.

What it costs, and who pays

The provider bills you directly on your own account. We never see your key, we never take a cut, and we cannot spend your credit.

Some providers include a free allowance that is enough for a small shop, and others charge per thousand words of conversation. These numbers change, so check your provider's own pricing page rather than trusting a figure written on any guide, including this one.

Set a monthly spend limit at your provider before you paste the key anywhere. That single setting is the difference between a surprise bill and a bad afternoon.

Which provider to choose

The dropdown in Tindahan lists six options. All but the last one are ready to use as soon as you paste a key.

  • Groq - fast and inexpensive for short shop answers
  • DeepSeek - inexpensive, good with plain questions about products
  • Google Gemini - generous starting allowance, needs the model name typed in
  • OpenAI - the best known, usually the most expensive of the five
  • OpenRouter - one key that reaches many models, needs the model name typed in
  • Other (OpenAI-compatible URL) - for a provider not listed, or your own gateway

If you have no strong preference, start with Groq or DeepSeek. Both are cheap enough that a small shop will not notice the bill, and switching later is a two minute job.

Create the key

The steps are the same everywhere, only the words on the buttons differ.

  1. Create an account at the provider's developer site, the one linked below, not their consumer chat product.
  2. Add a payment method or credits, and set a monthly spend limit while you are there.
  3. Find the API keys page, usually under account or developer settings, and click Create new key. On some providers only an owner or developer role can do this.
  4. Copy the key immediately. Several providers show it once and never again.
  5. Paste it somewhere private for a moment, such as a password manager. Never into a chat, an email or a screenshot.

Put the key into Tindahan

  1. In your dashboard, open Tindahan, then Assistant.
  2. Set the mode to Your own AI key.
  3. Choose your provider in Model provider.
  4. Paste the key into AI endpoint key. The field stores it masked: the middle is never shown again on this screen.
  5. If you chose OpenRouter, Google Gemini or Other, type the model name in the model field as well.
  6. Click Save.

There is also an AI relay endpoint field. Leave it empty unless you run your own relay service that returns JSON with an answer field. We do not sell relay credit, so if you are not sure what that sentence means, you do not need it.

Test it

  1. Open your site in a private browser window, as a shopper would.
  2. Ask something your shop data cannot answer, for example what would suit a winter wedding.
  3. If the reply is written in the model's own words rather than a tidy list from your shop data, the key is working.

Ask in another language and the answer comes back in that language, which is the point of using a model at all.

Keep the key safe

  • Never send a key in a chat, an email, a support ticket or a screenshot.
  • Set a spend limit at the provider before the key goes live.
  • The key lives in your site's database, so keep your site updated and keep the number of administrators small.
  • If a key is ever exposed, revoke it at the provider and create a new one. That takes a minute and closes the problem.
  • Use a separate key per site, so one can be killed without touching the others.

Keep the cost down

The assistant looks at your shop data first and only calls the model when that data cannot answer. Most questions therefore cost nothing, even with a key installed.

Set questions per hour per visitor in the Assistant settings. That one number stops a single curious visitor from spending your month. Choose the smallest model that answers well, and check your provider's usage page in the first week so you know your real cost rather than your guessed one.

If it does not work

  • The key is rejected: check that the provider selected in the dropdown matches the account the key came from, that you pasted the whole key with no space at either end, and that billing is active on that account.
  • Answers never come from the model: the shop data answered first, which is the normal path. Ask something outside your shop data to test it, or check the mode is still set to Your own AI key.
  • The bill is higher than expected: lower questions per hour per visitor, lower the spend limit at the provider, and look at the usage page to see which model is being called.
  • You want to stop using AI: switch the mode back to Built-in assistant. The key stays saved but is not called.

Provider links

Links checked on 27 September 2026. Providers move their dashboards from time to time. Wherever it has moved to, the key is always under your account's API or developer settings. If a link stops working, tell us and we will update this page.