I have the exact same questions. How are tariffs not inflationary ?

Well, they're certainly *stupid*. But I have to admit that there are situations or worldviews in which it's not inflationary. If you work at it.

First, textbook analysis suggest that a country introducing tariffs will see its currency rise. (we may not be in the textbook situation, but bear with me...). This reduces the cost of imports measured in local currency. It is conceivable that the quantities can work out that the two cancel, and end-user prices in the nominal currency of the tariff country don't change. (In fact, this is (perhaps coincidentally, possibly not) approximately what happened in the rounds of US tariffs 8 years ago). Of course, this is in the situation that the currency rose, so the prices have still actually risen in terms of global purchasing power, but it doesn't show up on the local price tags.

Alternatively, it depends on your definition of inflation. Imagine the price of left-handed widgets goes up because the main factory making them just burned down, but no other prices change. Since the price of left handed widgets is included in that country's CPI, the average price across all goods just went up. Is that inflation? To me, no, it isn't, since the most meaningful (to an investor) meaning for the term is "general purpose purchasing power", or monetary inflation: the portion which is not specific to any one product or service...how much of the world's stuff could a unit of currency buy? This is sometimes called "common inflation", and the most recent method for measuring it is the Fed's "MCT Inflation" metric, which has incidentally been rising a tiny bit for a few months now.
So...if a country sees a few specific goods rise in price (the imported ones), but the currency stays flat and still buys as much globally traded "stuff" as it always did, I would argue that in the most meaningful sense that there has been no monetary inflation. Just some idiosyncratic price changes.

FWIW, the MCT inflation page. newyorkfed.org: Multivariate Core Trend Inflation
Click on the graph and you get a detailed version that lets you see all the numbers. The most recent one, for March, is the first reading over 3% in a while, up from the lows under 2.5% last June. To oversimplify, this is the reasoning: imagine that the one year inflation rate of every product category is the sum of a single global "monetary" inflation number, plus a price change specific to that product or service. So, for 50 categories, you get 50 category-specific numbers and 1 global number for any given one-year interval. These can be estimated by doing a regression across the whole time series. The MCT inflation figure is the common one: the amount of overall inflation common across categories, ignoring product- or service-specific shocks.
One interesting thing is that MCT inflation never went negative during the pandemic.

Jim