See your real ROI —
cost-per-call & weekly trend, now in your IDE

The problem with a generic “$” number

Every cache hit and every recalled lesson is a model call you didn’t have to pay for. cachly has always estimated those savings — but the estimate leaned on a generic default price per call. If your team runs Opus or GPT-5.5, that default understates your savings by an order of magnitude. If you run a small model, it overstates them.

Either way, the number stops being believable. And an ROI number you don’t believe is worse than no number at all.

Cost per avoided call — your real price

You can now set a per-call price on each instance (Instance page → Cost per Call). Pick a model preset or type your own number. Every savings and ROI figure then multiplies cache hits by that price instead of the default.

The ROI Summary in the plugins now shows a Cost per avoided callrow, so it’s always clear which price the dollar figure is built on — your configured rate, or the default until you set one.

Week-over-week — is your Brain getting more useful?

A running total tells you the Brain has value. It doesn’t tell you whether that value is growing. The new week-over-week row compares this week’s Brain activity to the previous seven days and shows the direction with a ▲ / ▼ and a percentage.

It’s computed entirely from data the insights endpoint already returns — no new tracking, no extra calls. Until you have two full weeks of history, it simply shows this week’s activity with a “no baseline yet” note instead of a misleading percentage.

Where to find it

• VS Code (0.9.8+) — open the Brain Health panel; both rows appear under ROI Summary.
• IntelliJ / JetBrains (0.3.3+) — Tools → Show Brain Health, same ROI Summary section.

Both are part of the same idea we keep coming back to: the Semantic Cache and the Brain only earn trust when their value is visible and honest. Real prices, real trends — no marketing math.