Usage 0.22.13

HTML 報告的頂部多了一張「平均每則訊息」的統計卡片。它會把你燒掉的總 token（AI 閱讀與產出文字的計量單位，包含快取重送的 token）數量除以你的總訊息數量，讓你知道自己平均講一句話會消耗多少資源。這個數字反映的是「燒耗量」而不是你打字的長度——如果對話拖得很長，即使你只打幾個字，背後也可能帶著一大包歷史紀錄在燒錢。就像你的車子儀表板本來只有總油耗，現在多了一個「平均每踩一次油門花多少錢」的指標，讓你知道自己的開車習慣有多傷荷包。

The HTML report now features a new "Avg per Message" stats card at the top. It divides your total burnt tokens (the base unit AI uses to measure text, including cached tokens resent in the background) by your total message count, giving you a quick read on how many resources you consume per reply. This number reflects your "burn rate" rather than your literal typing length — if a conversation drags on, even a few typed words could be dragging a massive bag of history behind them, costing you heavily. It's like having a car dashboard that used to only show total fuel consumed, but now includes an "average cost per press of the gas pedal" meter, revealing exactly how expensive your driving habits are.