Two filter modes, chosen by the shape of what you type

A plain word or phrase does a straightforward case-insensitive substring match, which covers the common case — filtering a log down to lines mentioning "timeout" or "error" without worrying about exact casing. Wrapping the filter in slashes, like `/error|fail/i`, switches into full regular expression mode instead, letting you match alternation, anchors, and character classes for cases a plain substring search can't express — like matching either of two words, or a pattern that only counts at the start of a line. Detecting which mode to use from the input itself, rather than a separate toggle, keeps the common case (plain text) fast to reach for while still making the more powerful mode available the moment you need it.

Why filtering runs on every keystroke

Re-filtering live as you type turns log triage into a fast, iterative process — start broad, watch the match count, narrow the pattern, watch it narrow further — rather than a type-then-click-then-wait loop that breaks the flow of actually reading through a log. For most logs (up to a few thousand lines) this is imperceptibly fast; it only becomes noticeable on very large pasted logs, where the practical fix is trimming to the relevant time window or section first, both because it's faster and because a smaller, pre-narrowed log is usually what you actually want to be scanning anyway.