Repository navigation
Perf: sidebar search lowercases every message of every thread per keystroke #144
Description
Problem
electron/store.js:1021-1057 (searchThreads) does String(m.text).toLowerCase().includes(needle) across all threads' full message lists, with no index and no early cap on scan cost. The renderer fires it after a 250ms debounce (src/components/Sidebar.tsx:52,1203-1205).
Impact
Each keystroke burst allocates a lowercased copy of the entire store's message text on the main thread. With capped-but-large transcripts (1,000 messages x N threads) that is MBs of allocation and a full scan per query: a GC and latency spike while the user is typing.
Fix direction
Cache lowercased text per message (invalidate on append), or build a lightweight inverted index; also consider a max-results early exit.
Activity
currentbits commentedon Aug 21, 2026
Stale as of the sidebar redesign: search now matches only thread.title (src/components/Sidebar.tsx:1758,
thread.title.toLowerCase().includes(queryLower)) — the per-keystroke scan over every message of every thread no longer exists. Verified by grep in perf research round 31b (2026-08-21): no message-body search remains in the renderer. Transcript search is now the separate feature request #486.