phuryn/claude-usage v1.5.4 on GitHub

v1.5.4 — 2026-07-01

Added a Title column to the Recent Sessions table (after Project) and its CSV export, showing Claude Code's own session title: a user-set custom-title takes priority over an AI-generated ai-title. Long titles wrap within the column (min width 160px). Sessions without a title record show a muted Untitled placeholder — there is no fallback to the first user message, so prompt text never leaks into the table (#147, thanks @arojunior).
The Recent Sessions CSV export now includes full session IDs (the table still shows the 8-char prefix for readability, but an 8-char prefix isn't useful in an export) (#147).

The embedded dashboard now waits up to 20s (was 10s) for the server to become ready on cold start — the one-time topic backfill can slow the first launch — and a failed start now offers a Retry button (both in the error notification and on the panel) instead of only pointing at the command palette (#147).

The scanner now parses custom-title / ai-title transcript records into a new sessions.topic column (additive in-place migration; existing DBs upgrade without a rebuild). Titles are captured even when Claude Code appends them after the turns (picked up on the next incremental scan), and a title-only record can never create a token-less phantom session row (#147, thanks @arojunior).
On upgrading an existing database, the next scan runs a one-time topic backfill: it re-reads the title records already present in previously-scanned transcripts (which an incremental scan would otherwise skip) so old sessions get a Topic too, then records via a schema_meta flag that it's done so it never repeats. Only title records are read, so token totals are untouched (#147).