All releases

v0.0.10 - Insight Loop

v0.0.10 Latest View on GitHub ->

tokenuse 0.0.10, named Insight Loop, adds the local Insights workflow across the TUI and desktop app, including heuristic signals, manual CLI-generated advice, and local quota sidecars for Claude Code and Copilot.

Highlights

  • Added a new Insights page to the TUI and desktop app for local heuristic recommendations across model right-sizing, cache efficiency, anomalies, and quota pacing.
  • Added manual Generate Advice runs in desktop Insights using the configured Codex, Claude Code, or Gemini CLI.
  • Split TUI Insights into Advice and Signals views, with scrolling, redacted or prompt-snippet advice generation, and background run status.
  • Imported Claude Code and Copilot limit snapshots from local sidecars so usage can show quota context without external telemetry.
  • Added Config actions for Claude and Copilot limit sync, including a Claude Code statusLine sidecar installer.
  • Fixed usage limit rows so Claude Code and Copilot limits attach to their own console sections.

Insights

The new Insights page surfaces local recommendations with a severity, scope, and estimated weekly savings where the engine has a clear basis for that estimate. Each recommendation states its assumption inline, and the rule details live in docs/guides/insights.md.

The engine remains fully local. It does not call a network service or emit telemetry. Cursor and Copilot cache-efficiency rules explicitly silence themselves because their local logs do not expose cache buckets.

Press i from any TUI page to jump to Insights.

Advice Generation

Desktop Insights can now run configured Codex, Claude Code, or Gemini CLI prompts to generate local advice. Prompt templates are user-editable config files, runs are stored in the local archive database, and advice items can be marked done, reopened, or dismissed.

TUI Insights now has separate Advice and Signals views. Advice generation runs in the background in both TUI and desktop, so the interface reports progress and completion instead of freezing during a run. TUI users can press a for redacted data or A to include prompt snippets.

Limits And Status

Usage imports Claude Code limit snapshots from limits/claude-code.json and Copilot quota snapshots from limits/copilot.json.

Config now includes Claude and Copilot limit sync actions. Copilot sync is explicit, confirmed, feature-gated, and writes a local quota sidecar before refreshing archive limits.

Claude sync now guides users to set up the Claude Code statusLine sidecar when it is missing. The Config page can install or remove a wrapper for ~/.claude/settings.json; the wrapper writes Claude Code's status payload to the sidecar and forwards stdin through any previously configured status-line command, such as cship. Settings are backed up to settings.json.bak.<unix-ts> before editing, and a Generate wrapper only path is available for manual setups.

Notes

  • Local usage analysis remains local-only. Network calls are limited to explicit Config-page downloads, local CLI advice generation selected by the user, or maintainer refresh and release paths.
  • The release tag should be v0.0.10 and must match the root Cargo package, desktop Cargo package, desktop package.json, and Tauri config versions.

Full Changelog: https://github.com/russmckendrick/tokenuse/compare/v0.0.9...v0.0.10

Full Changelog: https://github.com/russmckendrick/tokenuse/compare/v0.0.9...v0.0.10