feat: ✨ Add context and quota gauges to the status bar.

[Flaykz]

Summary

This PR adds two lightweight usage indicators to the status bar to make session monitoring easier without opening the stats view.

• add a context gauge for the active session based on the latest assistant usage entry found in the session .jsonl
• add a 5h quota gauge using the existing usage data, with a shortcut to the Stats tab
• expose the new session token IPC bridge from Electron main to the renderer
• style both gauges and integrate them into the existing status bar layout
• ignore local JetBrains .idea metadata files

Why

The goal is to surface the most useful runtime limits directly in the main UI:

• current context consumption for the selected session
• current 5h quota usage at a glance
  This reduces friction when deciding whether to continue in the same session or switch context.

Notes

• the context gauge refreshes when a session becomes active and when Claude finishes responding
• the quota gauge refreshes automatically every 5 minutes
• both gauges stay hidden when no data is available

Testing

• UI change only
• not manually tested in this branch

[abasiri]

Merging this — thanks. Two notes on what happens next, so nothing is a surprise.

I pushed a merge commit (a64d468) resolving the only conflict. Your windows-latest → windows-2022 pin was correct and landed separately as #85, where you're credited — you found that break independently, and it turned out to be why every Windows build had been failing. This branch now just takes main's copy of the line.

The context gauge is coming out in a follow-up, for two reasons.

Claude Code already renders the context window in its own statusline (/statusline), and it does it better than we can: it knows its actual limit, whereas we have to guess. Right now that guess is a constant —

const { contextTokens } = result;   // `model` is returned by the IPC handler, then dropped
const CTX_MAX = 200000;

— and on a 1M-context model that reads about 500% and stays pinned red. The handler already returns model, so it's fixable, but duplicating what /statusline shows natively isn't worth the maintenance.

The quota gauge is the good idea here, and I want to extend it. While reviewing this I went digging in the usage API and found the Stats tab was also missing data — /usage shows a per-model weekly bar that Switchboard never rendered:

Current session      1% used
Current week (all)  13% used
Current week (Fable) 8% used   ← this one

The cause: transformUsageResponse asked for bucket keys by name (seven_day_sonnet, seven_day_opus), and those now return null server-side, along with nine codenamed buckets. The live data moved into a self-describing limits array where each row carries its own kind, percent, reset time, and — for model-scoped windows — the model's display name. Fixed on main in 43ab0d0, so the Stats tab renders whatever rows come back and new models appear without a code change.

That matters for your gauge: it reads usage.session, the 5-hour bucket, which is usually the least informative of the three. Mine sits at 1% and resets in a few hours, while the weekly ones are at 13% and 8% — those are what actually constrain you. So the follow-up will point the status-bar gauge at the limits rows instead, most likely showing whichever window is fullest.

Net: your quota gauge is the part that survives and gets better. Appreciate the contribution — the status bar was the right place for this.