feat: status bar (model, token count, cost estimate)

[andrewchumchal]

Summary

Add a persistent status bar at the very bottom of the app showing the current model, token count for the active conversation, and estimated cost.

Motivation

Users switching between expensive and cheap models have no at-a-glance signal of what they're running or how much they've spent in a session. The status bar provides ambient awareness without interrupting the flow.

Proposed behaviour

Layout (left → right)

[ 🟣 claude-sonnet-4-5 ]  [ ↑ 1,240 / ↓ 3,810 tokens ]  [ ~$0.014 ]  [ ⚡ Routing: ON ]

• Model indicator — provider colour dot + model name; click opens the model picker
• Token counter — input / output tokens for the active conversation; resets per conversation
• Cost estimate — running cost based on provider pricing; shown as ~$x.xxx
• Routing badge — shows when intelligent routing is active; click opens routing settings

Right side (status items)

• Error count if any IPC errors are present
• MCP tools active count (e.g. 🔧 3 tools)

IPC requirements

Providers need to emit token usage events after each response. Requires small additions to ipc.ts in the client and a new window.api.onTokenUsage channel in preload.ts.

uiStore changes

• No new state — reads from conversationStore (active conversation token totals)

conversationStore changes

• tokenUsage: { input: number; output: number } per conversation (accumulated)

Acceptance criteria

• Status bar renders at the bottom of the app at all times
• Current model name and provider colour shown
• Clicking the model opens the model picker
• Token count shown for the active conversation
• Cost estimate shown (can use hardcoded pricing table initially)
• Token count and cost update in real time as responses stream in
• Routing badge shown when routing is active

Prerequisites

• feat: app shell — activity bar + primary sidebar panel switching #14 App shell (status bar lives in the shell)
• Client ipc.ts / preload.ts token usage events