feat(vscode): show usage by routed model

[lambertjosh]

Why

The VS Code session header shows aggregate tokens, context pressure, and total cost, but it does not explain which provider route and concrete model consumed that usage. It also derives the token summary from loaded messages, so unloaded history and delegated subagent work can be omitted.

What changed

The Tokens row now uses the complete persisted task-tree totals from the local model-usage endpoint and becomes an accessible disclosure. Expanding it shows usage grouped by configured provider and model, including steps, direct cost, input/output/reasoning tokens, cache reads/writes, and cache rate. Provider headings keep direct subscriptions distinct from Kilo Gateway routing, while model rows use compact catalog display names with routed-ID fallback.

The extension host owns the authenticated SDK request and workspace routing. Webview state is keyed by session and request UUID, retains successful data during refresh, ignores stale responses, refreshes for relevant root/subagent events and reconnects, and falls back to the existing message-derived Tokens summary when the endpoint is unavailable.

Stack

This PR is stacked on #11608 and targets feat/tui-routed-model-usage because it reuses and extends the local session model-usage endpoint introduced there.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 files)

• packages/kilo-vscode/src/KiloProvider.ts

Previous Review Summaries (5 snapshots, latest commit d9a5125)

Current summary above is authoritative. Previous snapshots are kept for context only.

Previous review (commit d9a5125)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 files)

• packages/kilo-vscode/src/KiloProvider.ts

Previous review (commit a08fa5f)

Status: 1 Issues Found | Recommendation: Address before merge

Overview

Severity	Count
CRITICAL	0
WARNING	1
SUGGESTION	0

Fix these issues in Kilo Cloud

Issue Details (click to expand)

WARNING

File	Line	Issue
packages/kilo-vscode/src/KiloProvider.ts	3516	Brand-new sibling or subagent sessions are ignored by the refresh path because their IDs are absent from the previous usage snapshot, leaving aggregate usage totals stale until some already-known family member changes.

Files Reviewed (4 files)

• packages/kilo-vscode/src/KiloProvider.ts - 1 issue
• packages/kilo-vscode/webview-ui/src/components/chat/TaskUsage.tsx - 0 issues
• packages/kilo-vscode/webview-ui/src/context/session.tsx - 0 issues
• packages/kilo-vscode/webview-ui/src/types/messages/extension-messages.ts - 0 issues

Previous review (commit de28740)

Status: 2 Issues Found | Recommendation: Address before merge

Overview

Severity	Count
CRITICAL	0
WARNING	1
SUGGESTION	1

Fix these issues in Kilo Cloud

Issue Details (click to expand)

WARNING

File	Line	Issue
packages/kilo-vscode/src/KiloProvider.ts	1672	Removing the usage-family tracking makes a child viewer stop refreshing when parent or sibling sessions add more model usage, even though the displayed totals still aggregate the whole top-level family.

SUGGESTION

File	Line	Issue
packages/kilo-vscode/webview-ui/src/components/chat/TaskUsage.tsx	107	The new "Hit Rate" label overstates what is measured because the current formula includes cache writes in the denominator.

Files Reviewed (3 files)

• packages/kilo-vscode/src/KiloProvider.ts - 1 issue
• packages/kilo-vscode/webview-ui/src/components/chat/TaskUsage.tsx - 1 issue
• packages/kilo-vscode/webview-ui/src/context/session.tsx - 0 issues

Previous review (commit 980af98)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (3 files)

• .changeset/vscode-model-usage.md
• packages/kilo-docs/public/img/screenshot-tests/kilo-vscode/visual-regression/chat/task-usage-collapsed-chromium-linux.png
• packages/kilo-vscode/webview-ui/src/styles/task-header.css

Previous review (commit dd3aa14)

Status: 1 Issues Found | Recommendation: Address before merge

Overview

Severity	Count
CRITICAL	0
WARNING	0
SUGGESTION	1

Fix these issues in Kilo Cloud

Issue Details (click to expand)

SUGGESTION

File	Line	Issue
.changeset/vscode-model-usage.md	2	The changeset text is VS Code-specific even though this file also publishes notes for @kilocode/cli and @kilocode/sdk.

Files Reviewed (21 files)

• .changeset/vscode-model-usage.md - 1 issue
• packages/kilo-docs/public/img/screenshot-tests/kilo-vscode/visual-regression/chat/task-usage-collapsed-chromium-linux.png - 0 issues
• packages/kilo-docs/public/img/screenshot-tests/kilo-vscode/visual-regression/chat/task-usage-expanded-200-chromium-linux.png - 0 issues
• packages/kilo-docs/public/img/screenshot-tests/kilo-vscode/visual-regression/chat/task-usage-expanded-chromium-linux.png - 0 issues
• packages/kilo-vscode/src/KiloProvider.ts - 0 issues
• packages/kilo-vscode/tests/unit/model-usage.test.ts - 0 issues
• packages/kilo-vscode/webview-ui/src/components/chat/TaskHeader.tsx - 0 issues
• packages/kilo-vscode/webview-ui/src/components/chat/TaskUsage.tsx - 0 issues
• packages/kilo-vscode/webview-ui/src/context/model-usage.ts - 0 issues
• packages/kilo-vscode/webview-ui/src/context/session.tsx - 0 issues
• packages/kilo-vscode/webview-ui/src/stories/StoryProviders.tsx - 0 issues
• packages/kilo-vscode/webview-ui/src/stories/chat.stories.tsx - 0 issues
• packages/kilo-vscode/webview-ui/src/styles/task-header.css - 0 issues
• packages/kilo-vscode/webview-ui/src/types/messages/extension-messages.ts - 0 issues
• packages/kilo-vscode/webview-ui/src/types/messages/sessions.ts - 0 issues
• packages/kilo-vscode/webview-ui/src/types/messages/webview-messages.ts - 0 issues
• packages/opencode/src/kilocode/session/model-usage.ts - 0 issues
• packages/opencode/test/kilocode/cli/cmd/tui/model-usage.test.ts - 0 issues
• packages/opencode/test/kilocode/session-model-usage.test.ts - 0 issues
• packages/sdk/js/src/v2/gen/types.gen.ts - 0 issues
• packages/sdk/openapi.json - 0 issues

---

_{Reviewed by gpt-5.4-20260305 · Input: 87.2K · Output: 4.7K · Cached: 197.1K}

_{Review guidance: REVIEW.md from base branch main}

[lambertjosh]

@alex-alecu - can you review this since you are reviewing the first one? Also, it may be good to add the same model name to the auto efficient responses like you did on the TUI, but that can probably be a different PR.

[lambertjosh]

There is now an expandable section for the existing tokens line:

When it's expanded:

[lambertjosh]

@alex-alecu - this is ready for review

[alex-alecu]

Tested locally and works