release: 0.9.0

[pa-arth]

One merged PR since 0.8.0 (#32), and it adds a measurement rather than correcting one — so unlike 0.8.0 there is no number a user has already read that changes meaning here.

MINOR, not patch: new public surface (computeConcurrency, renderConcurrency, concurrencyKey, Session.parentSessionId, a concurrency field on AuditResult), all additive. Aggregate schemaVersion untouched at 10.

What makes it a release

cc-audit could say what sessions cost and how long they ran, and nothing about how many ran at the same time. For anyone running more than one, that single fact reframes the rest of the report — 180h of wall time carrying 487h of agent work is a different product than 487h of work.

The card prints the shape rather than just the average, because the average is the part that misleads: solo is routinely the tallest single bar while being a minority of the time.

╭─ CONCURRENCY  ·  how many ran at once, and what it bought ───────────────────╮
│ 2.7× average — sessions live whenever anything was live                      │
│ 180.7h of your time carried 486.7h of agent work (+305.9h)                   │
│ live     yours    agent  share              /agent-hr                        │
│ 1        58.9h    58.9h  █████████████  33%       6.4                        │
│ 2        41.8h    83.6h  █████████▎     23%       6.9                        │
│ 3-4      51.4h   175.8h  ███████████▍   28%       7.1                        │
│ 5-7      25.7h   142.5h  █████▊         14%       6.6                        │
│ 8+        3.0h    25.9h  ▊               2%       5.1                        │
│ solo 33% of your time · two or more 67%  — the tallest bar, still a minority │
│ bridge 1m/2m/5m/10m → 2.4 / 2.5 / 2.7 / 2.9× (5m used above)                 │
╰──────────────────────────────────────────────────────────────────────────────╯

The fix that predates the feature

Transcript paths were split on / only, while the loader builds them with join() from node:path. On Windows that returns a single segment, so the whole absolute path became the sessionId — shipped in every prior version — and with the new parent link every subagent would have counted as an independent session. Both sites now share one helper.

Verified, not assumed

• npm ci && lint && typecheck clean; 42 test files / 435 tests pass.
• Built the npm artifact and confirmed bundles/npm/package.json reads 0.9.0, so publish.yml's tag-vs-package guard will match v0.9.0.
• Ran the bundle against the real local corpus, not just --help: 2.7× average, 180.7h wall carrying 486.7h agent, peak 13, bridge sweep 2.4/2.5/2.7/2.9 — and the section renders inside its frame at every row, which is the failure this section actually has (three lines overflowed the card the first time it met real data).

Tag v0.9.0 after merge; .github/workflows/publish.yml guards tag == package.json and publishes + attaches the standalone binaries.

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

Updates the package metadata and release notes for version 0.9.0.

• Bumps the root package and lockfile versions from 0.8.0 to 0.9.0.
• Documents the new concurrency measurement, report output, public API additions, and Windows transcript-path fix.

Confidence Score: 5/5

The release metadata changes appear safe to merge, with no actionable defects identified.

The package and lockfile versions are synchronized, build artifacts derive their version from the updated root package metadata, and the documented post-merge v0.9.0 tag will satisfy the publishing guard.

Important Files Changed

Filename	Overview
CHANGELOG.md	Adds comprehensive 0.9.0 release notes describing concurrency reporting, session grouping, and the Windows path fix; no actionable issue identified.
package.json	Bumps the package version to 0.9.0, which is consistently consumed by the build and publishing paths.
package-lock.json	Keeps both root lockfile version fields synchronized with package.json at 0.9.0.

_{Reviews (1): Last reviewed commit: "release: 0.9.0" | Re-trigger Greptile}