[grdsdev]

Summary

supabase gen types --lang swift gives the Swift driver from @supabase/typegen a filtered environment. The driver reads SUPABASE_TYPEGEN, XDG_CACHE_HOME and HOME from that environment to find the supabase-typegen binary and its release cache. The generator layer passed only PATH, PATHEXT and ComSpec, so:

• SUPABASE_TYPEGEN had no effect, but the CLI's own error message tells users to set it.
• The cache path became relative (Library/Caches/supabase/typegen/<version>/ on macOS, .cache/... on Linux). It resolved against the working directory, so the downloaded 23 MB binary landed inside the user's project, and ~/Library/Caches was never read.

This PR adds the three variables to lookupEnv. SIDE_EFFECTS.md already documents them. It also lists SUPABASE_TYPEGEN in CLI_NON_CONFIG_ENV_NAMES, so the registry test accounts for it. Other languages do not change: the spawned tool already inherits the full environment, and the Dart driver reads nothing from this object.

How this fits the supabase-swift v3 alpha

For the alpha, the Swift driver will stop downloading the prebuilt binary. It will build supabase-typegen from source, at the supabase-swift commit pinned in the app's Package.resolved, and cache it by commit (SDK-3043, a separate change in @supabase/typegen). The resolution order becomes SUPABASE_TYPEGEN, then the source build, then PATH.

This PR is still needed for that design, and it lands first:

• SUPABASE_TYPEGEN stays the first step.
• The source build caches under XDG_CACHE_HOME or HOME. Without this fix, the cache path is relative, and the build output would land inside the user's project.

The download path that this PR also fixes stays in CLI 3.1.0 until the driver change ships.

Linked issue

Linear: CLI-3134

• The linked issue is open and carries the open-for-contribution label (or I'm a Supabase maintainer).

How it was tested

• New integration tests in types.generator.integration.test.ts, with a fake supabase-typegen. They cover three cases: SUPABASE_TYPEGEN, a cached release binary under XDG_CACHE_HOME, and one under HOME. All three failed before the fix and pass after it. The two cache tests skip on platforms that have no Swift release asset.
• test:integration and test:unit for src/commands/gen/types/ pass (176 and 34 tests).
• End to end, with a CLI binary built from this branch (bun scripts/build-binary.ts), a Postgres 17 container, and a real supabase-typegen from supabase-swift:
  • With SUPABASE_TYPEGEN=<binary> and nothing on PATH, CLI 3.1.0 fails with GenTypesToolNotInstalledError. This branch generates the Swift file.
  • In a project whose Package.resolved pins supabase-swift 3.0.0-alpha.1, with the binary in ~/Library/Caches/supabase/typegen/3.0.0-alpha.1/: CLI 3.1.0 does not find it, and this branch uses it. No files are written to the project.
  • Not covered: a real download. The first release with the asset (v3.0.0-alpha.1) is not published yet. The download writes to the same path the cache test reads. After SDK-3043, the alpha does not download at all.

Checklist

• The PR title follows Conventional Commits (e.g. fix(cli): …).
• Tests added or updated for the change.
• From the repository root, pnpm check:all passes; relevant package tests pass for every touched workspace, and pnpm types:check passes for each touched TypeScript workspace (or workspace declaring it). Locally, types:check, lint:check, fmt:check and knip:check pass. lint:actions:check did not run, because jactionlint is not installed, and this PR changes no workflow.

Why now

supabase-swift v3 launches as Public Alpha on 2026-12-11. Its demo flow is supabase gen types --lang swift. A CLI release with this fix and the SDK-3043 driver change must ship before that date.

[github-actions]

🤖 AI Review

Both independent reviews reported no findings. Reading the changed code confirmed that the three environment variables reach the typegen host, matching the existing documented Swift behavior. The new tests cover the explicit binary override and both cache locations. No conflict with pending next-branch work was identified. Tests were not run because dependencies are absent from the checkout.

Findings

No issues found.

Stats

Claude findings: 0 · Codex findings: 0 · Confirmed: 0 · Refuted: 0 · Uncertain: 0

---

Models: claude-opus-5-5 + gpt-6.1-sol · Trigger: auto · Workflow run

This review runs once per PR. A maintainer can request another with a /ai-review comment.

[github-actions]

CLI benchmark (Lightweight Commands): no changes found

Commit cb8d5f8 against develop at d8f66a0 · 10 runs each · report only, never blocks merging

Command	Workload	Before	After	Change
supabase --version	version	309 ms	315 ms	no real change
supabase --help	help	325 ms	303 ms	no real change (-7%, within normal variation)
supabase init	init	300 ms	325 ms	no real change (+8%, within normal variation)
supabase migration new bench	migration-new	297 ms	328 ms	no real change (+10%, within normal variation)
supabase config convert	config-convert	278 ms	277 ms	no real change
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	251 ms	268 ms	no real change (+6%, within normal variation)
supabase projects list	projects-list	230 ms	228 ms	no real change
supabase functions list --project-ref replayfixtureprojref	functions-list	258 ms	270 ms	no real change

API calls: unchanged for all 8 commands.

Binary size: 134.8 MB, unchanged.

Details: timing ranges, API calls per command, slowest phases

Timing ranges

Command	Workload	Before range	After range	Change	Verdict
supabase --version	version	237-336 ms	240-369 ms	+6 ms (+2%)	within noise
supabase --help	help	280-407 ms	266-356 ms	-22 ms (-7%)	within noise
supabase init	init	272-398 ms	270-420 ms	+25 ms (+8%)	within noise
supabase migration new bench	migration-new	263-353 ms	267-392 ms	+31 ms (+10%)	within noise
supabase config convert	config-convert	226-323 ms	209-310 ms	-1 ms (0%)	within noise
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	227-333 ms	222-356 ms	+16 ms (+6%)	within noise
supabase projects list	projects-list	185-260 ms	214-273 ms	-2 ms (-1%)	within noise
supabase functions list --project-ref replayfixtureprojref	functions-list	239-295 ms	217-289 ms	+12 ms (+5%)	within noise

API calls per command

Command	Workload	Before	After	Endpoint changes
supabase --version	version	0	0
supabase --help	help	0	0
supabase init	init	0	0
supabase migration new bench	migration-new	0	0
supabase config convert	config-convert	0	0
supabase gen types --lang typescript --project-id replayfixtureprojref	gen-types	2	2
supabase projects list	projects-list	1	1
supabase functions list --project-ref replayfixtureprojref	functions-list	2	2

Slowest phases

No command moved beyond noise, so there are no phase deltas.

Workflow run

Comment /bench to run the lightweight benchmark (this comment) again on the latest commit, or /bench full to also run the integration-heavy benchmark, which posts in its own comment.