[saioai]

Project rate-limit commands currently require list-rate-limits and update-rate-limit.
This adds list and update, with clearer rate units and project-specific empty results.

What changes

• Add friendly verbs across project command routes while preserving both existing names.
• Label returned limits as requests/minute, tokens/minute, requests/day, images/minute, audio megabytes/minute, and batch input tokens/day.
• Keep model names, IDs, exact values, nulls, and unfamiliar fields visible.
• Explain that batch input tokens/day differs from general tokens/day.
• Add no general daily-token or rate-limit reset flags. Existing model and hosted-tool permission commands remain unchanged.

Readable labels also change in pipes. Explicit machine formats retain API field names and values.

Commands

New convenience names:

openai admin projects rate-limits list --project-id proj_example
openai admin projects rate-limits update --project-id proj_example --rate-limit-id rl_example --max-tokens-per-1-minute 1000

Use a rate-limit ID returned by list, and only apply an approved project change.
The colon route, admin organization projects, and projects shortcut support both verbs.
Existing list-rate-limits and update-rate-limit remain available.
Pagination flags, JSON/JSONL/YAML/raw formats, field extraction, and request overrides retain their behavior.

update --help lists supported fields. Omitted fields remain unchanged; zero and null are not reset shortcuts.

Code

pkg/custom adds command decoration before route cloning and supplies the scoped empty-result message.
pkg/transformers changes readable field labels while preserving values and unfamiliar data.
Generated request handlers, dependencies, and startup wiring remain unchanged.
The synthetic recorder reuses scripts/demos/capture_and_render.sh.

Tested

On macOS arm64, 196 focused test/subtest results passed without skips. Focused race checks, compilation, and module verification passed.
An independent observer passed 123 CLI processes, including machine-output comparisons, combined output/API failures, and copied Bash/zsh help.
Copied scripts/run examples passed twice. Local runtime checks used c900d05; later changes affect only the recorder.
The final build matches the tested binary byte-for-byte.

Required CI, CodeQL, and native help checks passed at fbcae067. The trusted budget check passed at 21/1,000 custom lines.
The local budget reporter stalled; its CI counterpart completed successfully.
Native help covered macOS Bash/zsh, Ubuntu Bash, and Windows cmd/PowerShell/pwsh, with documented platform skips.
F32-specific native Windows workflows, graphical-terminal appearance, and live Admin acceptance remain unverified.
Public daily-token and rate-limit reset contracts remain separate API decisions.

Demo

Real CLI binaries with matching synthetic local responses, recorded as macOS Bash terminal replays.
Baseline: 5739ce0. Runtime: c900d05; later commits only correct the recorder.
Dark and light 100×30, plus narrow 40×36 captures, passed assertions and independent visual inspection.
The demo shows the shorter list command, project-specific empty results, and readable rate units.

Before:

After:

Recording recipe

[github-actions]

Castiron custom code

Evaluated main: 5739ce0f3a6cc5887092554a403bf53357eca9a4.

✅ No new custom-code files detected.

2 mixed files remain; 0 existing customizations changed.

Compared 5739ce0f3a6c → fbcae06752ca. Generated baselines verified.

2 existing customizations unchanged

• pkg/cmd/adminorganizationcertificate.go
• pkg/cmd/audiovoice.go

A changed generated baseline means this report cannot reliably identify which handwritten lines changed.

Inspect the custom-code diff

Download the exact patch produced by this run (requires repository access):

gh run download 38035012219 --repo openai/openai-cli \
  --name castiron-custom-code-38035012219-1 --dir /tmp/castiron-custom-code-38035012219-1
git apply --stat /tmp/castiron-custom-code-38035012219-1/custom-code.patch
cat /tmp/castiron-custom-code-38035012219-1/custom-code.patch

Or reproduce it from an SDK checkout containing the vendored reporter:

git fetch --no-tags origin 5739ce0f3a6cc5887092554a403bf53357eca9a4 fbcae06752ca45bbe1fb8d9fab1b50bfe96c9ca3
python3 scripts/castiron/custom_code_report.py report \
  --base 5739ce0f3a6cc5887092554a403bf53357eca9a4 \
  --head fbcae06752ca45bbe1fb8d9fab1b50bfe96c9ca3 --fetch --require-head-hash --public \
  --out /tmp/castiron-custom-code-fbcae06752ca
cat /tmp/castiron-custom-code-fbcae06752ca/custom-code.patch

This is the current full custom patch for mixed files, not an attribution of only the handwritten lines changed by this PR.

Full report and patch