spend-limit: name the organization in command output

[markbackman]

Summary

pipecat cloud spend-limit never said which organization the numbers described. The org appeared only in the transient spinner line, which disappears, so both the rendered table and the JSON payload were unattributed. That matters as soon as -o/--organization is used or more than one org is in play.

Changes

• _render_show takes the org and renders an Organization row as the first row, so show, set and clear all attribute their output. The parameter is required, so the signature enforces the invariant.
• --output json (and the deprecated --json alias) emits an organization key. The API response does not carry one, so the CLI fills it from the resolved org via a new _with_org helper. It uses setdefault, so a server-supplied organization wins if the field ever appears.
• The two confirmation prompts on set name the org. These are the invocations that block sessions, so the target is worth pinning before you hit enter.
• The "no spend-limit data available" message names the org instead of saying "this organization".

Note for reviewers

Adding organization to the JSON output is an additive change to a machine-readable shape. Anything doing an exact-dict comparison on spend-limit show --json will see the new key.

The empty payload is deliberately left alone. spend_limit_get passes not_found_is_empty=True, so an org with no spend-limit record yields (None, None) and show --output json emits {}. _with_org only decorates a real payload, so that stays falsy and if not payload: keeps working. Covered by a test.

Test plan

• New tests: the org row renders; a server-supplied organization is not overwritten; the 404 path still emits {}; set and clear attribute their JSON.
• Updated the existing tests that asserted the JSON payload matched the API response exactly.
• Full suite passes (446). ruff format --diff and ruff check clean.
• Driven manually against a local stub API in all three output modes, plus set and clear.
• pyright is not clean on this repo and is not in CI: 90 errors on main, 92 here. The two additions are reportUnusedCoroutine on the new tests calling the @synchronizer.create_blocking wrapper without await, matching the eight pre-existing occurrences in that same file. (An earlier version of this description claimed "pyright clean"; that came from running it scoped to the one changed source file.)

🤖 Generated with Claude Code

[Regaddi]

Small, focused, and does what it says. _with_org with setdefault is the right shape, the if org: guards are harmless (requires_login resolves an org before any of these commands run, so it is never None in practice), and naming the org in the two set confirmation prompts is a real improvement on the destructive paths.

Approving. Three things worth picking up, none blocking.

1. The JSON "no data" sentinel changed, and the PR does not say so

spend_limit_get is called with not_found_is_empty=True, so a 404 returns (None, None) with no error. On that path show --output json previously emitted {} and now emits:

{
  "organization": "acme"
}

Confirmed against a mocked 404.

The reviewer note frames the risk as "anything doing an exact-dict comparison will see the new key". This is a different break: a consumer doing if not payload: no_data() now takes the wrong branch, because the empty result became truthy. The human path handles this case with its own explicit message, so only JSON consumers are affected.

"limitCents" in payload still separates the two states, so it is recoverable, but the CHANGELOG entry and the reviewer note should say it out loud. The alternative is to keep emitting {} when data is None in show, and only merge the org into a real payload.

2. set and clear JSON output has no test

Both call sites moved to _with_org(data, org), but the new tests only cover show. Given that the point of the PR is that all three commands attribute their output, at least one of those two deserves an assertion.

3. "pyright clean" in the test plan is not accurate

uv run pyright reports 90 errors on origin/main and 91 on this branch. The extra one is the new test calling show(...) without awaiting (reportUnusedCoroutine, tests/test_spend_limit.py:265). It matches the eight pre-existing occurrences in that same file and pyright is not in CI, so it is not a problem in itself, but the claim should be corrected rather than left standing.

Verified locally

• 443 tests pass.
• ruff format --diff and ruff check clean.
• Rendered the table by hand: the Organization row lands first, above Limit.

Nit

_render_show(data, org=None) keeps a default that renders no Organization row. Every call site passes an org, so the default exists only to keep the old TestRenderShow cases compiling. Making it required would let the signature enforce the invariant this PR is establishing.

[github-actions]

Docs automation failed for this PR — its documentation is missing.

The update-docs workflow did not produce a docs PR: run log

Re-run it once the cause is addressed:

gh workflow run update-docs.yml -f pr_number=200 --repo daily-co/pipecat-cloud

If this PR genuinely needs no documentation, remove the
docs-automation-failed label.