[air] Add air list/get provisioned_capacity commands

[vinchenzo-db]

Adds two read commands for pre-provisioned AI Runtime capacity reservations (the same reservation a run targets via compute.provisioned_capacity_id):

• air list provisioned_capacity — the workspace's reservations (id, accelerator type, reserved accelerators).
• air get provisioned_capacity <id> — one reservation plus its accelerator usage summary (used / idle).

Both call AiWorkflowService's purpose-built public read API (GET /api/2.0/ai-training/provisioned-capacities[/{id}]) via a raw client.Do, matching the sibling air list workflows call. The platform tracks these internally as "guaranteed capacity"; the public surface (and this CLI) calls them "provisioned capacity". Usage is populated only by Get, so the list view omits it. Output uses the same JSON envelope + text pattern as the other air commands, and the SAFE-gate the server returns before rollout is surfaced as a clear FEATURE_DISABLED error rather than a retryable one.

Tests: unit tests for the projections, resource-name handling, int64-as-string decoding, pagination, and the JSON/text/not-found/feature-disabled paths; an acceptance test drives the real binary against a mocked endpoint (text + JSON goldens). Verified live against e2-dogfood: the CLI reaches the deployed handler and classifies its responses correctly (the API is SAFE-gated off there, so the successful list/get shape is covered by the goldens).

Changes

Why

Tests

e2e test

$ cd /home/v.chen/databricks-cli-wt-pcap                                                                                                                    
go build -o /tmp/air ./
# v.chen at ip-10-90-27-126 in ~/databricks-cli-wt-pcap (git:air-provisioned-capacity) [20:01:15]
$ /tmp/air experimental air list provisioned_capacity -p eng-nephos-e2
ID                                       ACCELERATOR    RESERVED
cap-uitest-nephos-a10g-test              GPU_1xA10      3
# v.chen at ip-10-90-27-126 in ~/databricks-cli-wt-pcap (git:air-provisioned-capacity) [20:07:54]
$ /tmp/air experimental air list provisioned_capacity -p eng-nephos-e2 -o json
{
  "v": 1,
  "ts": "2026-09-11T20:08:12Z",
  "data": {
    "provisioned_capacities": [
      {
        "provisioned_capacity_id": "cap-uitest-nephos-a10g-test",
        "accelerator_type": "GPU_1xA10",
        "reserved_accelerators": 3
      }
    ]
  }
}
# v.chen at ip-10-90-27-126 in ~/databricks-cli-wt-pcap (git:air-provisioned-capacity) [20:08:12]
$ /tmp/air experimental air get provisioned_capacity cap-uitest-nephos-a10g-test -p eng-nephos-e2 -o json
{
  "v": 1,
  "ts": "2026-09-11T20:08:40Z",
  "data": {
    "provisioned_capacity_id": "cap-uitest-nephos-a10g-test",
    "accelerator_type": "GPU_1xA10",
    "reserved_accelerators": 3,
    "used_accelerators": 0,
    "idle_accelerators": 3
  }
}
# v.chen at ip-10-90-27-126 in ~/databricks-cli-wt-pcap (git:air-provisioned-capacity) [20:08:40]
$ /tmp/air experimental air list --help
/tmp/air experimental air get --help
List your active runs for the current profile (use --all-status for finished runs)

Usage:
  databricks experimental air list [flags]
  databricks experimental air list [command]

Available Commands:
  provisioned_capacity List the pre-provisioned AI Runtime capacity reservations for the current workspace

Flags:
      --all-status           Show runs in all states (default: active only)
      --all-users            Show runs from all users
      --filter stringArray   Filter runs, e.g. experiment=foo* (repeatable)
  -h, --help                 help for list
      --limit int            Maximum number of runs to show (default 20)

Global Flags:
      --debug            enable debug logging
  -o, --output type      output type: text or json (default text)
  -p, --profile string   ~/.databrickscfg profile
  -t, --target string    bundle target to use (if applicable)

Use "databricks experimental air list [command] --help" for more information about a command.
Show status, configuration, and timing details for a specific run

Usage:
  databricks experimental air get JOB_RUN_ID [flags]
  databricks experimental air get [command]

Available Commands:
  provisioned_capacity Show a pre-provisioned AI Runtime capacity reservation, including its accelerator usage

Flags:
  -h, --help   help for get

Global Flags:
      --debug            enable debug logging
  -o, --output type      output type: text or json (default text)
  -p, --profile string   ~/.databrickscfg profile
  -t, --target string    bundle target to use (if applicable)

Use "databricks experimental air get [command] --help" for more information about a command.
# v.chen at ip-10-90-27-126 in ~/databricks-cli-wt-pcap (git:air-provisioned-capacity) [20:08:54]
$ 

[eng-dev-ecosystem-bot]

Integration test report

Commit: 204b91e

Run: 34642427409

	Env	🔄​flaky	💚​RECOVERED	✅​pass	🙈​skip	Time
💚​	aws linux		1	275	17	6:28
💚​	aws windows		1	277	15	5:56
💚​	azure linux		1	274	17	6:24
💚​	azure windows		1	276	15	4:47
🔄​	gcp linux	1	1	274	17	7:23
💚​	gcp windows		1	277	15	6:21

	Test Name	aws linux	aws windows	azure linux	azure windows	gcp linux	gcp windows
💚​	TestAccept	💚​R	💚​R	💚​R	💚​R	💚​R	💚​R
🔄​	TestSyncIncrementalSyncFileToPythonNotebook	✅​p	✅​p	✅​p	✅​p	🔄​f	✅​p

Top 3 slowest tests (at least 2 minutes):

duration	env	testname
4:37	gcp windows	TestAccept
3:24	azure windows	TestAccept
3:22	aws windows	TestAccept

[ben-hansen-db]

Some blocking comments, let's align first before merge.

[ben-hansen-db]

claude:

Both raw workspace calls omit the required workspace routing header. Unified hosts cannot reliably route these requests to Config.WorkspaceID.

[ben-hansen-db]

suggests instead

headers := auth.WorkspaceIDHeaders(w.Config)
apiClient.Do(ctx, http.MethodGet, provisionedCapacityPath, headers, nil, query, &resp)

[ben-hansen-db]

what about other not found errors or http 404?

apierr.ErrNotFound is broader

[ben-hansen-db]

this is very long... is the the phrase we want to use databricks air list provisioned-capacity??

databricks air list capacity or reservations, pools, commitments