NVIDIA · mesutoezdil · Jul 21, 2026
@@ -19,17 +19,16 @@ Full subcommand list. Always pair with [flags.md](flags.md) for global flags and
 | `self-hosted uninstall --compute-plane --cluster-name=X` | Per-plane primitive: `helmfile destroy` on one compute plane (no ICMS unregister, no drain) | progress | Just the helm releases on that compute plane |
 | `self-hosted uninstall --control-plane [--force-with-registered-clusters]` | Per-plane primitive: `helmfile destroy` on the control plane | progress | Refuses if compute planes still registered unless `--force-with-registered-clusters` |
 | `self-hosted uninstall --no-apply <plane>` | Render delete YAML via `helm get manifest` per release | YAML on stdout | GitOps; pipe to `kubectl delete -f -` or commit + Argo applies. Mirrors `install --no-apply` |
-| `self-hosted down --cluster-name=X [--drain-active=true\|false\|prompt] [--remove-persistent]` | Orchestrator: drain → uninstall --compute-plane → cluster delete in ICMS | progress | Symmetric companion to `up --cluster-name=X` |
+| `self-hosted down --cluster-name=X [--drain-active] [--remove-persistent]` | Orchestrator: drain → uninstall --compute-plane → cluster delete in ICMS | progress | Symmetric companion to `up --cluster-name=X` |
 | `self-hosted down --all [--confirm]` | Orchestrator: tear down every registered compute plane + control plane | progress | Bounded parallelism via `--all-concurrency` (default 4); `--confirm` required in non-interactive |
 | `self-hosted down --plan-only ...` | Orchestrator dry-run: phases + helm releases + ICMS rows + ETAs (no helm/helmfile contact) | JSONL | **Always run before any actual `down`** |
 
 ## Cluster management
 
 | Command | Purpose | Notes |
 |---|---|---|
-| `cluster register --name=X --nca-id=Y --region=Z [--ignore-existing]` | Register a JWKS+OIDC with ICMS | Used by `up` Phase 5 internally; standalone for manual registration |
+| `cluster register --name=X --nca-id=Y [--region=Z] [--ignore-existing]` | Register a JWKS+OIDC with ICMS | Used by `up` Phase 5 internally; standalone for manual registration |
 | `cluster list` | List registered clusters | Output is YAML by default; `--json` for machine |
-| `cluster get --cluster-id=ID` | Get one cluster's metadata | |
 | `cluster rotate --cluster-id=ID` | Re-fetch JWKS from K8s and PUT to ICMS | After K8s API server signing key rotation |
 | `cluster delete --cluster-id=ID` | Remove ICMS row | **DESTRUCTIVE: confirm with user** |
 
@@ -76,7 +75,7 @@ Task commands require `NVCF_API_KEY` set to a task-scoped key and `NVCF_BASE_NVC
 | `api-key generate [--for function\|task] [--description=…] [--expires-in=…]` | Mint API keys | Default (no `--for`): generates both a function key and a task key; `--for function` or `--for task` generates one only; `--scopes` requires `--for` |
 | `api-key generate --validate` | Mint and immediately validate each generated key | Validates each key against its correct audience (NVCF for function keys, NVCT for task keys) |
 | `api-key list` | List API keys for current owner | Lists NVCF-scoped keys only |
-| `api-key delete --id=ID` / `api-key revoke --id=ID` | Remove an API key | **Confirm with user** |
+| `api-key delete KEY-ID` / `api-key revoke KEY-ID` | Remove an API key | **Confirm with user** |
 | `api-key show` | Show currently saved API key from state | |
 
 ## Agent skill

@@ -8,7 +8,6 @@ Stable across subcommands. Use these to drive agent retry / surfacing logic.
 | `1` | Generic error | Helm render failed; network unreachable; file not found; YAML parse error |
 | `2` | Pre-flight check failed | Gateway API CRDs missing; kubectl not on PATH; default StorageClass absent |
 | `3` | Admin auth failed | No token + `--non-interactive` set; ICMS rejected JWT; init endpoint unreachable |
-| `4` | Cluster registration failed | ICMS unreachable; JWKS fingerprint conflict; missing NCA ID |
 | `5` | Manifest apply or `--wait` timed out | Helm install timeout; check polled but didn't pass before duration |
 | `130` | Cancelled by SIGINT/SIGTERM | User Ctrl-C; CI budget exceeded; pod evicted |
 
@@ -20,9 +19,8 @@ Stable across subcommands. Use these to drive agent retry / surfacing logic.
 | `1` | Surface stderr to user; ask before any retry. Generic errors usually need a human to read the message. |
 | `2` | Surface failed checks + their `hintURL`. Don't propose `up` until the user fixes prereqs. |
 | `3` | Suggest `nvcf-cli init` (interactive) OR `--token=$JWT` (CI). Don't auto-mint without user OK. |
-| `4` | Show `errMessage` + `remediation`. Branch on `retryClass`: `immediate` → may retry now (with user OK); `backoff` → wait `retryAfterSec` then retry (with user OK); `after_remediation` / `none` / `unknown` → operator action required, do not auto-retry. |
 | `5` | Ask user if they want to wait longer, dig into the stalled component, or cancel. |
-| `130` | Don't retry automatically — user explicitly cancelled. Confirm before re-running. |
+| `130` | Don't retry automatically; user explicitly cancelled. Confirm before re-running. |
 
 ## Where to find more detail
 

@@ -16,11 +16,11 @@
 | `--compute-plane-stack=…` | Compute-plane bundle source: local path, git URL, or `oci://` URL | embedded OCI URL pinned by CLI version |
 | `--env=local\|prd\|…` | Helmfile environment name | `local` for dev builds, `prd` for releases |
 | `--non-interactive` | Disable all stdin prompts | `false` |
-| `--token=$JWT` | Admin JWT, overrides stored session | — |
-| `--no-apply` | `install` only — emit YAML only, don't kubectl apply | `false` |
+| `--token=$JWT` | Admin JWT, overrides stored session | - |
+| `--no-apply` | `install` only - emit YAML, do not kubectl apply | `false` |
 | `--output=text\|json` | Legacy alias for `--json` (deprecated, removed in next major) | `text` |
 | `--plain` | Force plain streaming output | auto-detect |
-| `--wait DURATION` | `check` only; block until pass | — |
+| `--wait DURATION` | `check` only; block until pass | - |
 | `--control-plane-context CTX` | kubectl context for control plane (REQ-20) | current context |
 | `--compute-plane-context CTX` | kubectl context for compute plane (REQ-20) | current context |
 | `--icms-url URL` | Public ICMS URL; required when contexts differ | derived from `base_http_url` |
@@ -30,7 +30,7 @@
 
 | Flag | Purpose | Default |
 |---|---|---|
-| `--cluster-name NAME` | ICMS cluster row identifier (required) | — |
+| `--cluster-name NAME` | ICMS cluster row identifier (required) | - |
 | `--nca-id ID` | NCA account ID | `nvcf-default` |
 | `--region REGION` | Cluster region | `us-west-1` |
 | `--plan-only` | Dry-run; emit phase plan + ETAs without changes | `false` |
@@ -39,23 +39,21 @@
 
 | Flag | Purpose | Default |
 |---|---|---|
-| `--cluster-name NAME` | Limit to one compute plane | — (control + all clusters) |
+| `--cluster-name NAME` | Limit to one compute plane | - (control + all clusters) |
 | `--watch` | Live re-render | `false` |
 | `--watch-interval DUR` | `--watch` cadence | `5s` |
 | `--component NAME` | Filter components panel | all |
-| `--show-events DUR` | Recent events window | `5m` |
 | `--no-events` | Skip events panel | `false` |
 
 ## `cluster register`-specific
 
 | Flag | Purpose | Default |
 |---|---|---|
-| `--name NAME` | Cluster name (required) | — |
-| `--nca-id ID` | NCA ID (required) | — |
-| `--region REGION` | Region (required, non-empty) | — |
+| `--name NAME` | Cluster name (required) | - |
+| `--nca-id ID` | NCA ID (required) | - |
+| `--region REGION` | Region | `us-west-1` |
 | `--icms-url URL` | ICMS endpoint | from config |
 | `--ignore-existing` | Match-or-create instead of fail-on-exists | `false` |
-| `--identity-source psat\|spire` | Identity source | `psat` |
 
 ## `api-key generate`-specific
 
@@ -107,12 +105,12 @@ In JSON, `function update` accepts `modelUpdates[]` entries with `modelName` and
 
 | Flag | Purpose | Default |
 |---|---|---|
-| `--function-id ID` | Function ID (required) | — |
+| `--function-id ID` | Function ID (required) | - |
 | `--gpu NAME` | GPU family | `H100` |
 | `--instance-type TYPE` | SKU | `NCP.GPU.H100_1x` |
 | `--min-instances N` | Min replicas | `1` |
 | `--max-instances N` | Max replicas | `1` |
-| `--max-request-concurrency N` | Concurrency | `10` |
+| `--max-request-concurrency N` | Concurrency | `0` (unset; server default applies) |
 | `--clusters NAME[,NAME]` | Target specific compute clusters | (any matching SKU) |
 
 ## `task create`-specific
@@ -127,22 +125,22 @@ In JSON, `function update` accepts `modelUpdates[]` entries with `modelName` and
 | `--image IMAGE` | Container image | Mutually exclusive with `--helm-chart` |
 | `--helm-chart CHART` | Helm chart URL or OCI reference | Mutually exclusive with `--image` |
 | `--input-file FILE` | JSON task spec file | CLI flags override JSON values when both are set |
-| `--backend BACKEND` | Backend / CSP | — |
+| `--backend BACKEND` | Backend / CSP | - |
 | `--clusters NAME[,NAME]` | Specific clusters within the instance | Comma-separated |
 | `--container-args ARGS` | Args passed when launching the container | Single string, e.g. `"--epochs 10 --batch-size 32"` |
-| `--container-env NAME=value` | Container environment variable; repeatable | — |
-| `--secrets NAME=value` | Secret environment variable; repeatable; encrypted at rest | — |
+| `--container-env NAME=value` | Container environment variable; repeatable | - |
+| `--secrets NAME=value` | Secret environment variable; repeatable; encrypted at rest | - |
 | `--max-runtime DURATION` | Max wall-clock run time (ISO 8601, e.g. `PT4H`) | No default; omit for no time limit |
 | `--max-queued DURATION` | Max time in queue before cancellation (ISO 8601) | Server default `PT72H` |
 | `--termination-grace DURATION` | Grace period after stop signal (ISO 8601) | Server default `PT1H` |
-| `--result-strategy UPLOAD\|NONE` | How to handle task output artifacts | — |
-| `--results-location LOCATION` | Result upload target, format `org-name/[team-name/]model-name`; required when `--result-strategy=UPLOAD` | — |
-| `--models NAME:VERSION:URI` | Model artifact; repeatable | — |
-| `--resources NAME:VERSION:URI` | Resource artifact; repeatable | — |
+| `--result-strategy UPLOAD\|NONE` | How to handle task output artifacts | - |
+| `--results-location LOCATION` | Result upload target, format `org-name/[team-name/]model-name`; required when `--result-strategy=UPLOAD` | - |
+| `--models NAME:VERSION:URI` | Model artifact; repeatable | - |
+| `--resources NAME:VERSION:URI` | Resource artifact; repeatable | - |
 | `--tags TAG[,TAG]` | Task tags | Comma-separated |
-| `--description TEXT` | Task description | — |
-| `--logs-telemetry-id UUID` | Logs telemetry endpoint ID | — |
-| `--metrics-telemetry-id UUID` | Metrics telemetry endpoint ID | — |
-| `--traces-telemetry-id UUID` | Traces telemetry endpoint ID | — |
+| `--description TEXT` | Task description | - |
+| `--logs-telemetry-id UUID` | Logs telemetry endpoint ID | - |
+| `--metrics-telemetry-id UUID` | Metrics telemetry endpoint ID | - |
+| `--traces-telemetry-id UUID` | Traces telemetry endpoint ID | - |
 
 
@@ -52,14 +52,14 @@ api_keys_service_url: https://api-keys.nvcf.nvidia.com
 # Config key: api_key
 # Environment variable: NVCF_API_KEY
 # Scope: invoke_function, list_functions, queue_details, list_functions_details
-# Note: Stored in ~/.nvcf-cli-state.json after running 'nvcf-cli api-key generate'
+# Note: Stored in ~/.nvcf-cli.state after running 'nvcf-cli api-key generate'
 # api_key: nvapi-xxxxx
 
 # NVCF Admin Token / JWT (for admin operations: create, deploy, delete, update)
 # Config key: token
 # Environment variable: NVCF_TOKEN
 # Scope: admin:* (all admin operations)
-# Note: Stored in ~/.nvcf-cli-state.json after running 'nvcf-cli init'
+# Note: Stored in ~/.nvcf-cli.state after running 'nvcf-cli init'
 # token: eyJhbGciOiJSUzI1NiIsInR5cCI6IkpXVCJ9...
 
 # NVIDIA Cloud Account Client ID (your account/username)
@@ -213,7 +213,7 @@ api_keys_owner_id: svc@nvcf-api.local
 #    - No kubeconfig or kubectl needed!
 #    - Run: nvcf-cli init
 #    - Calls API Keys service directly: /v1/admin/keys
-#    - Tokens stored in: ~/.nvcf-cli-state.json
+#    - Tokens stored in: ~/.nvcf-cli.state
 #
 # 4. Token Usage:
 #    - NVCF_API_KEY: User operations (invoke, list, queue details)