1.16.0 (2026-11-03)
1.15.0 (2026-10-27)
1.14.0 (2026-10-22)
1.13.0 (2026-09-04)
1.12.0 (2026-08-05)
1.11.0 (2026-08-04)
1.10.0 (2026-08-03)
1.9.1 (2026-07-27)
1.9.0 (2026-07-27)
1.8.0 (2026-07-25)
1.7.3 (2026-07-25)
- correct GreenNode spelling (greenode -> greennode) (#45) (9c9f847)
- vks: --secondary-subnets takes CIDRs, not subnet IDs (#44) (233cc4c)
1.7.2 (2026-07-07)
1.7.1 (2026-07-07)
1.7.0 (2026-07-07)
- per-product CLAUDE.md (tier 2) and agent skills (#35) (3a1b7cd)
- rebrand VNG Cloud to GreenNode in VKS and shared text (#38) (d0fd019)
- vks: AWS-CLI-style reference format for all commands (#30) (1fe168d)
- vks: create-cluster gains the remaining optional CreateClusterDto fields: --secondary-subnets, --list-subnet-ids, --node-netmask-size, --service-endpoint, --az-strategy, and struct-valued --auto-upgrade-config/--auto-healing-config (shorthand or JSON, matching create-nodegroup). --subnet-id is now optional (per spec; pass --subnet-id or --list-subnet-ids or neither, the server validates)
- core: cli.ParseStructFlag gains bool-field coercion (ParseStructFlagTyped) so struct flags can carry boolean values in shorthand
- core: Guard --endpoint-url against sending the IAM bearer token to untrusted hosts (SEC-08): hosts outside vngcloud.vn/greenode.ai are warned over TLS, and blocked when there is no TLS protection (plain http or --no-verify-ssl) unless --allow-untrusted-endpoint is set
- core: Redact credential fields (embedded kubeconfig, tokens, client certs/keys, secrets) in --debug request/response logging so 'update-kubeconfig --debug' and similar no longer print long-lived credentials to stderr (SEC-07)
- vks: create-cluster now provisions the control plane only and no longer creates a default node group: removed --node-group-name/--flavor-id/--os/--disk-type/--ssh-key-id/--disk-size/--num-nodes/--private-nodes/--security-groups/--labels/--taints. Create workers separately with create-nodegroup
- vks: Add --dry-run to the remaining mutating VKS commands (config-auto-healing, config-auto-upgrade, update-nodegroup-metadata, upgrade-nodegroup-version, generate-kubeconfig); it previews the request payload and exits without calling the API (works offline)
- output: Implement the --color flag (on/off/auto, like aws): color status values (ACTIVE=green, ERROR/FAILED=red, CREATING/PENDING=yellow) in text and table output. auto colors only when stdout is a terminal and NO_COLOR is unset; JSON output is never colored
- core: Add --dry-run to destructive commands that lacked it (vServer server stop/reboot/delete, volume/vpc/subnet/secgroup/secgroup-rule delete, and vks delete-auto-upgrade-config) and unify the preview + confirmation flow across delete commands via shared helpers
- configure: Mask credential values in 'configure set' output so client_id/client_secret are no longer echoed in plaintext to stdout (consistent with 'configure list'); non-sensitive values are still shown
- vks: grn vks wait now aborts immediately on a permanent error (HTTP 403/401/400, or 404 for an active waiter) instead of polling until timeout; transient errors (5xx, network) still retry
- vks: Rename 'set-auto-upgrade-config' to 'config-auto-upgrade'; the old name still works as a deprecated alias
- core: On error, print a single concise 'Error: ...' line instead of cobra's duplicated error plus a full usage dump (use 'grn --help' for usage)
- vks: Document the --unhealthy-range format in config-auto-healing help: expects "[min-max]" (e.g. "[2-5]")
- configure: Fix panic (nil pointer) when running 'grn configure set/list --profile ' for a profile that does not exist yet. LoadConfig now reads the credentials and config files independently, so a profile created via 'configure set' (config file only) loads its region/output/project_id correctly, and 'configure get'/'configure list' on a truly missing profile report a clear 'profile does not exist' error (exit 1, like 'aws configure') instead of crashing
- core: Honor --cli-connect-timeout: the TCP connect and TLS handshake are now bounded by the flag (previously it was accepted but ignored, so a slow/unreachable endpoint hung for ~127s regardless). Wired via the HTTP transport's dialer in both VKS and vServer clients
- output: Reject an invalid --output or --color value (e.g. a typo like 'tabel') with a clear error, a 'maybe you meant' suggestion, and a non-zero exit, instead of silently falling back
- vks: Add struct-valued flags (shorthand or JSON) to create-nodegroup: --tags, --secondary-subnets, --auto-scale, --placement-group, --upgrade-config
- vks: update-nodegroup: drop deprecated --labels/--taints (use update-nodegroup-metadata); replace --auto-scale-min/max and --upgrade-strategy/max-surge/max-unavailable with struct flags --auto-scale and --upgrade-config
- configure: Fix panic (nil pointer) when running 'grn configure --profile ' for a profile that does not exist yet; start from empty defaults so the profile can be created
- vks: Return a non-zero exit code and 'unknown command' error for invalid subcommands of grn vks/vserver (and their groups) instead of silently printing help with exit 0
- vks: Remove get-nodegroup-events; network-type CALICO->TIGERA; --os adds rocky; replace boolean toggle flags with --private-cluster/--private-nodes/--load-balancer-plugin/--block-store-csi-plugin <enabled|disabled>
- vserver: Add vServer CLI: vpc, subnet, secgroup, volume, volume-type, flavor, image, and server commands
- vks: Align node group commands with latest API: remove --image-id (no longer in the API), add --os (ubuntu/linux, default ubuntu) for create-cluster/create-nodegroup; update-nodegroup no longer requires/sends image