feat(cli#128): selected-vs-connected — bind data commands to the active client (§7.3)

[saadqbal]

Closes #128. Implements RFC-0001 §7.3 ("'selected' is not 'connected'") — client use set only a local pointer while the data commands ignored it and targeted whatever the current kubeconfig pointed at, risking a silent wrong-target on multi-cluster / remote hosts.

Three logical commits (each builds + tests on its own):

1. refactor — extract resolveClusterTarget, collapsing the duplicated Load → NewClientset → DiscoverParentRelease (→ DiscoverSharedPVC) block + its exit-code contract (3 kubeconfig/clientset, 4 missing release/PVC) out of data ingest/list/delete. This is the resolveClusterTarget task from CLI quick wins: cluster-connect + kubeconfig-flag helpers, rename sweep, pinned lint + deadcode gate #127, landed here as the base (the rest of CLI quick wins: cluster-connect + kubeconfig-flag helpers, rename sweep, pinned lint + deadcode gate #127 stays a separate quick-wins PR, different epic). cluster doctor deliberately excluded (different 2/3 exit contract).
2. bind data commands (§7.3) — cache the active client's namespace + name in the env profile at client use/create (setActiveClient), so data commands bind without a backend round-trip (they run cluster-local, maybe offline). data ingest/list/delete default -n to it when the user gave neither --namespace nor --context; no active client → unchanged current-context behavior (backward compatible). A "no release here" miss on a bound namespace is rewritten to the §7.3 "active client runs on another machine" guidance (typed noParentReleaseError); a PVC-missing failure passes through.
3. client list state column (§7.3) — add online/offline/pending from the backend EdgeDevice.status code (the "connected somewhere" signal, C.4) beside the (active — this machine) marker, plus a hint separating selected (local pointer) from connected (backend heartbeat). Plain words, not emoji glyphs (§12 mojibake watch-item).

Design decisions

• Namespace source: cached at use/create (local) rather than a per-run backend lookup — keeps data commands cluster-local/offline-capable. Adds active_client_namespace + active_client_name to the config Profile (both omitempty → v1/v2 configs unaffected).
• "Connected" = backend heartbeat only (status), not per-row local cluster probing (that's the slow path flagged in §12); local reachability stays a per-active-client concern.

Not in scope

• Optional soft reachability probe on client use (issue marks it optional; would add a cluster dial to an intentionally cluster-free command) — left as a follow-up.

Verification

go build ./..., go vet ./..., gofmt, and the full go test ./... suite are green; new tests in clustertarget_test.go cover the state-label mapping, binding (defaults / explicit-override no-op / no-active-client no-op), explain (rewrite vs pass-through), and the setActiveClient cache.

🤖 Generated with Claude Code

---

Note

Medium Risk
Data commands may target a different namespace when an active client is set, which changes default cluster targeting for multi-cluster setups; behavior is backward compatible without a cached namespace.

Overview
Implements RFC-0001 §7.3 so data ingest, data list, and data delete no longer ignore the machine's active client and silently target whatever the kubeconfig's current context implies.

Active client binding: client create / client use now call setActiveClient, which persists active_client_namespace and active_client_name on the env profile (alongside active_client_id). When neither --namespace nor --context is set, data commands default the target namespace from that cache—no backend call—while explicit flags still win. No cached active client keeps the old current-context behavior.

Cluster resolution: Shared resolveClusterTarget replaces duplicated kubeconfig → clientset → parent release (→ PVC) logic and preserves exit codes 3/4. DiscoverParentRelease exposes ErrNoParentRelease so a true "no release in namespace" can be distinguished from RBAC or ambiguous multi-release errors; bound-namespace misses get a "active client runs on another machine" message via activeClientBinding.explain (PVC failures are not rewritten).

client list: Rows show backend state= (online / offline / pending as plain text), (active — this machine) on the selected id, and a short hint separating local selected from backend connected heartbeat.

^{Reviewed by Cursor Bugbot for commit e798dc8. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

👋 Heads-up — Code review queue is at 31 / 30

Above the WIP limit. The team convention is to review existing PRs before opening new work.

Open PRs currently in Code review (oldest first):

• averaging-service#143 — chore(release): promote develop → staging (week of 2026-07-06) · author: @aptracebloc · no reviewer assigned
• averaging-service#156 — feat(strategies): FedMedian Byzantine-robust aggregation (CLI: a mistyped subcommand (e.g. dataset pus) prints help and exits 0 instead of erroring #75) · author: @shujaatTracebloc · reviewer: @divyasinghds
• averaging-service#157 — feat(strategies): FedTrimmedAvg Byzantine-robust aggregation (CLI: a mistyped subcommand (e.g. dataset pus) prints help and exits 0 instead of erroring #75) · author: @shujaatTracebloc · reviewer: @divyasinghds
• averaging-service#158 — feat(strategies): Krum / MultiKrum Byzantine-robust selection (CLI: a mistyped subcommand (e.g. dataset pus) prints help and exits 0 instead of erroring #75) · author: @shujaatTracebloc · reviewer: @divyasinghds
• averaging-service#159 — feat(strategies): FedProx aggregation strategy (dataset push: --category help lists 5 of 9 categories, and unknown-vs-empty category errors give two different lists (one leaks unsupported segmentation) #74) · author: @shujaatTracebloc · reviewer: @divyasinghds
• averaging-service#160 — feat(strategies): FedOpt family — FedAdam / FedYogi / FedAdagrad (dataset push: --category help lists 5 of 9 categories, and unknown-vs-empty category errors give two different lists (one leaks unsupported segmentation) #74) · author: @shujaatTracebloc · reviewer: @divyasinghds
• averaging-service#161 — feat(strategies): resolve per-experiment aggregation hyperparameters (dataset push: --category help lists 5 of 9 categories, and unknown-vs-empty category errors give two different lists (one leaks unsupported segmentation) #74) · author: @shujaatTracebloc · reviewer: @aptracebloc
• backend#926 — chore(release): promote develop → staging (week of 2026-07-06) · author: @aptracebloc · reviewer: @saadqbal
• backend#927 — feat(validators): schema check for tabular/text/token classification (#924 G2) · author: @divyasinghds · no reviewer assigned
• backend#931 — feat(validators): generic all-category checks + validation matrix (#924 Phase 0/G12) · author: @divyasinghds · reviewer: @saqlainsyed007

Pull from review before opening new work. (This is a nudge from the kanban WIP check, not a block.)

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Want fixes drafted automatically? Bugbot Autofix can create code changes for findings. A team admin can enable Autofix in the Cursor dashboard.

^{Reviewed by Cursor Bugbot for commit 758af8d. Configure here.}

[aptracebloc]

Approving. Clean implementation of RFC-0001 §7.3 — the refactor is a solid DRY win and the binding is careful. A couple of non-blocking notes below.

Summary

• resolveClusterTarget centralizes the Load → NewClientset → DiscoverParentRelease (→ PVC) block + the 3/4 exit contract out of the three data commands (cluster doctor correctly excluded — different 2/3 contract).
• §7.3 binding: client use/create cache the active client's namespace+name; data commands default -n to it only when neither --namespace nor --context was given. A "no release here" miss on a bound namespace is rewritten to the "runs on another machine" guidance.
• client list shows state= (backend heartbeat) + (active — this machine) + a selected-vs-connected hint.

What stands out

• Precise error plumbing: the new cluster.ErrNoParentRelease sentinel + errors.Is/errors.As mean only the genuine "namespace has no release" binding-miss gets rewritten — RBAC / ambiguous-multi-release / PVC-missing errors pass through untouched. Not a blanket rewrite.
• Backward-compatible + non-presumptuous: no active client / no cache / explicit --namespace|--context → unchanged current-context behavior; explicit flags always win; new config fields are omitempty (v1/v2 configs unaffected).
• Offline-correct: namespace cached at use/create rather than a per-run backend lookup — and namespace is effectively immutable per client, so staleness isn't a concern.
• Good coverage of the binding (default / explicit-override no-op / no-active no-op), the explain rewrite-vs-passthrough, the sentinel, and the state-label mapping.

Non-blocking notes

1. Deliberate default-targeting change: once a user has run client use/create, a bare data … now targets the cached namespace, not the kubeconfig's current namespace. That's the §7.3 intent and it's mitigated (explicit flags override, clear message) — just flagging it changes behavior for already-migrated users.
2. Binding is namespace-string-based, not identity-based — it defaults the namespace but not the cluster/context, so a wrong cluster is detected via the discovery miss + "runs elsewhere" message rather than prevented. Narrow gap: a different cluster hosting a release under the same namespace name wouldn't be caught (silent right-namespace-wrong-cluster). Almost certainly fine given slug-derived namespaces — worth a one-line confirmation, not a change.

— drafted with Claude (Opus 4.8), sent by @aptracebloc