fix(cli): offline-vs-no-env classifier keys on the probe name, not the remembered label (bugbot #266)

[saadqbal]

What

Follow-up to #267. Cursor Bugbot flagged (learned rule "commands targeting the active client must use bindActiveClientNamespace"):

Preferring the remembered client label onto env.name before the offline-vs-no-env check makes that check treat a non-empty ActiveClientName/ActiveClientID as evidence of an environment, even when ActiveClientNamespace is empty… a leftover id/name alone can render Offline with the full menu instead of the no-environment installer path.

resolveHomeModel's classifier uses provisioned || env.name != "", where env.name != "" is meant to signal "the probe surfaced an environment name." #267 made env.name carry the remembered display label, so a stale ActiveClientName/ID with no cached namespace now flips the no-env installer path to Offline.

Fix

Capture the probe's own surfaced name (probeNamedEnv) before the display-name override, and classify on provisioned || probeNamedEnv. The remembered label still drives the display name; it no longer counts as evidence of a reachable environment.

(In the real flow the probe only surfaces a name on localLive/localDegraded, never on the localNoRelease/localUnreachable branch this classifier runs on — so this restores the intended "namespace is the signal" behavior while preserving the synthetic probe-surfaced-a-name case the existing tests pin.)

Verification

• gofmt + go vet clean; go test ./internal/cli/ green.
• New TestResolveHomeModel_LeftoverNameNoNamespaceIsNoEnv (leftover name, no namespace → homeNoEnv) — fails on current develop (env.name="stale-id" → Offline), passes on the fix. All existing state-machine tests still pass.

Once merged to develop, promote PR #266 picks it up.

🤖 Generated with Claude Code

---

Note

Low Risk
Narrow CLI home state-machine fix with a dedicated regression test; no auth, data, or cluster mutation paths.

Overview
Fixes a regression where the bare tracebloc home screen could show Offline with the full menu when config still had a stale ActiveClientName/ActiveClientID but no cached ActiveClientNamespace. After the display-name override, env.name was non-empty, so provisioned || env.name != "" treated that as evidence of an environment.

resolveHomeModel now records probeNamedEnv from the cluster probe before applying the remembered display label, and uses provisioned || probeNamedEnv for the offline vs no-environment branch. Remembered names still drive envName for display; they no longer affect classification.

Adds TestResolveHomeModel_LeftoverNameNoNamespaceIsNoEnv so leftover name without namespace maps to homeNoEnv (installer path), not homeOffline.

^{Reviewed by Cursor Bugbot for commit 0f791a4. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

👋 Heads-up — Code review queue is at 36 / 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#181 — feat(weights): normalize-on-read — mixed cycles average instead of rejecting (SEC-03 §8 step 2) · author: @shujaatTracebloc · no reviewer assigned
• averaging-service#182 — feat(weights): averaging writes SafeTensors (TF + PyTorch) — SEC-03 §8 step 4 · author: @shujaatTracebloc · no reviewer assigned
• backend#1079 — feat(global_meta): edge dataset_meta exposure + attributes contract at ingest (#924 G4a) · author: @divyasinghds · no reviewer assigned
• backend#1086 — docs(rfc): SafeTensors weight-format migration — SEC-03 Phase 1 (RFC 0004) · author: @shujaatTracebloc · no reviewer assigned
• backend#1093 — chore(deps): bump django from 5.2.14 to 5.2.15 · author: @dependabot · no reviewer assigned
• backend#1095 — feat(experiment): configurable preprocessing knobs incl. tabular scaler — RFC 0003 L1 + L1b (#1094) · author: @LukasWodka · no reviewer assigned
• backend#1100 — fix(boot): pin SDK install to tracebloc==0.11.2, drop 404 dev line (#1098) · author: @LukasWodka · reviewer: @saqlainsyed007
• backend#1104 — perf(experiments): remove in-request completion sweep from status ping (part 2/2, #984) · author: @aptracebloc · no reviewer assigned
• cli#266 — main - > enhance CLI features and tests · author: @saadqbal · no reviewer assigned
• cli#268 — test: coverage wave 2 — utility packages toward 95% · author: @LukasWodka · no reviewer assigned

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

[aptracebloc]

Verified the fix — approving.

probeNamedEnv := env.name != "" is captured before the display-name override, and the classifier now keys on provisioned || probeNamedEnv. That correctly decouples the classification signal from the display label: the remembered name still drives what's shown, but only a cached namespace (provisioned) or the probe's own surfaced name count as evidence of a reachable environment. So a stale ActiveClientName/ID with no cached namespace now falls through to the no-env installer path instead of Offline. TestResolveHomeModel_LeftoverNameNoNamespaceIsNoEnv is a proper mutation guard. This is the right structural fix (not just a patch) — it restores #244's "namespace is the signal" intent while keeping #267's display improvement.

Caveat / process note: the home-screen state machine (from #244) has now taken a few follow-on fixes off the #266 promote chain (#267 label unification → this classifier fix). Each is well-tested and it's converging, but the recurring theme is env.name being overloaded for both display and classification. Before #266 promotes, it's worth a quick holistic once-over of resolveHomeModel as a whole to confirm no other display/classification coupling is lurking — cheaper to catch now than as another promote-chain follow-up.

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