[architect] pkg/dashboard/contribute_ws.go god file: 7,175 lines, 140+ funcs, 8+ domains in one file (incl. 1,155-line HandleWS, 687-line selectTask)

[hivecommons-hive]

Architecture Finding

Type: tech-debt
Affected area: src/pkg/dashboard/contribute_ws.go

contribute_ws.go has grown to 7,175 lines and ~140 functions — the second-largest file in pkg/dashboard after the api_contribute.go split (#7435). It mixes at least eight separable domains in one compilation unit:

1. WS transport/session — HandleWS (a single 1,155-line function, lines 3624–4778), sendJSON, writeProtocolPing, closeWithReason, heartbeatLoop, connectionRegistered
2. Task-lease persistence — recordLease/renewLease/revokeLease/lookupLease, saveLeasesLocked/loadLeases, reclaimExpiredLeases
3. Cooldown & failure bookkeeping — recordTaskFailure*, isTaskInCooldown*, bookReleaseCooldown, failureCooldown*
4. Completed/failed task ledgers — load/save Completed/Failed tasks, markTaskCompletedVerdict*, no-work-verdict suppression, no-PR streaks
5. Task selection — selectTask (687 lines, 6228–6914), leasedIssueKeys, resumeGateReason
6. Task-prompt rendering — buildTaskPrompt*, taskBaseBranch, releaseLineFromTitle, sourceLabel, attributionPromptInstruction (~460 lines of pure functions)
7. Token/credential lifecycle — mintScopedToken, maybeRefreshToken, sendTokenRefresh*, deliverTaskCredential, contributorCanPush
8. Fleet/ops snapshots — FleetSnapshot, LiveStates, RoleBreakdown, ActiveConnections, CooldownCounts

Impact

Every fix in one domain (e.g., the #7457 identity work, token-refresh changes, lease semantics) rebases against unrelated churn in the same file; reviewers must load a 7k-line context to review a 30-line diff; test-to-code navigation is already split across 20+ contribute_*_test.go files while the code under test sits in one blob. HandleWS at 1,155 lines and selectTask at 687 lines are beyond safe-review size on their own.

Recommendation

Split by domain, verbatim moves, no behavior change — same recipe as the api_contribute.go split (#7435):

• contribute_task_prompt.go (pure prompt rendering — stage 1, PR incoming)
• contribute_leases.go (lease persistence + reclaim)
• contribute_cooldowns.go (cooldown/failure bookkeeping)
• contribute_ledgers.go (completed/failed/no-work/streak persistence)
• contribute_select.go (selectTask + gating helpers)
• contribute_tokens.go (mint/refresh/deliver credential paths)
• Longer term: decompose HandleWS into named phase helpers.

---

Filed by architect agent (ACMM L5 — hold-gated mode)

🐝 Hive Agent: architect | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=architect backend=copilot model=claude-opus-4-6

[clubanderson]

Progress: contribute_ws.go 7,175 → 3,339 lines on v4 after #7492 (task prompt), #7498 (lease/cooldown/ledger/token), #7511 (selection + fleet snapshot). HandleWS itself is still the big remaining body; leaving open.

[github-actions]

PR #7498 has merged (#7498) and referenced this issue with a non-closing Refs #7491 rather than a closing keyword.

Can this issue now be closed, or is there remaining work it should keep tracking?

[clubanderson]

Closing — every file in the recommendation exists on v4 (and on v5 via #7525/#7545):

contribute_task_prompt.go, contribute_leases.go, contribute_cooldowns.go, contribute_ledgers.go, contribute_select.go, contribute_tokens.go (plus contribute_failure_streak.go, contribute_turn.go, contribute_withheld.go, contribute_sse.go).

src/pkg/dashboard/contribute_ws.go on v4 @ 201c69c: 7,175 → 3,500 lines, 140+ → 56 funcs. The remaining 'longer term' item — decomposing the 1,208-line HandleWS into named phase helpers — is tracked separately so it doesn't keep this finding open indefinitely.

[github-actions]

@kubestellar-hive[bot] this issue was closed by @clubanderson as completed.

If this was closed by a merged PR, the fix only takes effect once it ships in a tagged release and your deployment updates to it — check your running version against the fix before concluding the fix failed.

If you are still seeing the problem, comment /reopen and it will be reopened for you — you do not need write access. Please add what you are still observing (running version, logs, steps) so it is not closed the same way again.