refactor: split oversized modules and enforce anti-slop lint by leoisadev1 · Pull Request #343 · opencoredev/akeru-bot

leoisadev1 · 2026-10-02T14:12:52Z

Problem

The codebase had grown 158 TypeScript files past 800 lines, hundreds of rule-evading escapes (431 as unknown as, 291 @effect-diagnostics directives, 32 @ts-* directives, 27 lint disables), and no lint guard to stop new slop or raw palette colors from coming back.

What changed

Split oversized modules along real seams (components, services, test fixture groups). On the main source tree, files over 800 lines went from 158 to 22 by raw line count, and every file now passes the max-lines lint rule, which counts code lines. The two generated protocol schemas are untouched.
Removed the escapes instead of moving them: 0 as unknown as, 0 @ts-* directives and 0 @effect-diagnostics directives remain in source. The few remaining lint-disable hits are generated files (routeTree.gen.ts, mockServiceWorker.js) and the suppression scanner's own pattern.
Typed untrusted boundaries with schemas, used Effect services such as Path, FileSystem and Clock instead of ambient APIs, and moved shared work-log and usage parsing into client-runtime for web and mobile.
Vendored oxlint-plugin-anti-slop, added shadcn lint and an Akeru plugin (no raw palette strings, no lint suppressions, no module mocking), and made the satisfied rules errors. scripts/check-lint-suppressions.ts runs in CI.
Replaced raw palette classes with semantic theme tokens. Default light and dark themes resolve to the same computed colors, borders and animations as before; new tokens were added where no existing one matched.
Removed dependencies with no remaining consumers and updated AGENTS.md and docs/internals/lint.md.

About 250k of the added and removed lines are moved code (the same line deleted in one file and added in another). The genuinely new or changed content is roughly 67k added and 24k removed, mostly import headers in split files, tests, and the vendored lint plugin.

Behavior changes included

These are real fixes rather than mechanical cleanup. Each has a focused test and its own changeset:

Migration 023 propagates real failures and matches existing columns case-insensitively, as SQLite does.
Positional SQLite queries honor SafeIntegers.
The orchestration engine hydrates persisted command state, and failed dispatches clean up uploads on defects and interruption, not just typed failures.
A new-chat bootstrap that fails, dies or is cancelled after creating its thread deletes that thread exactly once.
Subscription credential access and refresh reload health state again.
OAuth completion ownership, bounded image reads, Browserbase validation, Unicode payload bounds, legacy plan compatibility and bounded tool inspection.

Review

The branch went through several rounds of independent per-slice review (27 slices, then a second pass on the approved ones, then re-reviews of every fix commit) until each reviewer approved. Restoring lost behavior covered null handling in bot tool inputs, markdown prop forwarding on mobile, the Alert fragment crash, skeleton, shimmer and splash animations, thread ordering by parsed time, and eight default-theme color drifts.

For local review, git diff --color-moved=dimmed-zebra origin/main... separates moved code from changed code.

Verification

tsgo reports 0 errors and 0 Effect diagnostics for server, web, desktop, contracts, shared and client-runtime. Mobile reports the same 33 route-typing errors that exist on main.
scripts/check-lint-suppressions.ts passes, and lint and format are clean on touched files.
Focused tests for every fix pass.

Not covered

No before and after screenshots. The visual goal was exact parity in the default themes, and the reviews checked it by comparing token values, not in a browser.
Repository-wide tests run only in CI.
Two model-switch cases in orchestrationEngine.integration.test.ts fail on main as well and are unchanged here.

Created with Claude Opus 5.5 and GPT-6.1 Sol in Claude Code.

🤖 Generated with Claude Code

Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Split payload bounding and retention by responsibility, preserve delimiters as source escapes, and retain tool updates with one completion lookup per identity. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

…ilities Keep the existing module entry points, share the session settlement rule, and extract projection test cases and fixtures by behavior. Remove attachment materialization effects that only return decoded attachments. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

…d modules Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The memory approval reuses the composer's surface constant, the docked approval and question panels share one shape constant, the approval dot uses bg-warning, attachment tiles use border-border, and the channel warning label uses text-warning-foreground. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

useBotThreadRuntime and useGroupThreadRuntime carried identical threadTitle, readFileAsDataUrl, and attachment-reading code. Both now import them from threadRuntimeAttachments.ts, which has focused tests. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…runtimes useThreadPendingUserInput owns the answer drafts, question index, and responding request ids both runtime hooks carried in duplicate. groupModelSelection moves to groupThreadModel.logic.ts. Bot and group send lifecycles stay separate. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…e logic tests Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…w out of BotThreadLanding useBotLandingTimeline owns routine run paging, receipt projection, delegations, and the merged timeline order. RoutineReceiptRow gets its own module. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…modules BotSettingsForm owns the draft, save bar, and sheets. BotIdentityFields and BotEngineFields render the Identity and Model & usage sections from the draft. Source-contract tests read all four modules, and the shallow render tests expand the section leaves through a shared test-support helper. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Move snapshot history and lookup coverage into independent suites with shared fixtures. Keep command, telemetry, transaction, and restart engine coverage separate. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

…utton variants Adds avatar-edit, avatar-tile, default-muted-disabled, ghost-destructive, and outline-pill-muted variants plus an avatar size to ui/button, carrying the classes the call sites passed. The bot settings avatar action and new-bot shape tiles become Buttons; aria-pressed drives the tile's chosen state. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…nto ui/sheet SheetHeader, SheetTitle, and SheetPanel gain a compact variant carrying the classes BotSideSheet passed, and EmptyTitle gains a body variant for its sans medium title. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The BotSettingsPage split moved BotModelPicker and TraitsPicker into BotEngineFields. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…lity out of BotChatMessageRows BotChatMessageRows keeps the memoized assistant and user rows and re-exports the moved symbols. BotMessageAttachments uses the exact scale values for its arbitrary sizes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Keep bot settings and compatibility coverage at the existing test entry point and extract ownership cases with reusable fixtures. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

…est fixtures Compose reactor services from owned state and focused delivery, ingestion, checkpoint, and recovery helpers. Decode typed fixture events at publication, subscribe before triggers, and await observations or receipts instead of polling. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

Run owned-path lint autofixes and formatting. The diff is empty with git diff --ignore-all-space --ignore-blank-lines; all changes are whitespace-only. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

… modules DesktopOnboarding.tsx keeps gating, draft recovery, and the durable handoff resume; the subscription and identity steps and the overlay surface move to their own files. desktopOnboarding.logic.ts and goalPlan.logic.ts split into draft, engine, handoff, data, and classification modules with re-exports at the old paths, and the translator moves to a leaf to break the import cycle. Also uses success tokens for the connected check, a Button onboarding size for the tall rounded actions, classes instead of inline pointer-events, and exact scale values for the overlay width and z-index. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

VoiceCallProvider keeps connection, microphone, and teardown ownership at VoiceCall.tsx, which re-exports the moved helpers, hooks, and views so its importers and module mocks are unchanged. The hang-up button selects a new icon-sm-round Button size instead of restyling the primitive. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The search field, filter bar, page header, and dialog/page directory layout move to PluginDirectoryLayout.tsx, the catalog constants to pluginDirectoryCatalog.ts, and the MCP server draft helpers to mcpServerDraft.logic.ts. PluginsDialog.tsx keeps the environment-bound state owner, the server editor, and both surfaces, and re-exports moved names. Dialog gains directory header/panel variants that carry the catalog padding, including the pt-5! override, and a divided header variant for the border-b rule, so plugin call sites stop restyling the primitives. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

PluginsCatalog.tsx keeps the directory composition and re-exports the cards (PluginCatalogCards.tsx), Composio results, custom and removed MCP server lists, and the label translator, each now in its own module. Plugin action buttons select new Button pill sizes, the broker badge a quiet Badge variant, and the category picker a dense Select size, carrying the exact classes the call sites used to pass. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…nu modules sidebarContext.tsx owns the provider, contexts, cookie persistence, and keyboard-facing toggle; sidebarRail.tsx owns resize capture and width clamping; sidebarMenu.tsx owns the group, menu, and submenu parts. sidebar.tsx keeps Sidebar, the trigger, inset, and layout slots, and its export list is unchanged, so no importer changes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…ed modules toastState.ts owns the shared manager instances and the visible-time registry so every provider reads the same ones. ToastBody.tsx holds the body rendering, ThreadToastVisibleAutoDismiss.tsx the route-scoped visibility and timeout accounting, and AnchoredToasts.tsx the anchored provider. toast.tsx keeps ToastProvider and Toasts and its existing exports. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

CI runs scripts/check-public-dependencies.ts before installing dependencies, so importing effect/Schema failed with ERR_MODULE_NOT_FOUND. Parse manifests with Node built-ins only and add a test that guards the import list. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The test-file override for no-module-mocking matched only .ts and .tsx, so CI lint rejected the vi.mock seams in apps/desktop/scripts/dev-electron.test.mjs. Also format ResponseLoadingState.test.tsx, which failed fmt:check. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…old labels Restoring main's English work-fold labels removed the translated messages from thread-turn-fold-row.tsx, but the coverage snapshot still counted them. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

greptile-apps · 2026-10-02T14:33:04Z

Confidence Score: 5/5

[Medium risk] Refactors module structure and adds lint tooling configuration.

The PR appears safe to merge based on this review.

Summary

The PR splits oversized modules, strengthens lint enforcement, replaces raw palette colors with theme tokens, and includes targeted behavior fixes. Since the previous review, the desktop annotation stylesheet generator was updated to scan the preview modules containing overlay classes.

_{Reviews (7) · Last reviewed commit: "fix(desktop): build annotation styles fr..."}

The engine commits a queued command even after its caller stops waiting, so an interrupted bootstrap could delete a thread whose first turn was accepted, or miss deleting a thread whose create had committed. Create and turn start now await their results uninterruptibly and record them before cleanup reads them. The latest user message lookup now runs as one ordered SQL query instead of loading every message, image reads gain multi-chunk coverage, and the usage loading comment matches the static skeleton. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Claude turns select the 1M context variant and grok-build resolves to grok-4.6 before reaching Mastra, so comparing switches against the raw model slug failed for both providers on main as well. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

leoisadev1 · 2026-10-02T14:45:18Z

@greptile apex

Bootstrap awaits its create and turn start uninterruptibly, so a command stuck in the engine queue could hold a cancelled request and server shutdown forever. Closing the engine now interrupts the in-flight and queued results, releases their turn admissions, and fails later dispatches instead of leaving them waiting. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Cleanup after a failed dispatch also ran when the client cancelled, but the engine commits a queued turn start after its caller leaves, so the message kept pointing at deleted files. Turn starts with pending uploads now await the engine outcome and delete claimed files only when the command receipt shows the turn was not accepted. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…connection modules Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The worker could be interrupted after taking a command but before it installed settlement and admission release, leaving the caller waiting. Take uninterruptibly and install both before processing becomes interruptible. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…lled engine The dispatch ran uninterruptibly so its receipt could decide upload cleanup, which made a cancelled caller wait behind any stalled command. Run the dispatch detached with its cleanup attached and join it interruptibly. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…slop-cleanup # Conflicts: # apps/web/src/components/sidebar/ExperimentalSidebar.tsx

The PickPreload split moved overlay class names into sibling modules, so the generator dropped them on rebuild. Scan all preview sources instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

leoisadev1 and others added 30 commits October 1, 2026 01:33

refactor(server): split provider protocols and voice test suites

fad1699

Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

refactor(server): use typed predicates for tagged value guards

56331e1

Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

style(server): add readable spacing to provider layers

261ad52

Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

merge: swarm/server-core-root cleanup

c8cb646

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

refactor(server): retain preview failure causes

f568df4

Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

style(server): apply readable spacing to core modules

b9125b3

Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

merge: swarm/desktop-scripts cleanup

47df0b0

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

fix(server): preserve Unicode in bounded activity payloads

47a53d5

Split payload bounding and retention by responsibility, preserve delimiters as source escapes, and retain tool updates with one completion lookup per identity. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

refactor(web): split BotPromptComposer into logic, draft, and keyboar…

883830e

…d modules Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

test(web): split bot chat selection and retention cases out of runtim…

87ac734

…e logic tests Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

refactor(server): split snapshot and engine tests by behavior

32e7520

Move snapshot history and lookup coverage into independent suites with shared fixtures. Keep command, telemetry, transaction, and restart engine coverage separate. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

test(web): read the bot settings model controls from BotEngineFields

3e5a54b

The BotSettingsPage split moved BotModelPicker and TraitsPicker into BotEngineFields. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

refactor(server): separate bot ownership persistence tests

79f6efc

Keep bot settings and compatibility coverage at the existing test entry point and extract ownership cases with reusable fixtures. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

style(server): apply readable orchestration spacing

1da39d2

Run owned-path lint autofixes and formatting. The diff is empty with git diff --ignore-all-space --ignore-blank-lines; all changes are whitespace-only. Co-Authored-By: GPT-6.1 Sol <noreply@openai.com>

vercel Bot deployed to Preview October 2, 2026 14:13 View deployment

github-actions Bot added type:provider Agent provider contribution. area:directory Plugin directory and its user interface. area:connectors Plugin and MCP connector runtime. labels Oct 2, 2026

leoisadev1 and others added 3 commits October 2, 2026 10:15

greptile-apps Bot reviewed Oct 2, 2026

View reviewed changes

leoisadev1 and others added 2 commits October 2, 2026 10:40

greptile-apps Bot reviewed Oct 2, 2026

View reviewed changes

Comment thread apps/server/src/wsOrchestrationCommands.ts

leoisadev1 and others added 5 commits October 2, 2026 11:08

refactor(web): split the remaining oversized terminal, composer, and …

5647fe4

…connection modules Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

refactor(server): split the remaining oversized provider modules

8bb82c5

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

vercel Bot deployed to Preview October 2, 2026 16:05 View deployment

greptile-apps Bot reviewed Oct 2, 2026

View reviewed changes

Comment thread apps/server/src/orchestration/Normalizer.ts

leoisadev1 and others added 3 commits October 2, 2026 12:12

Merge remote-tracking branch 'origin/main' into t3code/codebase-anti-…

5dc6334

…slop-cleanup # Conflicts: # apps/web/src/components/sidebar/ExperimentalSidebar.tsx

fix(desktop): build annotation styles from every preview module

51a98e9

The PickPreload split moved overlay class names into sibling modules, so the generator dropped them on rebuild. Scan all preview sources instead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

leoisadev1 merged commit 66ba44d into main Oct 2, 2026
11 checks passed

leoisadev1 deleted the t3code/codebase-anti-slop-cleanup branch October 2, 2026 17:19

github-actions Bot mentioned this pull request Oct 2, 2026

chore(release): version packages #345

Open

Navigation Menu

Repository navigation

refactor: split oversized modules and enforce anti-slop lint - #343

refactor: split oversized modules and enforce anti-slop lint#343
leoisadev1 merged 392 commits into
mainfrom
t3code/codebase-anti-slop-cleanup

leoisadev1 commented Oct 2, 2026 •

edited by devin-ai-integration Bot

Loading

Uh oh!

greptile-apps Bot commented Oct 2, 2026 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

leoisadev1 commented Oct 2, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

leoisadev1 commented Oct 2, 2026 • edited by devin-ai-integration Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

What changed

Behavior changes included

Review

Verification

Not covered

Uh oh!

greptile-apps Bot commented Oct 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Confidence Score: 5/5

Summary

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

leoisadev1 commented Oct 2, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

leoisadev1 commented Oct 2, 2026 •

edited by devin-ai-integration Bot

Loading

greptile-apps Bot commented Oct 2, 2026 •

edited

Loading