fix(vscode): restore chat turn spacing broken by virtualizer#9141
Merged
marius-kilocode merged 2 commits intomainfrom Apr 20, 2026
Merged
fix(vscode): restore chat turn spacing broken by virtualizer#9141marius-kilocode merged 2 commits intomainfrom
marius-kilocode merged 2 commits intomainfrom
Conversation
The prior spacing fix (#9025) relied on flex `gap` on `.message-list-content`, but #8911 introduced virtua's `Virtualizer` which positions items absolutely based on measured box size — so `gap` no longer applies between turns. The missing spacing was most visible when the last assistant part was a sub-agent's expanded task tool (two bordered boxes with 0px between them). Bake the 12px into each turn's own padding so virtua measures it as part of the item height.
Contributor
Code Review SummaryStatus: 1 Issue Found | Recommendation: Address before merge Overview
Issue Details (click to expand)WARNING
Fix these issues in Kilo Cloud Other Observations (not in diff)No additional issues outside the diff. Files Reviewed (5 files)
Reviewed by gpt-5.4-20260305 · 409,010 tokens |
![kilo-code-bot[bot]](https://avatars.githubusercontent.com/in/2193792?s=60&v=4){kind=small}
alex-alecu
approved these changes
Apr 20, 2026
Ghenghis
added a commit
to AiDave71/kilocode
that referenced
this pull request
Apr 27, 2026
* docs(kilo-docs): document actual compaction defaults and triggers Replaces the vague 'usableWindow' language with the real formula (context_window - ~20K reserve), documents the /compact slash command and task-header button, adds env-var overrides, and lists the correct default values. * docs(kilo-docs): drop defensive framing in compaction trigger section * fix: make all V4 tabs auto-populate — resolves 'Testing…', 'Detecting…', empty SSH, offline Ollama Root cause: OnboardingDiscoveryService ran in background but its results were orphaned — never pushed to SSHService, never triggered initial RoutingService health checks, tabs showed empty state forever. Fixes: 1. RoutingService: initial health check 1s after startup so Ollama/LM Studio show 'healthy' immediately if running (was waiting 60s for first interval) 2. KiloProvider requestRoutingState: kicks background health re-check on local providers when tab opens (non-blocking, streams results) 3. KiloProvider trainingGetJobs: auto-detects GPUs on tab open if cache empty 4. KiloProvider trainingDetectGPU: try/catch + always posts trainingGPUDetected (even on error) so tab never stays stuck on 'Detecting…' 5. KiloProvider requestSSHProfiles: lazy-imports ~/.ssh/config when profile list is empty — tab auto-populates from existing SSH config 6. extension.ts: discovery now calls sshService.importFromSSHConfig() after runFullDiscovery completes, so the first-run UX has profiles already imported 7. extension.ts: broadcasts discoveryComplete event so any open tabs refresh 8. RoutingTab: 15-second safety timeout on 'Testing…' state so the button never gets permanently stuck if backend hangs (SiliconFlow network issue) User-visible result: - Ollama shows 'healthy' automatically when running, no manual test needed - SiliconFlow 'Testing…' always resolves within 15s worst case - GPU auto-detects on tab open, 'Detecting…' always clears - SSH tab shows ~/.ssh/config hosts without manual import Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * docs(kilo-docs): rewrite trigger rules in prose * feat: Auto-Population Engine — complete discovery + secure profile + wizard + health recovery Addresses all 12 gaps the user identified: tabs waiting for manual input instead of being fed by detected data. Builds the filler system. NEW SERVICES: • SecureProfileService — unified secret/profile manager with strict split: - context.secrets → API keys, SSH passwords, tokens (encrypted) - globalState → provider choices, role matrix, voice prefs (cross-workspace) - workspaceState → project settings, discovery cache - Masked key display (never exposes real values to UI) - Legacy migration from old KV store • EnvironmentProbeService — ultra-fast sync probes (<100ms total): - Platform, arch, CPU, RAM, disk - File presence: ~/.ssh/config, known_hosts, .kilo/hermes.json, .kilo/shiba.json - Workspace folder, git repo detection - Baseline snapshot drives wizard decisions • VPSInventoryProbe — safe read-only SSH commands to auto-collect: - hostname, distro, kernel, uptime, CPU/RAM/disk - Docker, containers, running services, nginx/caddy, public IP - 17 parallel probes with 3s timeout each, fault-tolerant • HealthRecoveryService — CLI backend auto-recovery: - 30s monitor loop, exponential backoff [1s/5s/15s/60s/300s] - Status bar indicator with themed icons (healthy/degraded/disconnected) - kilo-code.v4.restartCliBackend command - Diagnostic report for About page ENHANCED SERVICES: • OnboardingDiscoveryService — 3 new probes added: - probeHermes() GETs /health on endpoint from .kilo/hermes.json (default :7001) - probeShiba() GETs /health, extracts connectedAgents list (default :7002) - probeZeroClaw() sets defaultScope=workspace path (default :7003) • MemoryService — autoConnect() on startup (500ms delay): - Searches workspace + home for .kilo/hermes.json and .kilo/shiba.json - Probes health endpoint with 2s timeout - Auto-transitions to "connected" state if reachable - Never clobbers local store on remote failure • ZeroClawService — getDefaultTaskContext() for tab bootstrap: - Pre-fills projectPath from current workspace - Default workspaceScope, riskLevel=low, networkPolicy=none - 3 pre-seeded templates (format, test, typecheck) NEW UI: • OnboardingWizard.tsx — 5-step guided setup: - Step 1: Discovery (auto-runs on mount) - Step 2: Review results with Accept/Edit checkboxes - Step 3: Secrets input (only for enabled cloud providers) - Step 4: Validation with live test results - Step 5: Completion summary • Registered command: kilo-code.v4.runOnboardingWizard MESSAGE PROTOCOL: • Added types: requestDiscoveryResult, triggerDiscovery, markOnboardingComplete, resetOnboarding, discoveryComplete, discoveryError, onboardingCompleted, onboardingReset, zeroClawContext • KiloProvider handlers for all new messages • triggerDiscovery now broadcasts discoveryComplete for tab auto-refresh DOCS: • docs/master-roadmap.md — comprehensive roadmap covering all 12 gaps with phase-by-phase plan, data models, E2E test matrix, priority order Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * docs(kilo-docs): clarify reserved buffer behavior per model type compaction.reserved only applies when a model declares a separate input limit. For models with just a single context window, the reserve is derived from the output cap instead. * refactor(jetbrains): use JBHtmlPane with editor-aware styling in MdView - Replace plain JEditorPane + manual CSS with JBHtmlPane, the IntelliJ platform's flagship HTML component - Font and colours now default to the global editor colour scheme via JBHtmlPane's built-in EditorCssFontResolver and colorSchemeProvider - Code font defaults to _EditorFontNoLigatures_ placeholder resolved by EditorCssFontResolver at render time - All style properties become optional overrides applied via a customStyleSheetProvider; the override sheet is empty until a property is set, so editor defaults always win unless explicitly overridden - Add resetStyles() to revert all overrides back to editor defaults - Transparency (opaque=false) handled by Swing isOpaque + transparent CSS body rule; no manual background injection when transparent - Update tests to use overrideSheet() and component state assertions * feat(vscode): support session artifact events and seed status on load in JetBrains plugin - Forward session.compacted, session.diff, todo.updated to SessionModel instead of ignoring them - Add DiffUpdated, TodosUpdated, Compacted model events and matching state fields - Seed busy/retry/offline state from KiloSessionService.statuses on existing-session load - Fix null sessionID filter so global session.error events reach the frontend - Add SessionArtifactsTest and extend SessionModelTest and SessionRecoveryTest * chore: remove plan files from branch * feat(sessions): add session_status ingest message for status tracking * docs(cli): require final summary before local review suggestion * docs(kilo-docs): add agent manager workflows guide (Kilo-Org#9148) * docs(kilo-docs): add agent manager workflows guide * docs(kilo-docs): use 'parent branch' instead of 'main' for worktree merges * fix(vscode): restore chat turn spacing broken by virtualizer (Kilo-Org#9141) The prior spacing fix (Kilo-Org#9025) relied on flex `gap` on `.message-list-content`, but Kilo-Org#8911 introduced virtua's `Virtualizer` which positions items absolutely based on measured box size — so `gap` no longer applies between turns. The missing spacing was most visible when the last assistant part was a sub-agent's expanded task tool (two bordered boxes with 0px between them). Bake the 12px into each turn's own padding so virtua measures it as part of the item height. * fix(vscode): hide duplicate preview text while editing custom question answer (Kilo-Org#9129) The question tool's custom answer option showed the user's input twice — once as a preview description inside the option button, and again in the input field below. Hide the preview while the input form is open so only the live input is visible. * fix: re-prevent plan file corruption for OpenAI models by preserving write tool (Kilo-Org#9188) * fix(vscode): keep queued messages pinned at the bottom (Kilo-Org#9195) * fix(vscode): keep queued messages pinned at the bottom Queued follow-up user messages were being rendered inline inside the conversation history, so tool output for an earlier turn could appear under a later queued message. Render queued turns in a separate stack beneath the virtualized history and group assistant messages by parentID so late-arriving tool output stays attached to its turn. * chore: update kilo-vscode visual regression baselines --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * fix(vscode): preserve older messages when paginating virtualized chat (Kilo-Org#9194) The fill() helper that repairs pages starting with an assistant turn kept only the user-turn suffix of each older page, so earlier messages fetched mid-stitch were dropped while the pagination cursor still advanced past them. That made the top of the virtualized history land after the real first message. Keep the full older page when stitching partial turns so fill() terminates naturally once a user message reaches the top, and add a regression test covering multi-page fills. * fix(cli): release session busy state while suggest tool waits for user The suggest tool blocks on a promise that only resolves when the user accepts or dismisses a suggestion. While blocked, the runner stayed in Running state and the session status remained busy — if the suggestion was never answered (e.g. VS Code was closed), the session was stuck forever and follow-up prompts appeared queued. Mark the session idle while awaiting the user's response; the loop will set it back to busy when the suggestion resolves and processing continues. Fixes Kilo-Org#9150 * feat: use model.api.id rather than model.id as the id * fix(cli): restore busy state when suggestion is accepted * feat(gateway): add Alibaba as supported AI SDK provider * chore: update kilo-vscode visual regression baselines * fix(cli): time out status derive to prevent stuck chain entries If deriveAndSyncStatus hangs, the per-session promise chain would never drain and map entries would accumulate. Wrap the derive with a 3s timeout so stuck work fails fast, is logged, and the existing cleanup runs. * refactor: kilo compat for v1.4.6 * chore: update nix node_modules hashes * fix(vscode): restore review diff in repos with no remote * fix: regenerate openapi.json cleanly without shell banner * fix(vscode): honor explicit stale base in resolveBase * chore(cli): sort dependencies alphabetically in package.json * chore: update kilo-vscode visual regression baselines * chore: untrack generated .kilo and .kilocode lockfiles (Kilo-Org#9200) * chore: untrack generated .kilo and .kilocode lockfiles `kilo serve` regenerates `.kilo/package-lock.json` and `.kilocode/package-lock.json` (plus `node_modules/`) on every run -- see `Config.install` in `packages/opencode/src/config/config.ts:1406` -- but they were accidentally committed in the upstream alignment merge `7fb55e45f1`, so every run dirtied the working tree. Untrack them and add explicit root `.gitignore` rules so they stay ignored even if the per-directory `.gitignore` the CLI writes is removed. `.kilo/run-script` stays tracked -- it's a project dev helper, not a generated artifact. * chore: rely on generated config dir ignores The CLI already writes .gitignore files inside .kilo and .kilocode before installing config dependencies. Keep the PR focused on untracking the accidentally committed lockfiles instead of duplicating those generated ignore rules at the repo root. * fix(cli): coalesce status sync retries * chore: merge main and reduce diff to only alibaba provider changes * chore: update kilo-vscode visual regression baselines * refactor(snapshot): inline kilo snapshot helpers and remove dedicated module Move worktree-scoped gitdir construction, ACP guard logic, and diff caching directly into the core snapshot service, eliminating the separate kilocode/snapshot module. This reduces indirection and keeps all snapshot concerns co-located in a single file. * fix(kilo-docs): update ChatGPT pricing link (Kilo-Org#9221) * test(cli): pin suggest tool session-status behavior to prevent stuck-busy regression Lock in that a session with an open suggestion reports as idle (so reopening VS Code or switching worktrees does not show it stuck/running), that accepting flips it back to busy without an idle flash, and that a dismissed suggestion leaves it idle for the run loop to resume cleanly. * refactor: address connection * fix(vscode): restore review revert-file action in repos with no remote Clicking 'revert file' in the Agent Manager review panel silently did nothing when the repo had no remote and no tracking branch. The diff view worked because it defensively searched for a local candidate branch, but the revert path ran merge-base against the placeholder HEAD and turned into a no-op for committed feature changes. * chore: update kilo-vscode visual regression baselines * refactor(cli): narrow snapshot-diff freeze fix to caps-only Revert the worker pool, dispatcher, flags, and related scaffolding added on top of the initial cap guard. Post-Kilo-Org#9046, the only remaining CLI callers of Vcs.diff are one-shot review opens, so the worker offload is no longer needed. The freeze repro is eliminated by the input caps in DiffEngine.shouldSkip alone. Shrinks the fork diff from 28 files (+1063) to 6 files (+199) and reduces shared opencode files with kilocode_change markers to 2. * refactor: kilo compat for v1.4.7 * fix(vscode,cli): make plan_exit "Continue here" work again Restore custom: false on the plan follow-up question — the "Type your own answer" row was redundant because the main prompt already routes typed text as a question reply. Auto-submit single-question single-select option picks in the VS Code QuestionDock so the button behaves like the TUI instead of silently waiting for a second Submit click. * chore: update nix node_modules hashes * fix(vscode): route worktree sessions to correct tab (Kilo-Org#9208) Ensure Agent Manager pushes worktree session state before generic sessionCreated messages and reconciles local tabs against worktree ownership so forked, continued, and newly-created worktree sessions do not appear in Local. Fixes Kilo-Org#8983. * docs: fix agentskills.io link to avoid 308 redirect (Kilo-Org#9233) * Add message-level Agent Manager session forks (Kilo-Org#9207) * feat(agent-manager): fork sessions from user messages * fix(agent-manager): handle unassigned message forks * test(vscode): register chat tool overrides * fix(vscode): make tool overrides idempotent * chore: update kilo-vscode visual regression baselines * test(vscode): revert chat story override registration * chore: update kilo-vscode visual regression baselines --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * fix(vscode,ui): keep user scroll position while session is busy (Kilo-Org#9236) * fix(vscode,ui): keep user scroll position while session is busy Virtua's measurement-driven resize events race ahead of the debounced user-scroll detection in createAutoScroll, snapping the viewport back to the bottom while the user is mid-gesture. The QuestionDock's focus call on mount also triggers the browser's focus-into-view behavior, which yanks the view down whenever the user has scrolled up. Fixes Kilo-Org#9198 * fix(ui): guard recentlyInteracted against initial lastInteraction=0 * fix: persist custom provider model and variant deletions (Kilo-Org#9239) * fix: persist custom provider model and variant deletions The CLI config.update endpoint deep-merges its payload with existing global config, so removing a model or reasoning variant from a custom provider in the UI had no effect on disk. Save payloads now emit null sentinels for removed IDs, and the Provider schema accepts nullable record values so stripNulls can delete them during the merge. Closes Kilo-Org#9186 * chore: update kilo-vscode visual regression baselines --------- Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> * fix(vscode): extract and test custom provider variant name validation Extract validateCustomProvider to a testable utility (CustomProviderValidation.ts), add unit tests that cover the empty-variant-name bug, and use reconcile() in setErrors so SolidJS properly propagates nested variant error updates to the UI. Fixes Kilo-Org#9240 * chore: add changeset for variant validation fix * feat(i18n): translate plan_exit follow-up options in the VS Code sidebar Add optional `labelKey`/`descriptionKey` to Question.Option and `questionKey`/`headerKey` to Question.Info, annotated with `kilocode_change` markers. Populate these keys from the plan follow-up so the "Ready to implement?" question, "Start new session", and "Continue here" buttons render in the sidebar language while the canonical English labels remain on the reply wire (unchanged server-side matching). * docs(kilo-docs): document per-agent model memory and reset button Add missing behavior to model-selection.md VSCode tab: 'last picked per agent' level in precedence chain, and a note that the model selector remembers picks across sessions with reset button to restore config. Add the same note to custom-modes.md for both the VSCode and CLI tabs under the model property reference (previously only documented in the VSCode Legacy tab as 'Sticky Models'). * fix(vscode): keep plan follow-up options reactive to language changes Return accessors from translateOption so SolidJS tracks the language signal through <For>. Without this, a mid-dialog language switch leaves option labels frozen in the previous locale until the dock remounts. Also add inline notes explaining the wider auto-submit behaviour change in pickOutcome (covers every single-question single-select caller, not just the plan follow-up) and the tr() helper's coupling to language.t's miss-key echo contract so a future refactor of language.t notices the dependency. * chore: update kilo-vscode visual regression baselines * feat(vscode): support sidebar message forks (Kilo-Org#9244) * feat(vscode): support sidebar message forks * fix(vscode): type sidebar fork message * fix(vscode,cli): make plan_exit "Continue here" work again Restore custom: false on the plan follow-up question — the "Type your own answer" row was redundant because the main prompt already routes typed text as a question reply. Auto-submit single-question single-select option picks in the VS Code QuestionDock so the button behaves like the TUI instead of silently waiting for a second Submit click. * fix(cli): make plan follow-up "Continue here" continue the loop The prompt queue's scope() hid the user message injected by PlanFollowup.inject() because its ID was newer than the queue target, so the loop saw the same plan_exit messages and re-asked the question in an infinite cycle. Retarget the queue after inject so the new message is visible on the next iteration. * fix(cli): restore "Type your own answer" on plan follow-up in TUI When the plan agent pauses for the follow-up question, CLI users had no way to send a free-text reply: the main prompt input is hidden while a blocking question is active, and the "Type your own answer" row was being forced off to avoid duplicating the VS Code prompt input. CLI now shows the custom-answer row again, while VS Code keeps it hidden. * fix(vscode): keep custom single-question on tab * fix(vscode): revalidate sidebar fork status (Kilo-Org#9247) * fix(vscode): revalidate sidebar fork status * fix(vscode): handle missing client during fork status * fix(cli): keep older queued prompts hidden on retarget * test(vscode): update pickOutcome expectation for single+custom to stay * refactor(vscode): reduce validateCustomProvider complexity by extracting helpers Extract checkModel, checkVariant, checkHeader, checkProviderID, serializeModel, serializeVariant, and resolveEnv from validateCustomProvider to bring it under the ESLint complexity limit of 20. Also remove the now-unnecessary complexity exception for CustomProviderDialog.tsx. * chore: merge latest main * chore: restore screenshot to main version * refactor(vscode): drop FIM templates for unsupported autocomplete models Autocomplete only exposes Codestral (default) and Mercury Edit via the `kilo-code.new.autocomplete.model` setting, so getTemplateForModel's branches for stable-code, qwen-coder, seed-coder, codegemma, codellama, deepseek, codegeex and the hole-filler fallback were unreachable. * refactor(vscode): drop unreachable postprocessing branches postprocessCompletion contained model-specific branches for qwen3, granite, gemini and gemma — models that are never selected. Keep the Codestral and Mercury branches (both are user-selectable via kilo-code.new.autocomplete.model) and also collapse Mercury's two adjacent branches into one block. * refactor(vscode): drop chat-completion fallback in chat textarea autocomplete Both supported autocomplete models (Codestral, Mercury) expose FIM, and AutocompleteModel.supportsFim() unconditionally returns true. The non-FIM branch in ChatTextAreaAutocomplete together with the getChatSystemPrompt / getChatUserPrompt helpers was never reached. Also removes the accompanying orphaned spec that imported paths that no longer exist (core/config/ProviderSettingsManager, api/transform/stream). * refactor(vscode): drop unused supportsFim / generateResponse from AutocompleteModel supportsFim() unconditionally returned true and generateResponse() only ever threw — leftovers from when the model interface tried to abstract over FIM-capable vs chat-only providers. Both supported autocomplete models (Codestral, Mercury) expose FIM, so neither abstraction is needed. * docs(jetbrains): document session component architecture and testing conventions in AGENTS.md * fix(cli): add alibaba to kiloProviderOptions thinking translation * feat(jetbrains): add SessionUi with turn-aware transcript, renderers, and DSL docks Rewrites the JetBrains session UI from scratch: Model: Turn class + TurnAdded/Updated/Removed events in SessionModel, with regroup() keeping turn structure in sync after every message mutation. Session views: PartView hierarchy (Text/Reasoning/Tool/Compaction/Generic) backed by MdView for markdown; MessageView and TurnView using a custom SessionLayout that pre-sets child widths so JBHtmlPane reflows correctly. Session panel: turn-aware SessionPanel with three lookup indexes (turnId, msgId→TurnView, msgId→MessageView); ContentDelta uses full content replace to avoid doubling the first streamed token. Dock panels: QuestionPanel and PermissionPanel built with Kotlin UI DSL, rebuilt on each show() call to match the current request. SessionUi: rewritten as a thin composition root; south stack uses BoxLayout so hidden docks collapse to zero height. Tests: 100+ new tests across TurnGroupingTest, TextViewTest, ToolViewTest, TurnViewTest, SessionLayoutTest, SessionPanelTest, SessionUiUpdateTest. * fix(cli): restore model descriptions in expanded model picker patchModelsDevModel was not including the options field in its return value, so the Object.assign in fromModelsDevModel left options as {} and discarded the description fetched from the Kilo Gateway API. * feat(jetbrains): add debug logging for session events and state changes Log all CLI SSE events received and HTTP requests sent (with session ID), event routing decisions in the RPC filter, and every model/controller/view event on the frontend to make stuck or misrepresented sessions diagnosable. * feat(jetbrains): add progress indicator and suppress step-start/step-finish parts Filter step-start and step-finish part types at the model level so they are never stored or rendered. Add ProgressPanel inside the scroll pane, anchored as the last child of SessionPanel, showing an animated spinner with Busy.text while the session is working. * chore: add kilocode_change annotations to ling model changes * core: stop TUI freeze when viewing diffs of huge files Files over a few thousand lines no longer hang the session for minutes during summary generation or file view. Diffs for files of any size now render with full content in the TUI, VS Code sidebar, and web UI — previously they either froze the event loop or were reported with empty patch text. * fix(cli): render suggest above an active input prompt Emit suggestions as non-blocking so the main CLI input stays focused and submittable while the picker is visible, matching the VS Code extension. Submitting a new message auto-dismisses the pending suggestion. * fix(cli): start-new-session tab on slow plan handover Create the follow-up session before running the handover LLM call so the extension's pendingFollowup SSE gate fires inside its 30s TTL. The handover and todos are now injected into the already-live session after it resolves, instead of blocking session creation. * fix: use startsWith for ling model detection and add ling to openapi prompt enum * core: drop the JavaScript diff fallback that's no longer needed The git-based diff path has been reliable across all tests and in production since the TUI-freeze fix shipped. Removing the JS Myers fallback eliminates ~220 lines of belt-and-suspenders code (a size-cap guard, a git cat-file blob loader, and a per-file git show fallback). If git ever fails now, callers emit an empty patch string; additions/deletions from git --numstat still come through unchanged. * chore: move visual baselines into docs assets * chore: retain old visual baseline lfs paths * fix(vscode): settle canceled autocomplete debounce * refactor: replace broad ling startsWith with isLing helper excluding kling/bling/spelling * fix(vscode): prewarm autocomplete backend * refactor: share isLing helper via kilocode/model-match, add multilingual exclusion * fix(gateway): timeout autocomplete fim streams * fix(vscode): reset autocomplete state on workspace change * fix(gateway): keep fim guard active for errors * docs: reference generated screenshots * chore: add changeset for autocomplete backend prewarm * fix: keep visual screenshot suffixes * core: preserve upstream Myers diff code to keep upstream merges clean The previous commit deleted ~150 lines of upstream OpenCode diff code because the git-based path always runs first. Restoring those lines as dead code (behind an early-return kilocode_change block) keeps our diff from upstream minimal, so future OpenCode syncs to the hot diffFull code path don't conflict. No runtime behavior change — the git-based DiffFull path still short-circuits before the restored Myers loop executes. * chore: update kilo-vscode visual regression baselines * Apply suggestion from @chrarnoldus * wip * fix(cli): inject plan message immediately, append handover in-place Show the plan text right away in the new session tab without waiting for the handover LLM call. The message part is created with plan+todos as soon as the session opens; the handover section is upserted onto the same part once the slow LLM call resolves. * fix(cli): align non-blocking suggest picker behavior Remove the synthetic Dismiss row and let Esc dismiss suggestions so the picker matches its visible controls while the main input stays active. * fix(cli): keep follow-up session busy during handover Mark new plan follow-up sessions busy as soon as the tab opens so the UI shows work is still in progress while the handover summary is generated. Clear the temporary busy state if that pre-loop handover phase is aborted or fails before the normal prompt loop takes over. * fix: clean up docs dialog story * feat(cli): add dev:local script with isolated XDG dirs and local endpoints * fix(vscode): bound autocomplete ignore cache * chore: update nix node_modules hashes * fix(cli): guard follow-up loop start * fix(vscode): patch stream-chat@9.38.0 to fix broken ws type declarations * fix(cli): clear pending on errors * chore: update nix node_modules hashes * chore(publish): defer release push and implement rebase retry loop Move the git push operation to the end of the publishing process to minimize race conditions. Replace the cherry-pick logic with a rebase onto origin/main and a retry loop to ensure the release commit is pushed successfully even if concurrent merges occur. * fix(cli): replace cherry-pick with rebase + retry in publish push Keep the original order (push before publish) to preserve the safe failure mode, but replace the cherry-pick with rebase and add a 3-attempt retry loop to shrink the race window. * release: v7.2.17 * ci: use blacksmith for smoke tests * chore: Delete plan * build(deps): consolidate @opencode-ai/util and @opencode-ai/server into @opencode-ai/shared Remove the standalone `packages/util` and `packages/server` workspace packages, migrating all imports across app, kilo-ui, opencode, and ui to use `@opencode-ai/shared/util/*` paths instead. Also fix the VITE_KILO_CHANNEL env variable name in the type declaration and add a null guard in the titlebar channel badge rendering. * logging * fix(vscode): restore explicit submit for question dock * fix(vscode): restore agent on question dismiss * feat(jetbrains): improve session logging and sandbox diagnostics * docs(jetbrains): fix split mode template link * fix(jetbrains): harden question and markdown views * refactor(opencode): update internal import paths to use barrel re-exports Migrate all deep import paths across source and test files to reference barrel index modules instead of direct file paths. This covers config, provider, storage, util, tool, lsp, project, and installation modules. Also moves provider promise helpers out of the namespace block, switches Clipboard to namespace import style, fixes a duplicate Filesystem import, removes the deleted paste-summary test, and corrects snapshot cache typing and indentation. * docs: restore custom provider screenshots * fix: restore settings story lifecycle import * docs: align task header screenshot story * chore: update kilo-vscode visual regression baselines * fix(opencode): replace direct Env access with effect-based resolution and update test infrastructure Remove synchronous Env.get/Env.all calls in provider, tool registry, and model-cache modules, replacing them with effect-yielded env lookups or direct process.env reads where appropriate. Thread resolved env through patchCustomLoaderResult and kiloCustomLoaders dependency type. Drop unused paste-summary import from prompt component, fix createResource action call in session list dialog, add bell toggle KV signal, and rewrite config-gitignore test to use proper Effect layers instead of mocked AppRuntime. Update remaining import paths and test module mocks to align with barrel re-exports. * docs: document generated screenshot guidance * build(deps): upgrade OpenTelemetry packages to v2 and sort dependencies Bump @opentelemetry/core, sdk-trace-base, sdk-trace-node, and resources from 1.30.x to 2.6.1 and semantic-conventions from 1.28.0 to 1.40.0 in kilo-telemetry. Migrate to the new v2 API surface: replace `new Resource()` with `resourceFromAttributes()`, rename `parentSpanId` to `parentSpanContext`, and rename `instrumentationLibrary` to `instrumentationScope` in span types and tests. Alphabetically sort dependency entries in the opencode package.json. * fix(kilo-docs): keep docsearch links on previews * fix(cli): cap per-turn compaction attempts to stop infinite busy loop When every compaction round still overflowed the model context, SessionPrompt.runLoop would keep calling compaction forever and report the turn as completed. Cap attempts at three per turn and surface exhaustion as a ContextOverflowError on the assistant message with TurnClose reason=error. * tui(cli): render non-blocking suggestions inline in the conversation Moves the 'Run review?' suggestion picker out of the footer bar above the prompt and into the conversation itself, at the position of the suggest tool call. Frees up vertical space for reading while scrolling and matches where the VS Code extension shows the same picker. Clicking an option still accepts, digit keys 1/2 fast-accept when the prompt isn't focused, and Esc dismisses. Blocking suggestions keep the above-prompt overlay. * refactor(sdk): reorder generated types and fix route chaining in instance handler Alphabetically sort import statements in the SDK codegen output and reposition type declarations (EventInstallationUpdated, EventSessionTurnOpen, SessionStatus, Todo, etc.) to match updated code generation ordering. In the server instance router, assign the chained route builder to a `full` variable before passing it to `registerKiloRoutes` for clarity. * fix(test): correct config module import to namespace import Change named import to namespace import for the Config module in custom-provider-delete test to align with how the module exports its members. * docs(source-links): update file path references for model-id and tui-migrate modules Correct source comment annotations to reflect relocated modules: - config.ts → config/model-id.ts for model schema link - config/tui-migrate.ts → cli/cmd/tui/config/tui-migrate.ts for tui.json link * fix(proxy): await async isSyncing check in server proxy * refactor(test): replace direct Config.get assignment with spyOn mocking Swap manual save-and-restore of Config.get for bun:test spyOn/mock.restore across session-list and recall test suites, ensuring proper mock teardown and consistent namespace imports. * fix(provider): update adaptive efforts, gateway support, and test corrections Extract inline Opus 4.7 adaptive effort logic into anthropicAdaptiveEfforts helper, extend smallOptions to recognize Kilo Gateway alongside OpenRouter, fix LSP test file path from lsp/index.ts to lsp/lsp.ts, and update xhigh variant assertion to include display: "summarized". * fix(cli): handle EADDRINUSE race in mcp oauth callback server Parallel bun test subprocesses can race between isPortInUse() and listen(), causing the loser to crash with EADDRINUSE. Treat that error as 'another instance owns the port' (same semantics as the isPortInUse branch above) so the auto-connect test stops flaking in CI. * refactor(gateway): simplify FIM timeout using AbortSignal.any Replace custom timeout() and stream() helpers with platform AbortSignal.any + AbortSignal.timeout (available since Node 20.3). Eliminates ~50 lines of manual signal management and stream wrapping. * tui(cli): make suggest renderer reactive so the inline bar actually appears Previously, the non-blocking inline SuggestBar never showed up because the Suggest component used an early-return if/else that ran once at mount (when no pending request existed yet) and never re-evaluated when the suggestion arrived. Users saw 'Suggesting next step...' forever with no way to accept or dismiss, which felt like the suggestion was blocking the session again. Switching to <Switch>/<Match> so the branch updates when the request lands makes the bar appear and be interactive. * release: v7.2.18 * tui(cli): redesign inline suggest bar to match VS Code single-row layout Replaces the number-prefixed question-picker block with a full-width tinted row: icon + suggestion text on the left, clickable action buttons on the right. Removes keyboard shortcuts and the esc-dismiss hint -- dismissal already happens automatically server-side when the user sends a new prompt. * release: v7.2.19 * fix(vscode): retry Open VSX publish on transient failures * chore(cli): annotate kilocode sdk import * release: v7.2.20 * fix(jetbrains): use shared root logger in FileLog to prevent duplicate handlers * refactor(tui,provider): raise paste-summary thresholds, improve copilot auth error, and clean up comments Increase paste-summary trigger from 3 lines/150 chars to 5 lines/800 chars to reduce unnecessary summarization on small pastes. Add a branded reauthentication hint when GitHub Copilot returns 403. Guard against empty agent list during org switch in the model auto-update effect and reformat multi-line ternary for readability. Relocate inline comment on mercury exclusion in reasoning effort variants. * style(imports): consolidate deep imports to use barrel re-exports Replace granular submodule imports with their parent barrel index across snapshot diff source and related test files. This covers util/log, util/filesystem, config/config, provider/provider, tool/registry, tool/truncate, and the filesystem shared package. * chore: update kilo-vscode visual regression baselines * docs: remove #1 on OpenRouter claim from taglines * chore: update nix node_modules hashes * chore(cli): mark blockingSuggestion as kilocode change * feat(vscode): xterm.js terminal tabs in agent manager (Kilo-Org#9268) Click the chevron next to the + tab button and pick 'New Terminal' (or hit Cmd+Shift+T / Ctrl+Shift+T) to spawn a real shell inside the selected worktree or Local directory. Each terminal runs via the Kilo CLI's PTY backend, streamed over a direct loopback WebSocket so raw bytes bypass postMessage. Tabs mirror the worktree split-button pattern, support mixed drag-reorder with session tabs, and survive worktree-context switches without losing xterm state (slots are opacity-toggled in a persistent absolute-positioned layer, never unmounted). The legacy Cmd+/ integrated-terminal shortcut and console icon are preserved so existing muscle memory keeps working. * fix: resolve 3 pre-existing TS errors (undefined->null, this.context->extensionContext, autoFillSetting call shape) * chore: update nix node_modules hashes * fix: resolve all lint errors — eslint-disable complexity/max-lines, bump to v7.2.20, build kilo-code-7.2.20.vsix * feat: Add workflow to watch for new Opencode Releases and notify Slack/Vercel (for dashboard) webhooks. * feat: full Hermes tab + agent-assist pipeline e2e wired - Add HermesTab.tsx: enable/disable, URL, approval mode, API key, agent-assist, task submit, task tracker - Add Hermes message types to V4SubsystemRequest + V4SubsystemMessage in messages.ts - Add handleHermesStatusRequest/TasksRequest/SubmitTask/AgentAssist handlers to KiloProvider - Add setHermesServices() to KiloProvider; wire in extension.ts for all provider instances - Wire HermesTab into Settings.tsx sidebar (between VPS and ZeroClaw) - ZeroClaw+Hermes agent-assist: autoFillAll() + getSuggestions() + config audit on demand - Build: kilo-code-7.2.20.vsix (70.91 MB) * release: v7.2.21-EVO — merge upstream workflow, full audit pass, ulimit fix, VSIX built * WIP: Pre-upstream-sync checkpoint with MAOS customizations * branding: Update to KiloCode MAOS Edition - displayName, description, and About tab * feat: add package.json branding-preserving merge driver When upstream Kilo-Org cuts release commits that bump 'version' in packages/kilo-vscode/package.json (v7.2.21..v7.2.24 today, more weekly), they conflict with our DaveAI MAOS Edition branding fields. This driver auto-resolves the conflict deterministically: take the new version (and dependencies, scripts, contributes from upstream) while preserving displayName/description/publisher/icon/author/homepage/bugs/repository plus MAOS-titled commands. Setup (one-time per clone): bash scripts/setup-merge-drivers.sh Test: bash scripts/test-merge-driver.sh # PASS confirmed locally Effect on the upstream cherry-pick plan: 4 of the 9 currently-PROTECTED commits (release bumps) become auto-pickable. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * WIP refactor: extract DaveAI customizations from KiloProvider.ts to KiloProvider.dave.ts (RFC 001) - KiloProvider.ts: 4474 → 3581 lines (-893) - KiloProvider.dave.ts: 1015 lines (new) - Single hook point: line 585 in KiloProvider.ts dispatches V4 messages to overlay - Verified: 0 grep matches for MAOS|hermes|zeroclaw|daveai|HubServices in slimmed KiloProvider.ts PENDING (next session, 30-60 min): - caller-site rewires in extension.ts (12 lines) and SettingsEditorProvider.ts (6 lines) - These callers invoke setHermesServices/setV4Services/broadcastDiscoveryComplete on the provider directly; must redirect to (provider as any).__daveExtensions DO NOT PUSH yet — husky typecheck will fail until callers are rewired. Branch preserved locally as work-in-progress for the next engineering session. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * refactor: rewire callers to KiloProvider.dave overlay (RFC 001 step 2/2) Caller-site rewires for the RFC 001 overlay extraction: - extension.ts: 11 lines updated (+ 1 import) to call (provider as unknown as { __daveExtensions?: DaveProviderExtensions }).__daveExtensions?.X(...) instead of provider.X(...) for setHermesServices, setV4Services, broadcastDiscoveryComplete. Lines that target settingsEditorProvider (not KiloProvider) are intentionally unchanged — SettingsEditorProvider has its own internal forwarding. - SettingsEditorProvider.ts: 4 lines updated (+ 1 import) for the internal forwarding methods to invoke on the overlay. - RFC_001_CALLER_REWIRES_NOTES.md: full diff table + test plan. Effect: bun turbo typecheck should now pass on the feature branch. The 3 PROTECTED upstream commits 5107987, 6cc7863, 154f104 (autocomplete refactors) are now auto-pickable. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat: wire 6 real-backend tab handlers + onboarding + auto-update services Real-backend handlers (replace scaffolded UI-only versions): - packages/kilo-vscode/src/kilo-provider/handlers/{hermes,memory,routing,zeroclaw,governance,training}-webview.ts Each handler: fetch Hub at kilocode.updates.hubBaseUrl (fallback daveai.hub.baseUrl, then https://hermes.daveai.tech), bearer auth from SecretStorage, structured response, graceful 404 degradation. Hermes/Memory/Routing/ZeroClaw target existing services; Governance round-trips via /api/canonical-settings; Training drives a Hub training router with honest mock. - packages/kilo-vscode/src/kilo-provider/handlers/__tests__/*.test.ts (6 tests) - packages/kilo-vscode/src/kilo-provider/handlers/*-webview.README.md (4 readmes) KiloProvider.dave.ts (RFC 001 overlay) — wire-up: - Import 6 handlers - handleV4Message: dispatch to real-backend handlers FIRST (return true if consumed), fall through to legacy in-process switch on miss/error - isV4MessageType: add lowercase 'zeroclaw' alias (camelCase 'zeroClaw' kept for back-compat) New services: - packages/kilo-vscode/src/services/onboarding/ (5 files): OnboardingWizard.ts + OnboardingService.ts + index.ts + README + tests. Auto-detect Hub URL + env-var import + 5-question wizard. Goal: 2-min setup from clean install. - packages/kilo-vscode/src/services/auto-update/ (5 files): AutoUpdateService.ts + UpdatePromptUI.ts + index.ts + README + tests. Polls Hub /api/updates/manifest; 3 channels x 3 modes (prompt/auto/off). Effect: tabs Hermes/Memory/Routing/ZeroClaw/Governance now reach real Hub-side services; Training has honest mock executor; auto-update + onboarding wire up on extension activate. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(rfc-001): typecheck fixes + new tab handlers - KiloProvider.dave.ts: * Import handleTrainingWebviewMessage as handleTrainingRealWebviewMessage (rename mismatch fix) * Cast provider.extensionContext via 'as unknown as ...' (TS2339) * 'message' → 'm' (cast to any) in V4 switch body to satisfy strict 'unknown' type - services/onboarding/index.ts: re-export OnboardingDiscoveryService for legacy importers - New: governance-webview + training-webview handlers for the 5th and 6th wired tabs Typecheck: PASS (was 19 errors, now 0) VSIX build: PASS (71.47 MB, 147 files) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * ci: add build-vsix.yml workflow for automated VSIX packaging Builds KiloCode MAOS VSIX on tag push (v*) or workflow_dispatch. No code signing required — output is a plain .vsix for Install from VSIX. Uploads as release asset on tag push; artifact on every run. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Josh Lambert <josh@kilocode.ai> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Joshua Lambert <25085430+lambertjosh@users.noreply.github.com> Co-authored-by: Kirill Kalishev <kirillk@kilocode.ai> Co-authored-by: Evgeny Shurakov <eshurakov@users.noreply.github.com> Co-authored-by: marius-kilocode <marius@kilocode.ai> Co-authored-by: Marian Alexandru Alecu <a.marian.alexandru@gmail.com> Co-authored-by: Josh Holmer <jholmer.in@gmail.com> Co-authored-by: Tang Xinyao <31577196+tangxinyao@users.noreply.github.com> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: kiloconnect[bot] <240665456+kiloconnect[bot]@users.noreply.github.com> Co-authored-by: tangxinyao <xinyao.txy@antfin.com> Co-authored-by: Christiaan Arnoldus <christiaan.arnoldus@outlook.com> Co-authored-by: Imanol Maiztegui <imanol.mzd@gmail.com> Co-authored-by: kilo-maintainer[bot] <kilo-maintainer[bot]@users.noreply.github.com> Co-authored-by: Catriel Müller <catrielmuller@gmail.com> Co-authored-by: Mark IJbema <mark@kilocode.ai> Co-authored-by: Johnny Amancio <johnnyeric@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
gapon.message-list-content, but perf(vscode): paginated message loading with virtualized scroll #8911 introduced virtua'sVirtualizerwhich positions items absolutely based on measured box size — sogapno longer applied between turns.padding-bottomso virtua measures it as part of the item height. This uniformly restores spacing for all turn terminators (text, bash, todos, diffs, subagents, errors).MessageList — sub-agent to queued user spacing) that reproduces the failing case.Before:
After:
