Changelog

[v7.0.0] - 2026-10-09

Changed

Static system prompt. The system prompt now contains only the role definition, tool descriptions, tool guidance and a harness section. Per-session content — the skills catalog, the system info section and the git status — moves out of the system prompt and is delivered as <system-reminder> blocks on the first user message, so the prompt prefix stays byte-identical across sessions and the provider's prompt cache keeps hitting (src/core/prompts/system.ts, src/core/environment/contextReminders.ts).
No automatic workspace file listing. The environment details no longer enumerate up to 200 workspace files on the first message; that listing differed per workspace and session and rewrote the start of every request. The environment block now tells the model to explore with Bash (ls, rg --files, git ls-files) instead (src/core/environment/getEnvironmentDetails.ts).
Condense-and-retry on context overflow. The stale-tool-result pruner from v6.9.0 is removed: history is sent append-only, and when the provider reports a context-window overflow the task condenses the conversation and retries, up to three attempts. fitSummarySource trims the summary request to 60%, 30% or 15% of the context window when it doesn't fit, and the generic context-error pattern recognizes 400, 413 and 422 overflow responses from OpenAI-compatible gateways (src/core/task/Task.ts, src/core/condense/index.ts).

Added

Background shells. Bash commands can run detached with a background parameter; check_background and kill_background inspect and stop them, and a status bar above the chat input lists running shells with a stop action (src/integrations/terminal/BackgroundCommands.ts).
Reasoning effort selector. Models served through the MatterAI gateway expose a low/medium/high/max effort selector next to the model selector; the pick is saved per model in global state and sent on every request as the X-MATTERAI-REASONING-EFFORT header (packages/types/src/model-effort.ts).
Turn stats and refreshed marketing cards. The in-progress request row shows elapsed time and output tokens for the turn; the Axon Eido 3.2 marketing card is removed, and completed reasoning messages are filtered out so thinking is live-only (webview-ui/src/components/chat/TurnStats.tsx).

[v6.9.0] - 2026-10-03

Changed

Shell-first exploration. search_files and list_files are no longer offered to the model. It searches and lists with rg, find, ls and git through the shell tool, the way Claude Code does, and the system prompt (tool guide, capabilities, rules, objective, tool-use guidelines, environment details) teaches the common patterns. To keep this from becoming an approval prompt on every search, read-only commands skip the approval prompt in every command approval mode, including "Ask": rg, grep, find without -exec/-delete, ls, cat, head, wc, git status/diff/log/show/grep, and pipes / && / ; chains of these with no redirects or command substitution (src/core/tools/readOnlyCommand.ts). Anything unrecognised, or marked isDangerous, still follows the normal approval mode. The old executors stay in the codebase so previously rendered rows keep working.
execute_command is shown to the model as Bash. The tool is renamed to match what models are trained on. Internally it is still execute_command (approval, UI, mode groups, persisted history); src/shared/toolAliases.ts translates at the boundary: incoming native tool calls named Bash (or the old execute_command) map to the internal tool, and history sent back to the model uses Bash.
Commands run in bash. The command executor now prefers bash over the VS Code terminal profile (fish and csh choke on the &&, $? and $(...) syntax the model writes, and the cwd-tracking suffix already assumed a POSIX shell): bash on macOS/Linux, Git Bash on Windows, falling back to the configured shell (POSIX) or cmd.exe only when bash isn't installed. The system prompt states the shell, and warns the model when it is stuck on cmd.exe. Commands run with shell integration in the VS Code terminal still use the terminal profile.
Leaner system prompt. The "Verifying tool results and avoiding loops" and "Investigation efficiency" sections asked the model to deliberate before every tool call. They are replaced by a four-line "Working style" block (act directly, locate -> edit -> check once, batch independent calls, never repeat an identical call more than twice).

Added

Stale tool-result pruning. Once context passes 40% of the model's window, bulky results of read_file, execute_command (Bash), codebase_search, web_fetch and web_search older than the four most recent tool results are sent as one-line stubs. The stored conversation and task history are untouched; only the outgoing request shrinks, and the boundary advances in batches of six so the request prefix (and the provider's prompt cache) stays stable between prunes. It resets when the history is condensed. Results containing images are never stubbed.
Actionable "not found" edit errors. When file_edit / multi_file_edit cannot find old_string, the error now includes the closest region of the file (up to 7 numbered lines with their exact whitespace), so the model can retry without another read.

Notes

Already present in this extension and therefore not part of this release: whitespace/CRLF-tolerant edit matching, automatic context condensing, repeated-call detection, and process-tree kill on command abort. The bench/ harness and parallel read-only shell commands are not ported (parallel terminals share a single interactive ask).

[v6.8.6] - 2026-09-15

Fixed

Plan-aware default model. getKilocodeDefaultModel no longer returns the hardcoded openRouterDefaultModelId. It now resolves the default from the live MatterAI catalog (/v1/web/models): free accounts get the entry the backend flags freePlan, every other plan gets the first catalog entry (index 0, ordered by the catalog's sortOrder). The account plan comes from /axoncode/profile; a failed profile fetch is treated as free, and any catalog failure still falls back to openRouterDefaultModelId. The stale-model reset paths in ClineProvider now pass the KiloCode token and organization id so they resolve the same plan-aware default instead of the hardcoded one.

[v6.8.5] - 2026-09-04

Added

Generalized malformed tool-call JSON repair (src/utils/jsonRepair.ts). A new best-effort repair layer (parseToolCallArguments) is now the single entry point for every tool-call argument ingestion path — live streaming, stream finalization, and MCP argument validation. When a model emits almost-JSON arguments, the call is repaired and executed instead of being dropped (weaker models repeat the exact same mistake on retry, dead-looping the task). Repairs include: placeholder tags in structural positions (e.g. "limit": <longcat_arg_value>180 keeps the value; a tag-only value becomes null so the tool's default applies; a tag inside a key is stripped), unquoted keys and scalar values ({path: "src"}, {"file_pattern": *.ts}), single-quoted keys/values, Python literals (True/False/None), // and /* */ comments, trailing commas, missing commas between members, and lost key quotes ({"offset: 600, ...}). At stream finalization only, truncated arguments are recovered by closing dangling strings and open containers innermost-first, appending null after a dangling colon, and wrapping braceless bodies (path: "src"); during streaming an incomplete buffer stays unparseable so deltas keep accumulating. A single-object array is unwrapped when the model wrapped the tool object by mistake. Valid JSON is always strict-parsed first and returned verbatim (repaired: false) — the only exception is stripping a leaked placeholder tag from a key, since keys are structural parameter names; tags inside string values are legitimate content and are never touched.
Repair transparency. ToolUse blocks now carry a repaired flag; when a repaired tool call completes, the executed arguments are appended to the tool result as a note so the model sees what actually ran instead of repeating its malformed form on the next turn. useMcp_tool applies the same repair when validating MCP tool arguments and appends the same note.
Executor parameter hardening. read_file (native JSON, single file_path, and XML paths) and the kilocode read-file path now parse offset/limit through a parsePositiveInteger helper that floors floats, clamps to a minimum of 1, and falls back to the tool default on non-numeric input; search_files floors fractional max_results/context_lines into their allowed ranges.

Fixed

Placeholder-tagged tool arguments are repaired instead of dropped. When a model emits malformed native tool-call JSON with a placeholder tag in value position (e.g. "limit": <longcat_arg_value>180 or "limit": <longcat_arg_value>), the tag is stripped and the trailing value kept, or the value becomes null when no value follows so the tool's default applies. The repair runs only after strict JSON.parse fails, so valid JSON is never altered. Streaming partial previews are tag-cleaned as well.

[v6.8.4] - 2026-09-03

Added

Per-model usage visibility. The usage dialog and a new Settings → Model Usage section now show each tracked OSS model's share of the shared plan pool as weekly/monthly percentages (no credit amounts), alongside the model's plan-cost multiplier badge (e.g. 5x cost). The new settings section also renders the weekly/monthly plan windows with reset times and a refresh action, and is excluded from the save-button flow since it is read-only. Backend: /axoncode/profile now returns a modelUsage array (model, multiplier, weeklyPercentage, monthlyPercentage); charges for tracked OSS models are multiplied by their plan-cost multiplier before draining the shared pool.

Changed

Dynamic model catalog synchronization. Models are now dynamically fetched from the MatterAI backend /v1/models (or /v1/web/models) and registered into the client model registry (registerDynamicKilocodeModels), replacing static hardcoded lists while continuing to exclude deprecated axon models from active selection. Models automatically refresh on window focus, via the refresh button in the model selector, and through a 10-minute background poller. Cache bypass (forceRefresh) ensures database updates reflect immediately without restarting the extension.
OSS model catalog replaces Axon models. The KiloCode model catalog (extension kilocode-models.ts and the webview useOpenRouterModelProviders copy) now exposes seven OSS models — meta/muse-spark-1.2-contributor (Muse Spark 1.2 Contributor), deepseek/deepseek-v4-flash-0731 (DeepSeek V4 Flash), zai/glm-5.3 (GLM 5.3), zai/glm-5.3-flash (GLM 5.3 Flash), gpt-5.6-sol (GPT-5.6 Sol), gpt-5.6-luna (GPT-5.6 Luna), and gemini-3.7-flash (Gemini 3.7 Flash) — in place of the Axon context-window variants. Each OSS model carries its published per-token pricing (Muse Spark 1.2 Contributor $0.10/M input, $0.002/M cache read, $0.20/M output; DeepSeek V4 Flash $0.14/M input, $0.028/M cache read, $0.28/M output; GLM 5.3 $1.40/M input, $0.14/M cache read, $4.40/M output; GLM 5.3 Flash $0.15/M input, $0.03/M cache read, $0.50/M output; GPT-5.6 Sol $5/M input, $0.50/M cache read, $30/M output; GPT-5.6 Luna $0.20/M input, $0.02/M cache read, $1.20/M output; Gemini 3.7 Flash $0.75/M input, $0.075/M cache read, $3.75/M output). The default model is deepseek/deepseek-v4-flash-0731 (openRouterDefaultModelId in @roo-code/types, plus the CLI kilocodeModel defaults and web-evals MODEL_DEFAULT). Stored Axon model selections are now stale and reset to the default on next launch via the existing isValidKilocodeModel stale-model check.

Fixed

OpenRouter models no longer wiped on background refresh. refreshKilocodeModels (window-focus refresh, 10-minute poller, startup fetch) previously posted openrouter: {} to the webview, blanking the OpenRouter model list until the next manual reload. It now fetches both openrouter and kilocode-openrouter in parallel (mirroring the requestRouterModels handler) and posts the merged payload, logging and skipping only the provider that failed.
Stale model selections reset on launch. The dynamic catalog is now fetched once at provider startup (after ContextProxy initialization) and the webview state re-posted, so the isValidKilocodeModel stale-model check runs against a populated catalog instead of the empty-catalog bypass. Previously, a stored axon model survived until the first focus- or webview-triggered fetch completed.

[v6.8.2] - 2026-08-28

Added

Parallel execution of independent read-only tools. When a native-JSON provider returns several independent read/search calls in one assistant message (read_file, search_files, list_files, list_code_definition_names, codebase_search, lsp), they now run concurrently (up to 4 at a time via MAX_PARALLEL_READ_ONLY_TOOLS in the new src/core/tools/toolExecutionPolicy.ts) and their results are committed in model order so tool_use/tool_result pairing stays intact. Mutating, interactive, and external tools remain on the serialized path; read approvals inside a batch render as non-interactive progress rows instead of sharing the task's single ask promise. parallel_tool_calls is now enabled for native tool call providers.

Changed

search_files is one-shot and ripgrep-first. Ripgrep is now the default engine with FFF as the fallback; the default and maximum result count rose from 50 to 100; the model-facing cursor parameter was removed from both the XML and native JSON tool schemas. Pages that hit the cap say "additional matches omitted; refine the search pattern or path instead of paginating."
System prompt condensed. The RULES, tool-use guidelines, and objective sections were rewritten to be shorter and tool-agnostic: codebase_search is recommended when the target is unclear instead of being mandatory before any exploration, the smallest sufficient tool is preferred, and independent reads/searches are batched in JSON mode while edits and commands stay sequential. The legacy text-tool guidance block is now only included for XML tool style.
Native tool schemas tightened for strict mode. execute_command now requires cwd, message, and isDangerous (with destructive-action classification); file_edit/multi_file_edit require replace_all; generate_file requires path; check_past_chat_memories requires workspace (nullable where optional). The lsp description was condensed.
allowedTools metadata respected when empty. addNativeToolCallsToParams uses mode-filtered tool exposure whenever provided, including an intentional empty list for tool-less modes, instead of falling back to the full native tool set.
Write delay default reduced. DEFAULT_WRITE_DELAY_MS dropped from 1000ms to 250ms so routine edits do not add a second of latency; users with slower language servers can raise writeDelayMs in settings.
KiloCode notification polling disabled. fetchKilocodeNotificationsHandler now responds with an empty notification list; the native-notification fetch/display path was removed.

Fixed

Malformed tool-call JSON recovers instead of stalling. Tool calls whose JSON arguments fail to parse at stream finalization are tracked by AssistantMessageParser (getDroppedMalformedToolCalls); the task loop injects a corrective message with a truncated preview of the raw arguments and retries, bounded by the consecutive-mistake limit.
Tool-repetition auto-retry is bounded. The automatic "try again" response for repeated identical tool calls is capped at 2 attempts (MAX_TOOL_REPETITION_AUTO_RETRIES); after the budget is exhausted an error row tells the model to change approach instead of looping forever.

[v6.8.1] - 2026-08-21

Changed

Investigation efficiency guidance in system prompt. Added an "Investigation efficiency" section to the shared tool-usage instructions (applyDiffToolDescription in src/core/prompts/system.ts) that directs the agent to classify comprehension questions separately from implementation tasks, form a one-line hypothesis before searching, read call sites rather than implementation internals, avoid reading prose/content when the question is about control flow, and stop exploring as soon as it can answer.
Overlapping read detection in read_file. readFileTool.ts now detects when a requested file region partially overlaps with a region already read earlier in the same task (same file, same mtime). The exact-match short-circuit (unchanged) still blocks identical re-reads; the new overlap check prepends a notice to the served content telling the model which lines are already in context, so it does not waste a follow-up call re-reading the overlapping portion. This catches the common failure of reading lines 600-849 then 800-1059 of the same file.
Zero-result guidance in search_files. searchFilesTool.ts now appends actionable guidance when a search returns 0 matches, directing the model to tighten or simplify the regex, widen the path scope, try a different glob, or stop searching after 2+ failed attempts. This prevents the search to 0 results to slightly different regex to 0 results loop.
Native tool description improvements. The read_file native tool description now tells the model not to read file contents (prompt text, config values, prose) when investigating control flow, and not to re-read regions already read earlier. The search_files native tool description now tells the model to scope the path to the narrowest plausible directory and to stop after 2+ zero-result searches.

[v6.9.0] - 2026-08-04

Added

generate_file native tool. Orbital can now produce real file artifacts (PDF, DOCX, PPTX, XLSX, CSV, MD, TXT, HTML) directly from the chat. The tool delegates to the MatterAI backend (/axoncode/generateFile, Bearer-token auth) and holds the resulting binary in memory as base64 — no files are written to disk until the user explicitly saves. The chat row renders a file card with a type-specific icon, file name, type label, file size, and two text buttons: View File opens a dedicated editor-tab webview panel (ViewColumn.Beside) that renders each type natively — PDF via PDF.js canvas rendering (no browser plugin, avoids the VS Code webview sandbox restriction), HTML as live rendered content, Markdown via marked.js, PPTX presentation decks parsed from Marp-style slide markdown with accent colors and structured cards, CSV as an HTML table, TXT/DOCX/XLSX as formatted text — and Save File writes the binary to the OS Downloads folder. The tool is public (no experiment flag) and lives in the edit tool group alongside file_write and generate_image.

[v6.8.0] - 2026-08-01

Removed

Fireworks AI provider. Removed the Fireworks third-party provider (handler, settings schema, model fetcher entry, model selector, settings UI, docs, icon, CLI registry, JSON-schema entries, and fireworksApiKey / getFireworksApiKey i18n strings across all 22 locales). Existing fireworksApiKey values in stored profiles are no longer recognized; users will be prompted to pick a different provider on next launch.

[v6.7.2] - 2026-07-31

Added

Organization usage metrics. Orbital now reports metadata-only user-message events, accepted agent line counts with model/client/IDE dimensions, and newly observed Git commit line totals for AI-share, active-user, leaderboard, conversation, and client-version analytics. Prompt content and file contents are never included.
Axon Lumen 4 model support. Added support for axon-lumen-4-code in 200K and 400K context variants (axon-lumen-4-code-200k and axon-lumen-4-code-400k). Lumen models are available on Pro Plus and Ultra plans.
Plan gating for Lumen models. Lumen models now show the same warning icon and upgrade tooltip as 400k context variants for users on plans below Pro Plus, are disabled in the chat model selector, hidden from the settings model picker, and fall back to axon-eido-3-code-pro-200k when a restricted Lumen model is selected.

Changed

Flash model display name. Axon Eido 3 Flash now shows (200K context) in its display name, consistent with the other Axon model variants.
Orbital update restart sequence. The "Update & Restart" action now waits one minute after the update finishes installing, then restarts all extensions, reloads the window, and restarts all extensions again. The MatterAI progress animation remains visible during the wait.

[v6.6.9] - 2026-07-25

Changed

Remove free tag from Axon Eido 3 Flash. Removed (free) suffix from the Axon Eido 3 Flash model display name.

[v6.6.8] - 2026-07-24

Changed

File-edit tools reject ambiguous and guessed matches. file_edit and multi_file_edit now require old_string to be copied verbatim from a current read of the file and to identify exactly one location. The XML and native JSON tool descriptions, parameter descriptions, and the system-prompt editing discipline section were rewritten to forbid inventing, reconstructing, or guessing file content, indentation, whitespace, or escaping, and to direct the model to re-read the intended target and retry with exact surrounding context on a missing or multiple-match error.
replace_all is intentional-only. The replace_all parameter description now states it must be set to true only after verifying the requested change should apply to every occurrence, never merely to bypass a multiple-match error.
Clearer ambiguous-match errors. performReplacement now reports the match count and the line numbers where matches start (capped at five, deduplicated, with a "first N of M" suffix when truncated) instead of a generic "matched multiple locations" message, and stops at the first ambiguous candidate so a looser strategy cannot guess which occurrence was intended.
Clearer not-found errors. The not-found error now appends guidance that no edit was applied, that the model must not guess or invent a corrected old_string, and that it should re-read the intended target and copy the exact current text before retrying.

Added

getMatchDiagnostics in fileEditTool.ts. New helper that counts every match start (including overlaps) incrementally without allocating per-match prefixes or arrays, and returns the occurrence count plus a capped, deduplicated list of starting line numbers for concise error messages.
File-edit safety spec. New src/core/prompts/tools/__tests__/file-edit-safety.spec.ts asserts the text and native tool descriptions require verbatim unique matches and forbid bypassing ambiguity errors.
Ambiguous-match tests. New cases in performReplacement.spec.ts cover rejecting ambiguous exact matches, same-line duplicate line-number formatting, overlapping match locations, and allowing repeated matches only when replace_all is intentional.

[v6.6.7] - 2026-07-22

Added

FFF-backed search_files with ripgrep fallback. search_files now runs through the native @ff-labs/fff-node file finder first, falling back to a streaming ripgrep implementation when FFF is unavailable or fails. Results are compact (at most three matches per file), paginated, and capped at a configurable page size so the model reads only the relevant regions instead of a context-flooding dump.
- New src/services/search-files module (index.ts, format.ts, types.ts) is the single entry point: it tries FFF, falls back to ripgrep, and formats a compact Engine / Matches / Next cursor page.
- New src/services/fff/index.ts wraps the ESM-only FFF SDK, keeps it out of the CommonJS bundle via a dynamic import(), caches up to three FileFinder instances per base path, encodes the user regex for FFF's query parser, and continues through FFF pages until .orbitalignore-filtered matches fill the requested page.
- New searchFilesWithRipgrep in src/services/ripgrep/index.ts streams JSON ripgrep output, applies .orbitalignore before consuming the page budget, enforces the per-file match cap, and returns a ripgrep-engine cursor for continuation.
- New cursor, max_results (1-100, default 50), and context_lines (0-2, default 0) parameters on both the XML and native JSON search_files tool schemas; cursors are engine-scoped (fff:<offset> / ripgrep:<offset>) and validated by parseSearchCursor / normalizeNullableSearchString.
- searchFilesTool now parses the new params and delegates to searchFiles; toolParamNames and SearchFilesToolUse extended accordingly.
- esbuild.mjs copies the FFF runtime packages (and their platform-specific native libraries) under dist/fff/node_modules so the packaged extension can resolve the native SDK; root package.json declares pnpm.supportedArchitectures so all platform FFF binaries install.
- extension.ts disposes FFF finders and closes the native library on deactivate.
- Tests for the FFF wrapper, ripgrep fallback, search-files orchestration, and cursor/format helpers.

Changed

search_files system prompt rewritten. The tool description and capabilities section now describe compact, paginated matches and direct the model to read the relevant file region rather than relying on large context windows. The verbose file_pattern quoting examples were removed in favor of a concise parameter list.
ChatRow shows the file_pattern for search_files rows. Parallel searches are now distinguishable in the chat history by their glob, rendered as a <code> chip next to the regex.
Pinned todo list horizontal padding. ChatView pinned-todo container padding aligned with the shared mx-3.5 chat column padding.

[v6.6.3] - 2026-07-21

Changed

Orbital update banner now polls continuously. The in-app update banner no longer checks the Open VSX registry only once on mount. It now re-checks every five minutes (including while an update banner is already visible, so a newer release can replace a stale one), and whenever a long-lived view becomes visible again or the window regains focus. This keeps every open IDE window informed about subsequent releases without requiring an Extension Host restart. Polling is throttled to once per minute and pauses only while an installation or restart is actively in progress. The "Update & Restart" action now optimistically flips the banner to an installing state before the backend confirms, so the spinner appears immediately.

[v6.6.1] - 2026-07-21

Added

In-app Orbital extension update banner. When running inside the Orbital IDE, the chat view now polls the Open VSX registry on mount and shows a banner above the composer when a newer version of the extension is available. The banner surfaces the latest version, an "Update & Restart" action that installs the update via workbench.extensions.installExtension and reloads the window, and a retry button on failure. Progress states (downloading / installing / restarting) are reflected inline with a spinner. The check is gated to Orbital hosts (appName === "orbital" or uriScheme === "orbital") so VS Code installs are unaffected.
src/services/orbital/extensionUpdate.ts. New module exposing checkForOrbitalExtensionUpdate, installOrbitalExtensionUpdate, isOrbitalIde, and a semver-aware isNewerVersion helper that handles v-prefixed and prerelease tags.
orbitalUpdate i18n strings. New chat:orbitalUpdate.* keys (available, description, updateAndRestart, downloading, installing, restarting, failed, retry).

Changed

Primary button border-radius. vscode-button[appearance="primary"] border-radius bumped from 6px to 7px in webview-ui/src/index.css for a slightly softer corner.

[v6.6.0] - 2026-07-21

Added

Axon Eido 3 context-window variants. axon-eido-3-code-pro and axon-eido-3-code-mini are now exposed as separate 200K and 400K context options (-200k / -400k suffixes). Both variants share the same upstream model; the extension resolves the selected variant to its API model ID via getKilocodeApiModelId. The default model is now axon-eido-3-code-mini-200k.
400K context gated to Pro Plus and Ultra. The 400K variants are only available on Pro Plus and Ultra plans. The model selector disables them with an upgrade tooltip for lower tiers, the KiloCode settings panel filters them out and auto-falls back to the matching 200K variant, and the CLI /model command rejects selection with a clear error message.
SelectDropdown group headings. New DropdownOptionType.GROUP renders non-selectable section headings (e.g. "Context: 200k") inside the dropdown, with automatic hiding when the group has no visible items.
SelectDropdown custom value rendering. New renderValue prop lets callers customize how the selected option is displayed in the trigger (used by the model selector to dim the context qualifier).
Sticky search bar in SelectDropdown. The search input stays pinned at the top of the dropdown while options scroll.
New icons in utils/customIcons.tsx. BulbIcon, PlusIcon, and ArrowUp02Icon added to support the refreshed model selector and chat composer.

Changed

Model selector refresh. Axon models are now grouped by context window with a BulbIcon indicator, the selected option is highlighted, and the trigger uses a compact rounded style matching the command-approval selector. Tooltips fall back to the model's description when no curated tooltip is defined.
Chat composer refresh. The attachment button moved to the left of the model selector with a PlusIcon, the send button uses an ArrowUp02Icon, and the textarea padding was unified across edit and compose modes.
Follow-up question card. Restyled with a rounded-xl border, subtle shadow, and improved typography (15px / leading-6). The suggestion list now uses bordered cards with a focusable copy-to-input button.
Shared chat horizontal padding. New CHAT_CONTENT_HORIZONTAL_PADDING constant (px-3.5) in webview-ui/src/components/chat/chatLayout.ts is used by ChatRow, ChatTextArea, ChatView, ExplorationGroupRow, BrowserSessionRow, QueuedMessages, and the sticky user message so the chat column stays aligned.
Markdown list spacing. MarkdownBlock list/ordered/unordered line-height bumped from 1.35em to 1.7em for better readability.
Task header title opacity. KiloTaskHeader task title opacity raised from 70% to 100%.
Logout copy. Logout notification changed from "Logged out from Roo Code Cloud" to "Logged Out from Orbital".

Fixed

WebAuthService.logout state transition. Logout now explicitly calls transitionToLoggedOut() after clearing persisted credentials, so the in-memory state, session token, and user info are cleared even when SecretStorage.onDidChange does not fire. The auth-state-changed event is emitted with the previous state.

[v6.4.9] - 2026-07-21

Fixed

Accept-all edit telemetry. Accepting all pending edits from the IDE review bar now reports accepted line counts to the task metadata endpoint, matching the chat and per-edit accept actions.
Review-bar change navigation. The up and down buttons now move between individual edit locations in the active file, while the left and right buttons continue to navigate between edited files.

Added

/create-skill built-in command. Users can describe a repository-specific workflow in plain language and have Orbital create or update a reusable skill under .orb/skills/<skill-name>/, including any supporting scripts, references, or assets within that skill directory.

[v6.4.7] - 2026-06-30

Added

Official plugins marketplace. The former Skills Marketplace now installs each claude-plugins-official entry as a complete bundle under .orb/plugins/<plugin>/. Installed plugin skills and slash commands use plugin:capability names, bundled MCP servers are registered with namespaced IDs, and the UI reports the bundle inventory after installation.
Shared .orb/ convention with OrbCode CLI. AGENTS.md is now loaded from the repo-level .orb/ directory (alongside the project root and legacy .orbital/), so the Orbital IDE extension and the OrbCode CLI read the same project memory. Machine and settings locations are unchanged.
Linked-repositories service. A new src/services/links module is the single source of truth for the linked-repo feature, sharing .orb/links.json and the produced environment context between the IDE extension and the CLI.
/link built-in command. A new slash command for managing Linked Repositories: reads the current list, drives add/remove via ask_followup_question, and writes .orb/links.json with verbatim user input. Resolution and validation of the folder path happens at read time so links written by either tool stay portable.

Changed

/init reworked for cold-start. Replaces the verbose .roo/rules-* per-mode layout with a single, concise .orb/AGENTS.md (project structure, architecture, business-logic mapping, code patterns/conventions) capped at ~150 lines so it stays cheap to include in every future prompt. Existing AI assistant rules (CLAUDE.md, Cursor, Copilot) are still folded in. Refines an existing AGENTS.md rather than overwriting it.
Built-in commands spec. The expected command list is now commit, migrate, init, link; init is asserted to target .orb/AGENTS.md with the new concise structure, and link is asserted to manage .orb/links.json via ask_followup_question.

[v6.4.6] - 2026-06-25

Added

Tiered usage support in OutOfCreditsBanner. New selectResetIso() helper picks the correct reset time: when a tier is exhausted it surfaces the latest reset among exhausted windows; otherwise it falls back to the soonest upcoming reset, then to the legacy creditsResetDate. Banner label shortened from "Pro models limits reset at" to "Limits reset at".
OrbCode CLI marketing card. A third rotating marketing card in the welcome view pointing users at the CLI, joining the existing two on the 10-second rotation ((prev + 1) % 3).
New custom icons in utils/customIcons.tsx. Folder01Icon, Folder02Icon, ArrowDown01Icon, AddCircleHalfDotIcon and others added to support the refreshed MCP and history views.

Changed

Orbital → Orbital rebrand in copy. troubleMessage string in chat.json and the consecutive-mistake-limit description in settings.json now read "Orbital" where they previously read "Orbital".
Axon marketing copy. Welcome-view subtitle changed from "Frontier LLMs, fraction of the cost. Save 70% inference cost." to "Cut agent inference costs by 60% using Frontier Axon models".
History UI refresh (TaskItem). Visual and layout rework of the history list (90 insertions / 52 deletions).
MCP UI refresh (McpView). Major rework of the MCP view with a new icon set and updated i18n strings (283 insertions / 228 deletions across McpView.tsx, mcp.json, and customIcons.tsx).
Login UI refresh (KiloCodeAuth, ProfileView, WelcomeView). Consolidated the auth and welcome flows with refreshed i18n strings.
Chat agent UI refresh. Visual updates to ChatRow, ExplorationGroupRow, and ReasoningBlock (63 insertions / 61 deletions).
Button UI refresh. Updated button styles in index.css for consistency across surfaces (17 insertions / 5 deletions).

Fixed

Stream idle timeout bumped to 180s. STREAM_IDLE_TIMEOUT_MS raised from 60s to 180s to reduce false-positive timeouts during long-running streaming responses, especially with slower or more verbose model outputs.

[v6.4.4] - 2026-06-23

Added

Stream idle timeout: 60s idle window on model stream consumption. If no chunk arrives within the window (network drop, socket close, server stall), a descriptive error is thrown so the catch block can persist the failure and surface a streaming_failed row to the UI
Stream disconnection UI surfacing: New streaming_failed ErrorRow variant with a localized explanation of the likely cause for transport-level errors (ECONNRESET, socket hang up, fetch failed, ETIMEDOUT, ENOTFOUND, etc.), and a retry button matching the existing Provider error: behavior
chat:apiRequest.streamDisconnected i18n key

Fixed

Ripgrep binary resolution on modern VS Code / Orbital hosts: @vscode/ripgrep-universal nests the binary in a per-platform subfolder (bin/{os}-{arch}/rg) that the previous lookup chain did not check, causing getBinPath to return undefined and file searches to silently fail. Adds the universal package to the lookup chain with both node_modules and node_modules.asar.unpacked paths
Unified message queue: Replaced the local manualMessageQueue with the shared messageQueueService as the single source of truth. Queued messages now stay visible in the UI until they are actually dispatched; isWaitingForAskResponse is set before dequeue so a follow-up message resolves the ask rather than being re-routed into the queue

Changed

About footer actions: Export, import, and reset buttons now sit in a flex container with gap-1.5 so the icon and label share a single evenly spaced row. Removed ad-hoc pb-0.5 icon padding
Footer support copy: Removed the dead Discord link; the message now ends at the Reddit mention

[v6.2.3] - 2026-05-21

Fixed

Chat title display: Handle JSON-wrapped title strings from server, now supports plain string, JSON object, and stringified JSON responses, with normalized-title fallback at every consumer layer to clean already-persisted malformed titles

Changed

Updated matterai models, llm router spec tests, exploration group row, out of credits banner, pretty model name, and openrouter provider hooks

[v6.2.0] - 2026-05-21

Added

Tool call result pairing safety net: New backfillMissingToolResults function in OpenAI format transform that backfills placeholder tool messages for unanswered parallel tool calls, preventing provider rejections when tool_call/tool_result pairs are mismatched
Tool call result pairing module: Pure functions (reconcileAssistantToolUses, toolUseIdsRequiringResults, allToolResultsCollected) to keep assistant tool_calls and tool_results paired 1:1, with comprehensive unit test coverage
Task.ts reconciliation: Gating API requests until every tool_call in the assistant message has its matching tool_result collected, preventing partial result sets from being sent to providers

Changed

AssistantMessageParser performance: Avoided O(n²) string slicing in the streaming hot loop by deferring content updates to end-of-chunk; added 20KB max extract length threshold for large accumulated arguments during streaming
presentAssistantMessage optimization: Replaced deep clone with shallow copy for streaming content blocks to reduce overhead during large file streaming
FileWriteTool partial display: Truncated large file content during streaming preview (5KB limit) to prevent IPC bottlenecks and UI freezing
Model list cleanup: Removed deprecated axon-code-2-pro and axon-code-2-pro-high model entries from provider configurations

Fixed

ReasoningBlock arrow icon rotation styling
OutOfCreditsBanner border-radius styling consistency

[v6.0.0] - 2026-04-10

Highlights

This major release introduces the Agent Manager, a powerful new feature for managing multiple AI agents across workspaces with a dedicated sidebar and file viewer panel.

Added

Agent Manager

Agent Manager Sidebar: New collapsible sidebar for managing multiple agent tasks across workspaces
- Workspace-based task organization with expandable folders
- Quick access to recent tasks with compact timestamps
- "New Agent" button for starting fresh conversations
Agent File Viewer: Resizable file viewer panel for reviewing diffs and file changes
- Drag-to-resize functionality with min/max width constraints
- Automatic viewport-aware sizing
- Pull request-style diff view integration
Agent Toggle: Toggle button to show/hide the agent manager sidebar with smooth animations

Exploration Groups

Elapsed Time Display: Exploration groups now show elapsed time during and after exploration
- Live timer updates during active exploration (e.g., "Exploring for 1m30s")
- Summary with total time on completion (e.g., "Explored 5 files, 2 searches for 45s")
- Auto-collapse when exploration completes

Model Selection

Axon Model Tooltips: Hover tooltips in ModelSelector showing details for Axon models
Enhanced Model Provider Hook: New useOpenRouterModelProviders hook for better model management

Image Handling

ImageAttachment Interface: Refactored image handling to use a unified ImageAttachment interface
Improved Thumbnails: Enhanced thumbnail component with better image preview support

UI Components

Custom Icons Utility: New custom SVG icons including Folder01Icon, Folder02Icon, ArrowDown01Icon, AddCircleHalfDotIcon
Copy Button Update: Updated copy icon styling

Changed

Streamlined Chat UI: Cleaner, more focused chat interface with improved visual hierarchy
Better Diff View: Enhanced diff view for the edit tool with improved readability
Maximize States: Fixed maximize state handling for better window management
ChatRow Component: Enhanced with agent manager mode support and improved rendering
ChatTextArea: Refined for better user interaction
ReasoningBlock: Minor styling improvements

Fixed

Maximize state persistence issues
Various UI inconsistencies across components

Breaking Changes

Minimum VS Code version requirement updated

[v5.7.6] - 2026-04-05

Added

Native tool call helpers for kilocode provider
Enhanced assistant message parsing with improved tool handling
LSP tool integration for better code navigation
File write tool with improved content handling
Multi-file search replace strategy for complex edit operations
New tool type definitions in packages/types
Git branch display in bottom API config showing current repository branch
GitBranchIcon custom SVG icon for branch visualization
New message types for git branch request/response (fetchGitBranchRequest, gitBranchResponse)

Changed

Major Tool Architecture Refactoring: Consolidated multiple file editing tools into a unified, simplified tool system
- Merged edit_file, insert_content, search_and_replace, write_to_file, apply_diff tools into streamlined file_write tool
- Consolidated plan file tools (list_plan_files, read_plan_file, plan_file_edit) into core task management
- Removed redundant native tool prompt definitions
- Simplified getAllowedJSONToolsForMode with cleaner tool selection logic
Refactored system prompts for better tool organization and maintainability
Updated assistant message parsing with new parseAssistantMessageV2 implementation
Enhanced diff strategies with improved multi-search-replace functionality
Updated task handling with improved checkpoint service integration
Enhanced webview message handler for better task coordination
Improved ChatRow, ChatView, and CommandExecution UI components
Updated i18n translations for en locale
Refactored BottomApiConfig component with improved layout and branch display
Enhanced ProgressIndicator component styling
Updated index.css with improved styling utilities

Fixed

Shadow checkpoint service stability improvements
Assistant message presentation edge cases
Browser action tool compatibility
Webview message type definitions
File write tool: proper workspace path detection for partial display during streaming
File write tool: ensure tool result is always pushed on user rejection
GitHubDiffView: dynamic margin calculation for proper diff alignment
ToolUseBlock: improved component structure and styling
Git utilities: properly distinguish between current branch and default branch
- Added currentBranch field to GitRepositoryInfo interface
- Fixed webviewMessageHandler to use currentBranch instead of defaultBranch
- Updated tests to reflect the correct field name

Removed

Deprecated individual file editing tools (editFileTool, insertContentTool, searchAndReplaceTool, writeToFileTool, applyDiffTool, multiApplyDiffTool)
Deprecated plan file tools (listPlanFilesTool, readPlanFileTool, planFileEditTool)
Redundant tool prompt files (edit-file.ts, insert-content.ts, search-and-replace.ts, write-to-file.ts)
Native tool prompt definitions (apply_diff.ts, edit_file.ts, insert_content.ts, list_plan_files.ts, plan_file_edit.ts, read_plan_file.ts, search_and_replace.ts, write_to_file.ts)
Associated test files for removed tools
FastApplyChatDisplay component (functionality consolidated)

[v5.7.3] - 2026-04-02

Added

OAuth authentication support for MCP servers
LSP tool for code definition navigation
file_write tool implementation

Changed

Modernized settings panel UI with new card-based design
Reduced VSCodeButton size by 5% across all variants for better visual consistency

Fixed

Allow changing 3p models for existing tasks and fixed model display names

[v5.7.2] - 2026-03-29

Added

Third-party provider support with Fireworks, Ollama, and OpenCode integrations

[v5.7.1] - 2026-03-28

Changed

Minor release updates and stability improvements

[v5.7.0] - 2026-03-20

Changed

Major release with enhanced provider integrations

[v5.6.5] - 2026-03-14

Changed

Updated git repository URL across all localization files
Minor stability improvements

[v5.6.4] - 2026-03-13

Changed

Name cleanup and branding consistency updates
Improved CSS styling for better UI consistency

[v5.6.3] - 2026-03-12

Changed

Improved reasoning flow with enhanced time tracking
Better reasoning block display and interaction

[v5.6.2] - 2026-03-10

Added

Fade-up and shimmer animations for ChatRow, ChatView, and ReasoningBlock components
Enhanced visual feedback with smooth animation effects

Changed

UI cleanups across chat components
Improved chat message rendering and layout

[v5.6.1] - 2026-03-08

Changed

Monthly usage percentage display update in settings
Enhanced usage tracking UI in BottomApiConfig component

[v5.6.0] - 2026-03-08

Added

Split high think models for better model organization
Enhanced OpenRouter model providers hook

Changed

Tool cleanups and optimizations
Improved model selection handling

[v5.5.7] - 2026-03-06

Added

Web fetch and web search tools for enhanced information retrieval
Model state per task - each task now maintains its own model selection state independently

Changed

Cleaned up chat theme for improved visual consistency
Enhanced webview message handling for better task management

Fixed

Fixed @ mentions when editing messages

[v5.5.6] - 2026-03-05

Changed

Enhanced task state management with improved model selection handling
Updated webview message handler for better task coordination

[v5.5.4] - 2026-03-02

Added

Better follow up question UI for improved user interaction

Changed

Enhanced queue and sending prevention for better chat UX
Improved file list with scrollable container and max height

Fixed

Fixed activity bar icon display
Fixed authentication navigation issues

[v5.5.3] - 2026-02-26

Added

Plan file tools (readPlanFile, listPlanFiles) for better plan management
Image input modality support for enhanced interactions
Plan memory manager for improved plan tracking

Changed

Enhanced UI improvements across multiple components
Updated PostHog telemetry integration
Improved folder mention chip UI

Fixed

Limited background task height for better layout management

[v5.5.2] - 2026-02-26

Added

Background tasks support for multi-chats - now you can run multiple chat tasks simultaneously without blocking each other
Single file accept metrics - track detailed metrics when accepting individual file edits
Model state per task - each task now maintains its own model selection state independently
Bottom anchor for changes and navigation - improved UI with anchored navigation controls
Todo IDE plan view - enhanced plan viewing experience in the IDE

Changed

Chat performance improvements - optimized rendering and message handling for smoother chat experience
UI updates across multiple components for better consistency and user experience
Enhanced file edit review controller with improved functionality
Updated task metadata handling for better state management

Fixed

Fixed plan implementation issues for better plan execution
Fixed chat new lines and paste cursor positioning for improved text input experience
Fixed mention chip alignment for better visual consistency
Auto reject tools and send new message - improved tool rejection workflow

[v5.4.1] - 2026-02-13

Added

Sticky user messages for better chat navigation
Pro model info display in settings
Custom icons utility for improved UI components

Changed

Updated profile page with usage data
Streamlined model names and auto model selection
Improved codebase indexing tool UI
Enhanced task history list
Updated notifications integration
UI updates for code reviews

Fixed

Fixed file path bug in code reviews

[v5.4.0] - 2026-02-12

Added

Open plan in editor functionality for better plan management
Better fuzzy search for improved file context matching

Changed

Cleaner chat interface with improved UI
Better diff view for edit tool
Edit tool improvements with context search enhancements
Minor UI update to task header

[v5.3.4] - 2026-02-03

Fixed

Fixed multi-window authentication synchronization
Fixed beta model access for axon-code-2-pro

[v5.3.3] - 2026-02-02

Fixed

Fixed beta model access for axon-code-2-pro

[v5.3.2] - 2026-02-02

Added

Beta models gating for axon-code-2-pro with backend API integration
Custom icons utility for improved UI components

Changed

Set temperature to 0.2 for OpenRouter provider
Enhanced file edit tool with improved functionality
Updated kilocode models configuration
Cleaned up chat text area component
Updated and optimized test files
UI improvements across chat components

[v5.3.1] - 2026-01-30

Added

Retry button for 5xx streaming failures

Fixed

Fixed use_skill tool being included in system prompt when no skills are available

[v5.3.0] - 2026-01-29

Added

New skill tool functionality for enhanced code assistance
Skill parser and type definitions for skill management
Comprehensive test coverage for skill tool components
Enhanced code index orchestrator with skill integration
State manager for skill tool operations

Changed

Updated tool type definitions to support skill tools
Enhanced assistant message presentation with skill support
Improved code index manager with skill-related functionality
Updated dependencies in package.json

[v5.2.9] - 2026-01-27

Changed

Minor update to system prompt for long running command tool

[v5.2.8] - 2026-01-27

Fixed

Minor patch in line diff counter logic

[v5.2.7] - 2026-01-27

Changed

Minor update to system prompt for long running command tool

[v5.2.6] - 2026-01-21

Changed

Minor update to system prompt for long running command tool

[v5.2.5] - 2026-01-21

Added

Multi-line chip rendering for improved mention display

Changed

Clean up profile page
Enhanced chat rendering utilities

Fixed

Fixed plan hover popup display

[v5.2.4] - 2026-01-19

Fixed

Fixed SourceControlPanel.tsx - addressed review comments
Fixed chat text UI issues

[v5.2.3] - 2026-01-16

Fixed

Fixed chat text UI with improved component structure

[v5.2.2] - 2026-01-16

Added

Model selector moved to improved location

Changed

UI improvements and IDE theme fixes
Enhanced ChatTextArea component
Updated translations for all supported languages
Improved select-dropdown and popover components

[v5.2.1] - 2026-01-15

Added

Updated translations and added memories locale files for internationalization support

Fixed

Fixed issue where messages were being queued after rejecting exec_cmd tool

Changed

Code cleanup and improvements

[v5.2.0] - 2026-01-14

Added

New Auto-Generated memories for past chats: Orbital can now generate and reference memories in chats when building or updating codebase to ensure any previously used context can be quickly remembered.

Changed

Removed marketplace
Minor fixes to stream tool calling in cases where tools got stuck

[v5.1.0] - 2026-01-12

Added

Line counter tracking for file edit reviews (lines added, updated, deleted)
Server integration to report line change statistics
Automatic language detection from edited files for analytics
MatterProgressIndicator component with animated loading dots
"View Diff" button in file edit review to view pending changes in VS Code editor
Elapsed time tracking for reasoning blocks (shows "Thinking for Xs" during streaming)
"Thought" translation key for completed reasoning blocks

Changed

Enhanced file edit review accept all functionality with detailed metrics
Improved webview message handling for file edit operations
Updated "API Request..." label to "Generating..." for better UX
Fixed sendingDisabled behavior when canceling tasks or aborting commands
Enhanced reasoning block UI with time display and collapse functionality

Fixed

Removed debug console.log statement from kilo config loading
Fixed unused parameter warning in ReasoningBlock component

[v5.0.3] - 2026-01-11

Added

Context window usage tracking
Show credits usage on hover
Fetch and show chat title

Changed

UX improvements for chat
Headers for repo

Fixed

Notify LLM if files have been changed by the user post its own edits
On exec tool reject, do nothing

[v5.0.2] - 2026-01-09

Changed

Minor improvements to file read tool
Settings UI Cleanup

[v5.0.1] - 2026-01-09

Changed

Minor improvements to file edit and code reviews
File changes list does not dissapear on task abortion

[v5.0.0] - 2026-01-05

Added

Harness context agents and file edits are revamped to reduce error rates! Tools updates, read file, edit file, search and list.
Enterprise Code Review support for self-deployed platform
Improvments in AI code review system

Changed

Major UI improvments
Improved auth flow + review only mode

[v4.210.0] - 2026-01-02

Added

AI Code Reviews

1-click AI code reviews for all your agentic and manual changes
Get all the review comments, suggestions and 1-click Apply all

Changed

Improved ripgrep with file extension normalising
UI improvments

[v4.206.0] - 2025-12-29

Added

Enhanced file edit tool with substring replacer for inline edit file tool calls
Improved task handling with better file edit integration
Enhanced ripgrep service for better file search capabilities

Changed

Updated file edit tool implementation for more robust inline editing
Improved task processing with enhanced file edit capabilities
Enhanced package version management

[v4.205.0] - 2025-12-26

Added

New chat renderer utility for improved message rendering
Retry button functionality for API streaming failures
Enhanced checkpoint handling system

Changed

Migrated KiloTaskHeader and TaskItem components to use ReadOnlyChatText
Refactored ChatTextArea component with code cleanup
Improved ChatRow component with better checkpoint integration

Fixed

Fixed duplicate code removal in ChatTextArea
Resolved race condition with isUserInput flag
Fixed previous commands still showing run/reject buttons
Improved UI consistency and functionality

[v4.204.1] - 2025-12-20

Added

CLI authentication wizard with browser-based OAuth flow
Browser authentication utilities for secure token management
Welcome message utilities for CLI user onboarding

Changed

Refactored AcceptRejectButtons component for improved multi-edit handling
Enhanced ChatTextArea with better cancel button functionality
Updated ChatView component for improved message queue management
Improved Logo component with better styling and responsiveness
Enhanced AuthWizard with comprehensive authentication flow
Updated CLI package configuration and validation

Fixed

Fixed code paths return value issue in AuthWizard
Improved cancel button UI and functionality
Enhanced CLI authentication initialization process

[v4.204.0] - 2025-12-20

Added

Support for axon-code-2-preview model in OpenRouter provider
Enhanced file edit review controller with better diff handling
Improved UI refactors across multiple components

Changed

Refactored kilocode-models configuration for better model management
Updated useOpenRouterModelProviders hook for enhanced provider integration
Enhanced FileEditReviewController for improved file edit reviews
Improved QueuedMessages component for better message handling
Updated CSS styling for consistent UI appearance
Enhanced slash commands utilities for better command processing

Fixed

Cleaned up AI code reviews card for better user experience
Fixed various UI inconsistencies across components
Improved error handling in webview message handler

[v4.203.0] - 2025-12-19

Added

Enhanced chat text area with improved mention chip functionality
New mention chip demo component for better user interaction

Changed

Refactored ChatTextArea component for better performance and maintainability
Updated ChatRow component with streamlined UI elements
Improved CodeAccordian component for enhanced code display
Enhanced useOpenRouterModelProviders hook for better model management
Updated prettyModelName utility for improved model name formatting
Refined KiloTaskHeader component with better task management
Improved GhostServiceSettings component for enhanced configuration

Fixed

Better handling of chat text area interactions
Improved mention chip display and functionality
Enhanced UI cleanup and consistency across components

[v4.202.0] - 2025-12-18

Added

Enhanced UI components with improved user experience
Better error handling and display components
Improved model provider integration

Changed

Streamlined ChatRow component with enhanced functionality
Updated ErrorRow component for better error display
Refined AcceptRejectButtons component for improved user interaction
Enhanced CodeAccordian component for better code presentation
Improved ToolUseBlock component for better tool usage display
Updated ModelSelector component for enhanced model selection
Enhanced select-dropdown component for better user experience
Improved prettyModelName utility for better model name handling

Fixed

Better handling of fill-diff viewer edge cases
Enhanced UI consistency and visual improvements
Better error handling in various components

[v4.201.0] - 2025-12-16

Added

Enhanced file edit review functionality with improved user experience
Better webview message handling for file edit operations

Changed

Improved file edit review interface with refined accept/reject buttons
Updated ChatTextArea component for better user interaction
Enhanced OpenRouter provider integration
Streamlined UI components for improved consistency
Updated translations for ar, ca, cs, de, es, fr, hi, id, it, ja, ko, nl, pl, pt-BR, ru, th, tr, uk, vi, zh-CN and zh-TW

Fixed

Better handling of file edit review edge cases
Improved localization support for chat components

[v4.200.0] - 2025-12-15

Added

New file edit review system with accept/reject functionality
Enhanced monthly quota management within chat interface
Improved mentions cleanup and handling
New AcceptRejectButtons component for better user interaction

Changed

Updated UI components across dialogs, dropdowns, popovers, and selects
Enhanced chat interface with improved user experience
Refined chat row component for better readability
Updated tool use block styling for improved visual consistency
Streamlined chat message layout and interactions

Fixed

Better handling of file edit operations
Improved quota tracking and display
Enhanced error handling in chat interface

[v4.123.1] - 2025-12-07

Changed

Refined chat UI components with improved styling and consistency
Updated todo list display with better iconography using CheckCircle components
Enhanced timestamp display with improved font sizing
Streamlined chat message layout by removing unnecessary borders
Improved color scheme consistency across UI components
Updated font weights and sizes for better visual hierarchy

Fixed

Better handling of todo list status indicators
Improved color variable references for consistent theming

[v4.123.0] - 2025-12-03

Added

Enhanced chat interface with improved user experience
Cleaner message display and layout optimization

Changed

Refined chat row component for better readability
Updated tool use block styling for improved visual consistency
Enhanced internationalization support for chat components

Fixed

Improved message formatting and spacing
Better handling of long messages in chat interface

[v4.122.0] - 2025-12-03

Added

Enhanced credit management system with improved warnings
Better error handling for insufficient credit scenarios
Comprehensive test coverage for error utilities

Changed

Updated credit warning messages to be more user-friendly
Improved low credit warning UI components
Enhanced error messaging for better user guidance

Fixed

Better handling of credit limit scenarios
Improved error message display and formatting

[v4.121.0] - 2025-11-30

Added

Refined codebase search functionality with improved accuracy
Enhanced search result filtering and scoring
Better integration with codebase indexing system

Changed

Optimized search result limits (reduced from 200 to 5 max results)
Improved search score thresholds (minimum score: 0.5)
Enhanced model provider integration
Streamlined codebase search UI components

Fixed

Better handling of search result pagination
Improved search result display formatting
Enhanced model selection interface

[v4.120.0] - 2025-11-27

Added

Complete codebase indexing system with backend integration
Vector store migration to backend services
Enhanced embedding endpoint functionality
Improved package cleanup and optimization

Changed

Migrated vector store operations to backend for better performance
Refactored codebase indexing architecture for scalability
Enhanced UI components for better user experience
Improved model provider configurations

Fixed

Resolved embedding endpoint connectivity issues
Fixed codebase indexing performance bottlenecks
Enhanced error handling in vector store operations

[v4.119.0] - 2025-11-20

Added

Hardcoded model IDs for improved Axon integration
Enhanced tool optimization for better performance
Credit usage display per model for better cost tracking
Improved model selection interface

Changed

Optimized API calls to reduce latency
Enhanced model provider architecture
Improved credit management and display
Streamlined tool execution pipeline

Fixed

Better handling of model selection edge cases
Improved credit tracking accuracy
Enhanced error handling in model provider operations

[v4.118.0] - 2025-11-13

Added

Initial release with core functionality
Basic chat interface and model integration
Foundational codebase indexing capabilities

Features

AI-powered code assistance
Multi-model support through OpenRouter
Basic autocomplete and code suggestions
Initial marketplace integration

FilesExpand file tree

CHANGELOG.md

Latest commit

History

CHANGELOG.md

File metadata and controls

Changelog

[v7.0.0] - 2026-10-09

Changed

Added

[v6.9.0] - 2026-10-03

Changed

Added

Notes

[v6.8.6] - 2026-09-15

Fixed

[v6.8.5] - 2026-09-04

Added

Fixed

[v6.8.4] - 2026-09-03

Added

Changed

Fixed

[v6.8.2] - 2026-08-28

Added

Changed

Fixed

[v6.8.1] - 2026-08-21

Changed

[v6.9.0] - 2026-08-04

Added

[v6.8.0] - 2026-08-01

Removed

[v6.7.2] - 2026-07-31

Added

Changed

[v6.6.9] - 2026-07-25

Changed

[v6.6.8] - 2026-07-24

Changed

Added

[v6.6.7] - 2026-07-22

Added

Changed

[v6.6.3] - 2026-07-21

Changed

[v6.6.1] - 2026-07-21

Added

Changed

[v6.6.0] - 2026-07-21

Added

Changed

Fixed

[v6.4.9] - 2026-07-21

Fixed

Added

[v6.4.7] - 2026-06-30

Added

Changed

[v6.4.6] - 2026-06-25

Added

Changed

Fixed

[v6.4.4] - 2026-06-23

Added

Fixed

Changed

[v6.2.3] - 2026-05-21

Fixed

Changed

[v6.2.0] - 2026-05-21

Added

Changed

Fixed

[v6.0.0] - 2026-04-10

Highlights

Added

Agent Manager

Exploration Groups

Model Selection