Latest Version

​

Download

• Windows x64
• Windows x64 (system Python/Git)
• GitHub Releases

​

Changes

​

v0.8.2 (Added)

• Added GPT-6 Astra to ChatGPT subscription models, with reasoning effort and Fast mode controls. The configured context window now applies to all subscription models.
• Added a per-session Multi-Agent control in the model selector. The selection is preserved across reloads, split panes, detached windows, and session forks.
• The Python SDK can list, query, wait for, and cancel tasks; send messages between parent, child, and sibling Agents; and resume failed or cancelled subagents in their existing conversations. Tasks support short names and full output logs. Background execution is enabled by default on upgrade and can be disabled in Experimental settings.
• Explorer can inspect live Unity data through bounded, read-only C# queries, alongside code and asset searches.
• Archived sessions and Unity Inspector content now open in Workbench editor tabs. Unity selections and file attachments are sent to the most recently focused composer in the matching workspace.

​

v0.8.2 (Fixed)

• Aligned Codex subscription requests with the current Responses protocol, including tool namespaces, WebSocket prewarming, HTTP fallback, and compressed HTTP requests. Reopened and reconnected conversations preserve canonical response items and encrypted reasoning; failed or incomplete responses stop before tool execution.
• Corrected remote-compaction retention budgets for text and images, and applied configured subscription context limits consistently in the frontend and backend.
• Subagents now use their configured model, provider, credentials, reasoning effort, and speed settings instead of inheriting the parent provider route.
• Background results and Agent messages are persisted before delivery, retry without duplicate notifications, and survive foreground-round finalization. Session migrations preserve older context exports and explicitly mark missing historical fields as empty.
• Kept Agent pages, Workbench editors, session controls, and detached windows bound to the selected workspace. File previews revalidate after disk changes while preserving unsaved edits.
• Fixed virtual console row measurement, scrolling, and resizing with large log histories. Added runtime stall and bridge diagnostics.
• Unity recompilation results remain available across retries and are matched to the originating operation. Plan mode now blocks recompilation, test execution, and Editor state changes while allowing read-only inspection.
• Consolidated Unity guidance around verified edit targets, Prefab instance scope, structured asset edits through Unity APIs, and the existing authorization for the task.

​

v0.8.2 (Removed)

• Removed the standalone get_task_status and cancel_task tools; task controls are built into the Python SDK.
• Removed the separate legacy Inspector surface and archived-session Settings page after moving these workflows into the Workbench.
• Removed superseded bundled Unity rule files. Upgrades clean up the retired bundled copies while preserving user Agent definitions and project overrides.
• Removed the retired dev Agent compatibility path. Its definitions, rules, and model preferences are no longer loaded or merged into Unity; upgrades remove the bundled agent/dev directory.

​

v0.8.1 (Added)

• Text zoom and Markdown editing. Ctrl+wheel text zoom is available in Markdown editors and previews, text viewers, diffs, and merge views. Workspace Markdown files now open in rendered CodeMirror mode.

​

v0.8.1 (Fixed)

• Restored missing Workbench session editor capabilities, including compaction, interrupted-run resume, Plan mode, full-session and message-level forks, history navigation, context export and review, and pane-owned Thinking and Changes panels.
• Scoped session panes preserve drafts, execution settings, todos, background-task updates, undo and diff state, scroll history, and session actions across editor remounts.
• File, Knowledge, Unity, and native-drop actions remain bound to their originating workspace and session.
• Knowledge document outlines align with the document body and available viewport.
• Restored system resources return to the beginning of the workspace root.
• ONNX Runtime preparation resumes partial downloads and validates package and runtime integrity.
• Improved keyboard access for Changes panel resizing, undo selection, command lists, and close controls.

​

v0.8.1 (Removed)

• No user-facing workflows or compatibility paths were removed.

​

v0.8.0 (Added)

• Concurrent multi-workspace operation. Multiple projects and Git worktrees can run in one Locus instance, with parallel sessions, Unity connections, Knowledge indexes, asset databases, and workspace services. Each workspace keeps an independent execution context, runtime state, and resource cache.
• VS Code-style split workbench. Sessions, files, Knowledge documents, Unity assets, and Views open as editor tabs. Tabs can be reordered, split in four directions, moved between panes, or detached into native windows. Window layouts, tabs, drafts, scroll positions, and unsaved content can be restored.
• Development workspace tree. Sessions, Knowledge, collaboration, and files now share one workspace tree, with single- and multi-workspace layouts, saved presets, virtual folders, mounted directories, and per-node visibility controls.
• Persistent Markdown editing. Knowledge documents now use CodeMirror 6 for Live Preview and source editing, preserving selections, undo history, scroll position, drafts, and unsaved changes across document switches and split panes.
• Expanded Knowledge workflows. Added Plan as a fifth knowledge type, writable project Skill packages under Locus/skills, automatic Skill refresh, document outline navigation, cached loading, and conflict-aware collaborative editing.
• Workspace and Unity mentions. Chat @ search now covers workspace files and folders as well as live Unity scene objects, with ranked results and explicit object references.
• Richer Unity inspection. unity_yaml_read can display selected hierarchy metadata, including active state, Prefab source, world transforms, RectTransform bounds, tags, layers, and components. SubScene reads can follow the authoring Scene hierarchy, and adjustable serialized fields can render as inline property controls.
• Unity coordination and incremental synchronization. A workspace change journal selects targeted imports, full refreshes, or no-op recompiles from verified file events. Experimental unity_lock and unity_release coordinate multiple Agents sharing one Editor.
• Runtime and provider controls. Sub-agent defaults now include model, reasoning effort, and response speed. Custom Responses providers can enable Codex V2 remote compaction, with normalized reasoning replay for supported DeepSeek V4 routes.
• Session SDK and diagnostics. The Python SDK can list running sessions and send sourced input to another active run. Debug mode adds DevTools control, CDP port discovery, WebView bridge diagnostics, and bounded log batching.

​

v0.8.0 (Fixed)

• Strengthened isolation for session events, Unity services, MCP, C# analysis, asset databases, Knowledge, diff/merge state, Views, and caches across workspaces.
• Switching workspace scopes now refreshes the Unity plugin install or update notice. Late status checks and transient failures no longer overwrite newer event-driven notices.
• Preserved streaming sessions, input drafts, pending confirmations, queued attachments, image previews, and unsaved files while moving tabs, transferring windows, or remounting editors.
• Failed, cancelled, or backend-revoked optimistic sends now restore complete local-file attachments and absolute paths to the composer.
• Failed or timed-out window transfers now leave the tab and editor state safely in the source window.
• Manual context compaction can queue behind an active run and executes before queued follow-up input. Child Agent sessions remain independently controllable.
• Responses API citations now persist through session storage and context export. Incremental UTF-8 decoding preserves multibyte characters split across streaming chunks.
• Streaming Markdown, syntax highlighting, collapsed tool interaction, and tool file previews remain stable while content grows or finishes streaming.
• Interrupted or failed sub-agents return persisted partial output. Deferred tool schemas load by exact wire name, and image tools validate model vision support before execution.
• Unity recovery now exposes Safe Mode and Editor log paths, waits for modal-dialog transitions, and cancels interrupted Unity Test runs through the Test Framework lifecycle.
• Collapsed tool rows use concise UI labels, while expanded rows retain exact monospace tool identifiers.

​

v0.8.0 (Removed)

• Removed the legacy Project entry and second-level workspace bar; the Development workbench and workspace tree now provide the unified navigation model.
• Removed the Vditor-backed Markdown editor and its legacy rendering and compatibility paths after the CodeMirror migration.
• Removed the manual skill_reload tool; Locus now refreshes Skill manifests automatically after file changes.

​

v0.7.4 (Added)

• Configurable Codex context windows. ChatGPT subscription models now accept context windows from 16K to 1M tokens, defaulting to 272K. Effective context and automatic compaction limits scale with the configured value, and legacy Extended Context settings migrate to 372K.
• Prompt cache diagnostics. Session context usage now records cache invalidations from provider-reported usage, including model or provider changes and abnormal uncached input growth. These records are included in context exports, with optional in-app warnings available under Notifications.
• Prefab-aware Unity Property Tree. unity_yaml_read now identifies live Editor and disk YAML sources, supports configurable array limits, merges Prefab source and Variant data, marks overridden properties, folds identical repeated Prefab instances, and returns readable semantic subasset paths.
• Safer structured output from unity_execute. printJson now supports anonymous objects, stored fields, dictionaries, materialized collections, shared and cyclic references, and nested Unity objects. Deferred enumerables and target property getters remain unevaluated.
• Fast development installers. bun run build:installers creates incremental zlib-compressed development installers. Release installers retain ThinLTO optimization and both embedded-runtime flavors.

​

v0.7.4 (Fixed)

• Unity recompilation lifecycle. Recompile requests now wait for Unity’s compilation-start event, persist intermediate state across domain reloads, reconcile after bridge reconnection, and report detailed timeout diagnostics.
• Knowledge query output. Summary results use one line per document, detailed results use compact path-and-line headers, and frontmatter matches expose only the relevant summary or maintenance-rule value. App Skill packages are grouped under their real application roots, and Skill caches refresh after knowledge or plugin changes.
• Tool-call handoffs. Provisional provider tool calls omitted from the completed response no longer appear as empty completed calls.
• Rendered Markdown editing. Activating the editor from rendered content preserves the scroll position, document height, and clicked caret location.
• View runtime imports. Default .vue imports remain stable with TypeScript 6 CommonJS interop.
• File-read context. Absolute line numbers now use compact prefixes, reducing context usage and matching compaction excerpts.

​

v0.7.4 (Removed)

• No user-facing workflows or compatibility paths were removed.

​

v0.7.3 (Added)

• Codex automatic approval review. Codex users can opt into a structured review of pending tool actions. Reviews use the latest user request and bounded local inspection data; safe actions can proceed automatically, while escalated actions show the review rationale and risk level.
• Dangerous command interception. Forced deletion patterns in Bash, PowerShell, and CMD request approval by default.
• Four-state knowledge editing. Knowledge documents now support inherited, disabled, confirmation-required, and automatic AI editing modes, with legacy metadata migration and consistent enforcement across knowledge and filesystem tools.
• Local simulated models. Debug mode exposes deterministic streaming, tool-call, and error-response presets without network access.
• Atomic multi-edit batches. The edit tool can apply multiple independent replacements against one original file snapshot and commits the update only when every replacement is valid.
• Combined Unity Test discovery. Unity Test tools discover Edit and Play Mode tests together by default and render assembly, suite, and test paths as a tree.
• Staged context review. Session context review creates a dedicated chat, exports the context asynchronously, and places the resulting file in the composer for inspection before sending.
• Testing workflows. Added isolated-instance onboarding bypass and WebView stress drivers for streaming, session switching, and live tool calls.
• Unity Property Tree. unity_yaml_read and unity_yaml_search now use one path-addressable property tree for scene hierarchies, components, subassets, and serialized fields, returning child paths for continued inspection when output exceeds the current budget.
• Agent customization. User Agents and tool-description overlays can load from user-agents; the Agent panel can control environment, rule, knowledge-context, and lazy-tool prompt injections and reload changed definitions immediately.
• Session context usage. A context-usage window breaks token consumption down across system prompts, runtime injections, messages, and tool results, with per-tool distribution and effective model output speed.
• Per-Agent model defaults. Model and reasoning effort defaults persist independently for each Agent, while Codex and custom providers expose prompt-prefix cache lifetime controls.

​

v0.7.3 (Fixed)

• Abnormal web-search continuation. Fixed Codex hosted web_search results being replayed as local tool calls, which could trigger duplicate follow-up turns.
• Windows handle leak. Updated to the upstream-fixed Tauri/Wry runtime so window-border DPI queries no longer leak Windows device-context handles.
• Windows Rust test startup. Embedded the Common Controls v6 manifest in Rust unit-test binaries, preventing STATUS_ENTRYPOINT_NOT_FOUND while loading TaskDialogIndirect.
• Child-process lifecycle. Shell, Skill, development-server, and other child process trees are tracked by session, workspace, and application lifecycle and terminate with the owning operation.
• Read-only additional directories. File writes, Shell working directories, and command path targets now consistently enforce read-only additional-workspace constraints.
• Message layout stability. Improved message layout stability and fixed most flicker caused by streaming chat rendering, covering Markdown height reconciliation, stream-end scrolling, tool state settlement, and expansion-state retention.
• Parallel tool execution. Workspace mutations use path-aware coordination, same-file edits are batched safely, independent writes can overlap, and read-only subagents can run alongside compatible tools.
• Hosted tools and background tasks. Provider-hosted tool calls stay out of client function history, completed hosted rounds stop correctly, immediate async startup failures return directly, and notify-mode results arrive without status polling.
• Image-only chat turns. Compaction, paging, and turn navigation retain image prompts and exclude internal reminder turns.
• Shell and search environment. Search paths resolve consistently against the workspace, PowerShell guidance reflects the detected runtime, and pwsh is directly available from the Bash environment.
• Unity hot-reload feedback. Edit feedback appears only while the hot-reload pipeline is active.
• Knowledge metadata refresh. Document file metadata remains visible while save responses reconcile with the following refresh.
• History graph layout. The workspace WIP lane stays anchored to HEAD ahead of descendant branch commits.

​

v0.7.3 (Removed)

• No user-facing workflows or compatibility paths were removed.

​

v0.7.2 (Added)

• Live background task updates. Async and notify tool calls now update their original tool block as execution progresses. Shell output streams live, final results persist in session history, and background shell and Unity Test runs continue until completion or cancellation without synchronous execution deadlines.
• Built-in Unity project-file generation. The Locus Unity plugin can generate complete .sln and .csproj files directly from Unity’s CompilationPipeline without Visual Studio, Rider, or VS Code integration packages. Generated workspaces retain project references, compilation defines, analyzers, and Unity project-generation callbacks.

​

v0.7.2 (Fixed)

• Reduced premature session termination. Agent runs now continue after server-tool rounds and when Codex/Responses returns end_turn: false, reducing cases where a session ends before the model produces its final response.
• Codex request compatibility. Removed the unsupported max_output_tokens parameter from Codex requests, preventing the server from rejecting checkpoint-compaction requests.
• Codex context compaction. Codex sessions now use Remote Compaction V2 over the streaming Responses protocol with complete session identity and routing headers. Server compaction failures preserve the original context and stop the current run, preventing incompatible checkpoint state.
• Compacted-session forks. Forks created before the latest compaction reconstruct the correct active prompt window, and forks from assistant tool groups retain their matching outputs. This fixes empty fork context, failed compaction, and incorrect history selection anchors.
• Nested subagent stack overflows. Each child agent now runs its full loop in an independent Tokio task, separating the parent tool loop, tool execution, and child LLM loop onto distinct task stack boundaries. Ending the parent execution also aborts the child task.
• Concurrent agent scheduling stalls. Foreground subagents in mixed tool rounds complete before local sibling tools run. Workspace mutations are coordinated by normalized file path: writes to different files can overlap, same-path writes serialize, and read-only tools bypass mutation waits. Long waits identify the blocking sessions and tools.
• Unity parallel-edit deadlocks. Completed asset changes are imported whenever any overlapping edit session ends, while remaining sessions continue normally. This prevents finished files from remaining outside the AssetDatabase behind a long-running sibling session.
• Interrupted tool history. Resuming, forking, or continuing interrupted sessions recovers missing local tool results from recorded output and inserts explicit interruption results when necessary. A schema migration repairs terminal tool rounds in older sessions, keeping model requests and context exports valid.
• Persistent error notifications. Error notifications remain visible until dismissed explicitly. Success replacement, per-operation clearing, and global clearing preserve unresolved errors.
• Unity C# snippet compilation. Common System.IO types now use explicit aliases in unity_execute and unity_run_states, preventing Unity Mono’s internal MonoLinqHelper.ToArray from conflicting with LINQ ToArray when non-public access is enabled.
• Unity YAML inspection guidance. When a direct read is redirected from a Unity YAML asset, Locus now recommends both semantic YAML tools and unity_execute for Editor-side C# inspection, while retaining the explicit raw-read retry path.

​

v0.7.1 (Added)

• Live background task updates. Async and notify tool calls now update their original tool block as execution progresses. Shell output streams live, final results persist in session history, and background shell and Unity Test runs continue until completion or cancellation without synchronous execution deadlines.
• Built-in Unity project-file generation. The Locus Unity plugin can generate complete .sln and .csproj files directly from Unity’s CompilationPipeline without Visual Studio, Rider, or VS Code integration packages. Generated workspaces retain project references, compilation defines, analyzers, and Unity project-generation callbacks.

​

v0.7.1 (Fixed)

• Nested subagent stack overflows. Each child agent now runs its full loop in an independent Tokio task, separating the parent tool loop, tool execution, and child LLM loop onto distinct task stack boundaries. Ending the parent execution also aborts the child task.
• Concurrent agent scheduling stalls. Foreground subagents in mixed tool rounds complete before local sibling tools run. Workspace locks are scoped to normalized project paths, allowing separate projects to execute independently. Long lock waits identify the waiting tools and blocking sessions.
• Unity parallel-edit deadlocks. Completed asset changes are imported whenever any overlapping edit session ends, while remaining sessions continue normally. This prevents finished files from remaining outside the AssetDatabase behind a long-running sibling session.
• Interrupted tool history. Resuming or continuing an interrupted session now records explicit interrupted results for unfinished local tool calls, preserving a valid tool-call sequence.
• Unity C# snippet compilation. Common System.IO types now use explicit aliases in unity_execute and unity_run_states, preventing Unity Mono’s internal MonoLinqHelper.ToArray from conflicting with LINQ ToArray when non-public access is enabled.
• Unity YAML inspection guidance. When a direct read is redirected from a Unity YAML asset, Locus now recommends both semantic YAML tools and unity_execute for Editor-side C# inspection, while retaining the explicit raw-read retry path.

​

v0.7.0 (Added)

• Python Agent SDK. The bundled standard-library-only locus package exposes models, agents, tools, workspaces, sessions, custom Python callbacks, asynchronous runs, events, cancellation, input handling, and reusable workflow orchestration.
• Expanded Unity automation. Added dedicated Play/Edit Mode control, multi-level Console queries, Unity Test Framework discovery and execution, external script-editor integration, richer YAML inspection, improved viewport capture, and broader CLI parity testing.
• Graphics and UI debugging workflows. Added RenderDoc and Unity Frame Debugger capture, texture analysis, and a UI Toolkit API for inspecting, querying, interacting with, and serializing live panels.
• Runtime debugging. Added cooperative PlayerLoop debugging with cross-frame conditions, pause, frame stepping, inspection, and native Windows debugger discovery for CDB and WinDbg workflows.
• Session and workspace navigation. Pages and sessions can open in standalone windows. Added user-turn navigation, paginated history, interrupted-run resume, context export/review, compacted-context inspection, file previews, and model usage statistics.
• Agent execution controls. Added parallel tool scheduling with workspace locks, same-file edit batching, standardized subagent execution, optional background tasks with status/cancellation, inline todos, and local tool-failure logs.
• File-driven knowledge workflows. Design, Memory, Reference, and Skill now use file-driven Markdown operations, with a unified knowledge panel, Markdown previews, canonical Skill packages, /grill, and /review-context.
• Codex and provider controls. Added extended Codex context settings, optional GPT-5.6 session-title generation, native deferred tool loading for compatible custom providers, and provider-specific compaction output limits.
• Checkpoint compaction. Long sessions now compact into an anchored summary plus recent verbatim context, retain structured migration/export data, and expose the checkpoint contents in the compaction viewer.

​

v0.7.0 (Fixed)

• Hardened Codex WebSocket recovery, proxy handling, compressed and continued frames, stale request cleanup, continuation state, and compaction synchronization.
• Unity execute requests can reattach after native bridge disconnects or domain reloads without running accepted requests twice.
• Improved Unity YAML parsing, serialized-value fidelity, prefab/reference handling, viewport capture, and cold-path inspection safety.
• C# diagnostics now target Unity-managed source files and append semantic warnings to hot-reload or recompile results for the current edit batch.
• Markdown file references cache path status, handle folders correctly, and avoid repeated filesystem probes during rendering.
• UI Toolkit panel snapshots refresh explicitly, stale panels are rejected, and JSON serialization has bounded depth, members, collections, strings, and total output.
• Replaced native select controls with the shared dropdown implementation and reused the provider configuration flow during onboarding.

​

v0.7.0 (Removed)

• Removed the legacy Agent Graph, graph_view, and sheet workflows. Agent delegation now uses subagent, and custom interactive output uses View and Skill packages.
• Removed dedicated knowledge_* CRUD tools and the old standalone Todo panel. Knowledge changes use unified file-driven operations, while todos render directly in the session.

​

v0.6.0 (Added)

• MCP servers. Connect external MCP servers over stdio or HTTP and use their tools from agent sessions. A new MCP settings page manages servers and their tools; MCP tool calls show confirmation cards and status in chat. Server entries can be imported from Claude Desktop, Claude Code, and Cursor configs (imported entries arrive disabled). The new /connect command starts a guided flow for connecting external software, picking between MCP, CLI, and HTTP channels and verifying the connection actually works.
• Multi-model providers and a bundled models.dev catalog. Custom models are now grouped into providers, each hosting multiple models, replacing single custom-endpoint entries. Providers can be added from the bundled models.dev catalog (70+ services with prefilled official endpoints) or configured manually in a two-step dialog with per-field validation. Reasoning options such as enable_thinking and thinking type can be set per model, and the model selector groups custom models by provider. Existing configurations migrate automatically; old model ids keep working.
• External skill discovery. Optionally discover read-only agent skills from Claude Code, Codex, and .agents directories — user-level and per-workspace — plus installed Claude Code plugins. Off by default. Skill activation is now split into separate auto-inject and command channels with per-skill controls; external command skills support $ARGUMENTS and $1–$9 substitution.
• Extra working directories. A workspace can attach additional directories through its context menu; file tools can read and edit within them, and the workspace selector shows attached paths inline.
• Unity object preview straight from disk. Files Unity does not serialize (markdown, plain text, source code, binaries) now preview from disk, so previews keep working for assets Unity has not imported — such as AI-generated files under symlinked folders. Preview folding is now optional, and the asset inspector keeps previews expanded. (#114)
• Copy selection in chat. Right-clicking selected text in a message offers copying the selection as the first menu item.
• View log bar toggle. The log bar in view windows is hidden by default and can be enabled in display settings; the choice syncs across windows.

​

v0.6.0 (Fixed)

• Command-tool friction on Windows: managed Python environment variables now apply only to Python invocations instead of the whole command environment; command output is decoded using the system ANSI code page, fixing garbled Chinese (GBK) output; long-output truncation keeps both head and tail instead of only the head.
• Deferred tools now load through Anthropic native tool references and Codex native tool search, replacing the previous meta-tool activation.
• The builtin Unity profiler sampling skill now defaults to summary injection so agents can discover it on demand; it previously shipped with injection off.
• Helper windows (plan review, diff review, import dialogs, and others) open through a pre-warmed window pool and appear faster.
• UI locale catalogs are code-split and loaded on demand.
• Session and knowledge-index databases vacuum automatically once free pages dominate the file, reclaiming disk space.
• Context menu items now carry icons with a unified layout.
• Plugin toggles apply immediately in the plugin list instead of waiting for the backend round-trip.
• The scroll wheel no longer changes focused number inputs in settings.

​

v0.6.0 (Removed)

• The single-endpoint custom model configuration is gone, replaced by the multi-model provider dialog; existing custom endpoints migrate automatically.

​

v0.5.8 (Added)

• Standalone plan review. Plan approvals can open automatically in a dedicated window selected from Display settings. The window supports reviewing the Markdown plan, approving implementation, or sending it back with an optional note.

​

v0.5.8 (Fixed)

• Resolved intermittent DeepSeek HTTP 400 failures on Windows. Upgraded schannel from 0.1.28 to 0.1.29, preventing pending request data from being resent during TLS 1.3 renegotiation and corrupting large JSON bodies. This resolves the 400 Bad Request: Failed to parse the request body as JSON failures tracked in issue #48. (PR #110)
• Plan approval flow. Approval cards now render Markdown, attach optional feedback to Send back, and preserve approved plans or send-back feedback in the transcript with the correct lifecycle state. Selecting /plan in an idle session immediately enters sticky Plan mode.
• Fast-capable Codex models now show Fast in their model labels while Fast mode is enabled.
• Exact edit tool contracts. edit and knowledge_edit now expose one exact replacement per call. knowledge_read can return the stored summary, body, or maintenanceRules section verbatim—including whitespace and empty content—so edits can match the selected section exactly. Existing persisted batch file-edit calls remain executable.

​

v0.5.7 (Added)

• GPT-5.6 models and Fast mode. ChatGPT subscriptions now include GPT-5.6 Sol, Terra, and Luna with a 353,400-token Codex runtime context budget and reasoning levels up to Max. The model picker also offers Fast mode for faster responses with increased subscription usage.
• Manual Codex usage-limit resets. ChatGPT subscription settings list available usage-limit reset credits and expiry dates. After confirmation, a credit can be consumed to reset currently eligible Codex usage windows and refresh the displayed quota.
• Chat input history. The chat composer recalls previous user messages with Up/Down, restores their attachments and composer intent, and returns to the current draft at the end of the history.
• Double-press shortcut to stop responses. A configurable shortcut now stops the active response after a second press, with Esc as the default; the first press shows a confirmation notice.

​

v0.5.7 (Fixed)

• Unity documentation imports now recognize the current cloudmedia-docs.unity3d.com download host and legacy Google Storage URLs.
• Escape closes active previews, popovers, and diff overlays before the global response-stop shortcut is evaluated.

​

v0.5.6 (Added)

• Import local files as knowledge references. External sources can now import a local file or folder as a reference folder, in two modes: live link mounts the path through an OS link (junction/symlink), so content always reflects the latest files on disk and stays read-only; snapshot copies the current content once, with optional AI editing and a manual re-sync action. Markdown and text documents (.md / .markdown / .txt) become searchable in knowledge. Deleting a live-linked reference only detaches the link — source files are never touched. (#85)
• LaTeX math in chat. Messages render LaTeX formulas with KaTeX, recognizing $…$, $$…$$, \(…\), and \[…\] delimiters.

​

v0.5.6 (Fixed)

• Chat-completions requests are sent byte-for-byte as serialized, and request logs capture the same bytes that went on the wire. When a provider rejects a request with a JSON parse error, the message now explains that the body was most likely corrupted in transit (proxy, VPN, security software, or a gateway) and suggests a curl replay to confirm. (#106)
• Long streaming replies no longer slow down as they grow: live text, thinking, and the typewriter append new chunks instead of re-rendering the whole message.
• The message list no longer flashes for a frame when a streaming reply finishes.

​

v0.5.5 (Added)

• Per-file revert. The changes panel, the diff review window, and a new row context menu can restore a single file to its snapshot from before the current round, without popping the undo stack. If the file has later edits, Locus asks before rolling them back. (#100)
• Automatic retry for model requests. Connection errors, timeouts, HTTP 5xx, and 429 rate limits retry automatically, honoring Retry-After on 429. Retries only happen before any output has streamed; the attempt cap is adjustable in General settings (default 3, 0 disables). (#101)
• Inline <think> tags routed to the thinking channel. On the chat-completions and OpenRouter paths, inline <think>…</think> segments in streamed replies are stripped from the message body and shown as thinking instead (enabled by default).
• Subagent limits. General settings cap subagent nesting depth and per-session concurrency (default depth 1). Calls over the limit return an error explaining the current caps.
• /dream memory consolidation skill. A built-in skill that merges duplicate memories, fixes stale facts, and prunes the Memory index.

​

v0.5.5 (Fixed)

• Auto-compaction scales its retention and request budgets to the model’s context window and recognizes more provider “prompt too long” errors, so compaction converges on small-context models.
• Replayed tool-call arguments are normalized before being resent to chat-completions/OpenRouter providers — empty or malformed JSON becomes {}; Codex replays stay verbatim. (#94)
• Workspace file paths resolve lexically, so symlinked files open and reveal correctly. (#100)
• Smoother streaming of long replies: finished markdown blocks are frozen and only the tail re-parses during streaming.
• Context token estimates no longer undercount CJK text.
• Auth re-checks no longer tear down open tabs, and views preload to avoid a flash on first visit.

​

v0.5.4 (Added)

• Sticky plan mode. Plan mode is now a persistent, per-session read-only mode: the agent researches and writes the plan but cannot edit files or run non-read-only tools until you approve it from an in-chat card. A status bar shows plan mode is active with a one-click exit, and subagents spawned during planning inherit the read-only restriction.
• Main log written to a file. Locus mirrors its log (including the frontend console) to a rotating file at %APPDATA%\locus\logs\locus.log, flushed on crash. Console settings has a new “Locate Log File” button.
• Model catalog refresh. Added Fable 5 and Sonnet 5 (1M context) and made Opus 4.8 the default; removed the older Sonnet 4.6, Haiku 4.5, and Opus 4.7 entries.

​

v0.5.4 (Fixed)

• Model names no longer show the internal [1m] context-window marker in the pickers.
• Unity YAML parsing no longer panics on a multi-byte character right after a GUID, and single-quoted scalar names are unquoted correctly.
• Scene diffing now detects prefab source swaps and reparenting instead of reporting them unchanged, and skips rebuilding byte-identical scene targets.
• Faster asset re-scans: unchanged files are no longer re-parsed during watcher reconciliation, and the asset database rebuilds once on first launch after updating (schema v13).
• Smoother response streaming: cached diagnostics/trace checks, shallower tool-call watches, batched tool-result updates, and stabilized message groups cut redundant transcript work.
• Hot reload now fails closed on unsafe edits: ref-struct instance members, members referenced inside expression trees, namespace-scoped using changes, and cold-diff reverts.
• Hardened the Unity hot-reload bridge: applied detours roll back on apply failure, recompile completion is gated on a compile epoch so pre-request compiles can’t finish the request, components are skipped on GameObjects being destroyed, and convergence races / monitor-stop handoff are fixed.

​

v0.5.4 (Removed)

• Removed the old one-shot “plan → continue implementation” confirmation card; it’s replaced by the sticky, approval-gated plan mode above.

​

v0.5.3 (Added)

• Tuanjie (Unity China fork) support: Locus now discovers, launches, and connects to the Tuanjie editor (Tuanjie.exe) alongside standard Unity — covering process matching, the native engine module and its PDB, and window-title detection. Standard Unity installs are unaffected.
• Claude subscription support: Sign in to a Claude subscription or import an existing Claude Code / Codex CLI login (including a custom endpoint), and see remaining subscription quota in the Anthropic panel.
• Gateway-truncation diagnostics: Failed Anthropic requests now log transport facts and the upstream JSON parse offset, and when a 400 looks like a mid-transit truncation Locus saves the exact request bytes plus a runnable curl so the failure can be reproduced and reported. (#48)

​

v0.5.3 (Removed)

• Removed the “Claude subscription paused” disclaimer modal and its notice; the subscription sign-in flow replaces it.

​

v0.5.2 (Added)

• unity_code_usages member lookups now read from a serialized member-binding index built during asset scanning, so they catch broken/unbound UnityEvent calls named only by a type string (m_Target: {fileID: 0}, no GUID) and resolve AnimationEvent function names across the whole project, instead of re-scanning the referencing scenes/prefabs and every .anim clip on each query. The asset database rebuilds once on first launch after updating (schema v12).
• Unity editor resolution now falls back to the Unity Hub install cache (editors-v2.json), so editors installed on non-default drives (for example D:\ or F:\) are found.

​

v0.5.2 (Fixed)

• Unity 6.5 (6000.5) compatibility: GetInstanceID() and IsLoadingAssetPreview(int) became errors and the 32-bit InstanceID was replaced by the 64-bit EntityId. A LocusObjectIdentity shim folds the EntityId back to its low 32 bits (keeping the existing InstanceID and wire protocol), and the deprecated build-target APIs are gated by editor version. Older editors are unaffected. (#95)
• Fixed asset-database lock poisoning on CJK input: a bare CJK search term, or a .meta file with CJK text after guid:, sliced inside a multi-byte character and panicked while the asset-database lock was held, which broke every later asset query until restart. Slicing now happens on character boundaries. (#97)

​

v0.5.2 (Removed)

• unity_code_usages member mode no longer reports a UnityEvent call whose serialized type name points to a different class; it now matches only the exact class short-name or the bound script GUID.

​

v0.5.1 (Fixed)

• The grep tool now streams search results instead of reading whole files into memory and keeps only the first N matches in path order; on very broad patterns it stops early and reports that the results are an incomplete, non-deterministic subset.
• Fixed a focus flicker when clicking the input box in the embedded Unity window; the window is only force-activated when it doesn’t already hold the foreground, so the overlay no longer bounces between blur and focus.
• Fixed the native-plugin overlay failing to mount because its control-pipe name wasn’t normalized to the full pipe path, so the client never connected.

​

v0.5.0 (Added)

• Editor hot reload is now available: the agent can modify C# scripts at runtime and apply the new logic live, matching the capabilities of mainstream commercial hot-reload plugins.
• The Unity IPC connection has moved from the C# plugin to a native plugin written in Rust, which keeps the connection stable across domain reloads and while Unity’s main thread is blocked, and reports Unity state in real time.
• C# compilation has moved out of the Unity process into a standalone CoreCLR process; the old path is retained for now as a compatibility fallback.
• The Rust backend’s memory allocator was switched to mimalloc.
• Settings has a new Testing page. This release involves significant architectural changes that may cause compatibility regressions, so if you run into problems you can run these tests for reference.
• Exposed more APIs to the View Package.

​

v0.5.0 (Fixed)

• Asset database performance and correctness improvements.
• Fixed a case where the message history could hide the last tool call.
• The grep tool now resolves paths against the workspace directory.