Hmbown/CodeWhale v0.8.59

on GitHub

CodeWhale is the canonical project, command, npm package, and
release-asset name. The legacy npm package deepseek-tui is
deprecated and receives no further releases. Users coming from
v0.8.x legacy deepseek / deepseek-tui names should migrate
with docs/REBRAND.md.

Install

Recommended — npm (one command, both binaries)

npm install -g codewhale

The wrapper downloads both binaries from this Release and places them in the same directory.

Docker / GHCR

docker run --rm -it \
  -e DEEPSEEK_API_KEY="$DEEPSEEK_API_KEY" \
  -v ~/.deepseek:/home/codewhale/.deepseek \
  ghcr.io/hmbown/codewhale:v0.8.59

The image ships the codewhale dispatcher and codewhale-tui runtime. The latest tag is also updated on release.

Cargo (Linux / macOS)

cargo install codewhale-cli codewhale-tui --locked

Both crates are required — codewhale-cli produces the codewhale dispatcher and codewhale-tui produces the interactive runtime that the dispatcher delegates to. Installing only one binary will fail at runtime with a MISSING_COMPANION_BINARY error.

Manual download — platform archives (recommended)

Each archive below contains both the codewhale dispatcher and codewhale-tui runtime, plus an install script:

Unix (Linux / macOS):

tar xzf codewhale-<platform>.tar.gz
cd codewhale-<platform>
./install.sh

Windows:

• For the installer path, run CodeWhaleSetup.exe; it installs both binaries under %LOCALAPPDATA%\Programs\CodeWhale\bin and adds that directory to the current-user PATH.
• Extract codewhale-windows-x64.zip
• Run install.bat (copies to %USERPROFILE%\bin)
• Add %USERPROFILE%\bin to your PATH

The portable Windows archive skips the install script — extract and run from any directory. The NSIS installer is currently unsigned and may trigger Windows SmartScreen until a signing certificate is wired into the release pipeline.

Individual binaries are also attached below for scripting and the npm wrapper. The legacy npm package deepseek-tui is deprecated and is not republished. For migration from v0.8.x legacy binary names, see docs/REBRAND.md.

Verify (recommended)

Download the checksum manifests from this Release and verify:

# Linux — archive bundles
sha256sum -c codewhale-bundles-sha256.txt

# Linux — individual binaries
sha256sum -c codewhale-artifacts-sha256.txt

# macOS
shasum -a 256 -c codewhale-bundles-sha256.txt
shasum -a 256 -c codewhale-artifacts-sha256.txt

What's in v0.8.59

Added

• Moonshot Kimi K2.7 Code model. The Moonshot/Kimi provider now defaults to
  kimi-k2.7-code, recognizes kimi/kimi-k2 aliases for that model, keeps
  explicit kimi-k2.6 selectable, and adds the OpenRouter
  moonshotai/kimi-k2.7-code registry row.
• Concise verbosity mode (#3052). CLI noninteractive launches now default
  to concise prompt/output discipline unless overridden by config, env, or
  --verbosity, while interactive TUI launches remain normal by default.
  Thanks @cyq1017 for the PR.
• Ephemeral generated project context (#3058). Opening CodeWhale in a
  directory with no instruction files now keeps the bounded generated project
  overview in memory instead of creating .codewhale/instructions.md.
• ACP registry auth metadata (#1447). The ACP stdio adapter now advertises
  terminal authentication setup in initialize.authMethods, matching the
  registry's validation requirement.
• Sidebar context menus (#3065). Right-clicking the sidebar no longer shows
  Paste; clickable sidebar rows now offer their row command as the first
  context action.
• Sidebar hover popovers (#3088). Streaming turns now keep sidebar hover
  popovers responsive while continuing to throttle transcript/body mouse
  motion.
• Dark-theme selection contrast (#3074, thanks @drpars). Session, config,
  help, context-menu, and approval selections now use the muted selection
  background instead of the bright accent color.
• Cursor-style activity metadata rows (#3146). Dense successful tool-run
  summaries now render as a single muted Explored ... / Updated metadata
  row, include short command-family labels for successful generic verifier
  groups, and keep keyboard/mouse expansion and detail inspection intact.
• Provider-wait observability (#3095). Footer stall reasons now name the
  active provider/model route, idle seconds vs stream budget, and whether a
  fanout plan is still at 0 running or dispatch is pending. Structured
  provider-wait incidents log once per turn from the main tick loop (not on
  every footer redraw).
• Interactive fanout launch gate (#3095). Direct sub-agent children queue
  behind a configurable semaphore ([subagents] interactive_max_launch,
  default 4) with a visible queued: waiting for an interactive fanout slot
  reason before their first model step.
• Goal lifecycle controls. /goal is now the primary command surface for
  session goals, with pause, resume, complete, blocked, and clear
  controls while /hunt remains a compatibility alias.
• Persistent thread-goal API. App-server clients can now set, get, and clear
  durable thread goals through thread/goal/set, thread/goal/get, and
  thread/goal/clear, backed by the state store with Codex-style status and
  token/time accounting fields.
• Command-boundary ownership layers (#2888/#3055). Built-in slash command
  metadata now lives in commands/registry.rs, slash parsing in
  commands/parse.rs, and handlers under group-owned command areas, preserving
  the existing dispatch surface while reducing future commands/mod.rs churn.
• Approval-rule source metadata (#1186/#2971). Runtime API
  approval.required events now include optional matched_rule metadata when
  an execution-policy rule caused the prompt. Thanks @greyfreedom for the PR
  and @Ram9199 for the audit-semantics discussion.
• Localized tool-family labels (#2901). Tool activity labels for read,
  patch, run, find, delegate, fanout, RLM, verify, think, and generic tool
  work now route through the shipped locale tables. Thanks @gordonlu for the
  PR.
• Localized config section labels (#2918). The interactive config view now
  localizes section and session/saved scope labels while preserving English
  search terms. Thanks @gordonlu for the PR.
• Localized config editor labels (#2919). The config editor modal now
  localizes edit labels, default/unavailable placeholders, and effective
  currency hints. Thanks @gordonlu for the PR.
• Hotbar number-key dispatch (#3056). Bare 1-8 now trigger bound
  hotbar slots only when the composer is empty, while Alt+1-Alt+8 trigger
  slots regardless of composer text and overlays keep key ownership. Thanks
  @reidliu41 for the PR.
• Voice dictation commands (#3051). /voice, /voice-send, and
  /voice-control now record through sox/rec/arecord, transcribe via the
  active provider's chat-completions API, and insert transcripts at the
  composer cursor. The voice.toggle hotbar action dispatches the real voice
  command, with help and status text localized across all seven shipped
  locales. Thanks @HUQIANTAO for the PR.
• Thread rewind and snapshot restore API (#2808). GUI clients can now call
  POST /v1/threads/{id}/undo, /patch-undo, and /retry to fork, roll back,
  or rerun recent thread turns, plus POST /v1/snapshots/{id}/restore to
  restore a workspace snapshot by id. Thanks @bengao168 for the PR.
• Active provider fallback chain (#2773). Configured fallback_providers
  now build an ordered primary-plus-fallback route that the TUI can report,
  advance through, and reset with /provider fallback reset, including footer
  visibility for fallback state. Thanks @idling11 for the PR.
• Provider metadata registry (#3005). Built-in provider ids, display names,
  defaults, env vars, config keys, aliases, and wire formats now live in a
  shared metadata registry, with the provider drift check covering the registry
  contract. Thanks @sximelon for the PR.
• Hugging Face provider route (#2879). Hugging Face Inference Providers now
  have first-class config, env, docs, and registry coverage for the
  OpenAI-compatible router, including huggingface/hugging-face/
  hugging_face/hf aliases and HUGGINGFACE_*/HF_* env fallbacks. Thanks
  @mvanhorn for the PR.

Fixed

• SSE data lines without spaces (#3152). Chat Completions, Responses, and
  Anthropic stream readers now accept both data: {...} and data:{...} SSE
  frames, matching the spec and preventing providers that omit the optional
  space from streaming empty output. Thanks @wgeeker for the PR.
• Runtime thread detail N+1 reads (#3141). get_thread_detail now scans
  persisted turn items once and groups them by turn instead of reading the
  items directory once per turn, preserving item order while keeping large
  thread detail loads responsive.
• Project-local hook trust boundary (#3140). .codewhale/hooks.toml is now
  loaded only after the workspace is trusted in user-owned config, matching the
  project-local MCP trust model while preserving the documented shell-command
  hook contract.
• Skill registry sync latency (#3139). /skills sync now syncs registry
  entries with bounded ordered concurrency, so network latency no longer stacks
  one skill at a time while output order stays deterministic.
• SiliconFlow China provider config (#2893/#2895). siliconflow-CN
  now reads its own [providers.siliconflow_cn] / [providers.siliconflow-CN]
  table and falls back to [providers.siliconflow] only for unset
  api_key/base_url/model fields. Thanks @Artenx for the report and
  @idling11 for the PR.
• Self-update download timeout (#3006). codewhale update now applies a
  five-minute HTTP client timeout so blocked or very slow GitHub release
  downloads fail instead of hanging indefinitely. Thanks @New2Niu for the PR.
• Legacy deepseek update migration (#2960/#3013/#3053). Running
  deepseek update or deepseek-tui update from a pre-rebrand install now
  returns copy-pasteable npm, Cargo, Homebrew, and manual-binary migration
  steps instead of trying to spawn a missing codewhale binary. README and
  rebrand docs now cover the same upgrade path. Thanks @jazzi and
  @tiangangQiu for the reports, @cyq1017 for the update-path PR, and
  @angus-guo for the README PR.
• Short codew shim delegation. The codew convenience binary now
  prefers the sibling codewhale dispatcher installed next to it before
  falling back to PATH, preventing fresh local builds or installs from
  accidentally invoking an older global dispatcher.
• Constitution trust wording (#2950/#3008). The base prompt now explains
  that "begins with an A" means a baseline of trust, not a literal output
  formatting rule. Thanks @cyq1017 for the PR.
• TUI provider-source recovery (#3007/#3011). Unsupported interactive
  providers now report whether the value came from --provider, environment,
  or config. Config-sourced unsupported providers fall back to DeepSeek without
  forwarding stale keyring secrets. Thanks @cyq1017 for the PR.
• Exec auto-model handoff (#3148). codewhale exec --model auto now
  survives the CLI/TUI boundary by honoring the CodeWhale model env alias and
  legacy DeepSeek model handoff before falling back to provider defaults.
  Thanks @hongchen1993 for the PR.
• macOS shortcut modifiers (#2938/#2943). Ctrl-like shortcuts that are
  reported as SUPER by macOS terminals now work for backgrounding tasks and
  sidebar-focus chords without rewriting clipboard shortcuts. Thanks @idling11
  for the PR.
• TUI mouse-report leak (#3063/#3067). Strip raw SGR mouse coordinate
  tails from the composer even when use_mouse_capture is false, covering
  orphaned terminal reporting state after crashes or focus races.
• Interrupted sub-agent lifecycle (#3080). API-timeout interruptions now
  emit MailboxMessage::Interrupted, render terminal interrupted cards, and
  reconcile stale running fanout counts from manager snapshots.
• OpenAI Codex stream diagnostics and active tool collapse (#3146). The
  Responses bridge now reports nested response.failed /
  response.incomplete errors instead of unknown, and dense successful
  in-flight tool bursts collapse into the same calm activity metadata row as
  committed history.
• OpenAI Codex reasoning tiers. Switching from DeepSeek to openai-codex
  now normalizes stale reasoning state into Responses-compatible
  low/medium/high/xhigh tiers. Startup, /config, and the model
  picker now display Codex labels instead of leaking DeepSeek
  off/max names, while Codex still reports as a Responses payload
  provider. The Responses request builder also clamps legacy minimal input
  to low and has regression coverage that Codex requests use
  reasoning.effort, not DeepSeek thinking fields.
• OpenAI Codex context metadata (#3070). The gpt-5.5 default and
  CodeWhale aliases now use OpenAI's documented 1,050,000-token context window
  and 128,000 max-output metadata for context pressure, prompts, and doctor
  capability output.
• OpenRouter Nemotron 3 Ultra preset. The OpenRouter preset and model
  registry now emit nvidia/nemotron-3-ultra-550b-a55b while keeping the old
  Ultra aliases compatible.
• OpenRouter auth after MiMo switches (#3064). Switching from Xiaomi MiMo
  to OpenRouter now has regression coverage for preflight key failures and
  Bearer auth header isolation before any request can be dispatched.
• Responses strict-tool schema compatibility (#3062/#3017/#1883). Responses
  function tools now preserve per-tool strict-mode compatibility, keep optional
  strict-schema fields nullable, and append deterministic constraint notes when
  root composition groups must be flattened for Responses.
• Runtime prompt autonomous loop guard (#3061). Runtime policy reference
  now explicitly forbids initiating new work when <runtime_prompt> is the
  only new turn content and no tool/sub-agent handoff is pending.
• Goal runtime status sync. Goal token budgets and active/paused/complete
  status now sync into the engine alongside the objective, and model-visible
  update_goal can only mark goals complete or blocked.

Contributors

• Devin session work on #3080/#3095 (PRs #3103, #3104, #3106) — Hunter Bown
  (maintainer integration/cherry-pick on codex/v0.8.59-release-ready).
• Nightt (@nightt5879) for the Responses strict-tool schema hardening in PR
  #3062.
• yekern (@yekern) for the #3061 runtime-prompt loop safety report and repro
  that shaped the dispatch guard.
• Paulo Aboim Pinto (@aboimpinto) for the staged command-boundary design and
  Layer 3 registry/parser extraction in PR #2888, plus the #2851/#2791/#2870
  architecture stream that guided the grouped command areas in #3055.

Contributor credits for this release live in the changelog entry above —
thank you to everyone whose reports, PRs, reviews, and reproductions shaped it.

See CHANGELOG.md for full notes and docs/CHANGELOG_ARCHIVE.md for older releases.