v0.8.67

> **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)

```bash
npm install -g codewhale
```

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

### Docker / GHCR

```bash
docker run --rm -it \
  -e DEEPSEEK_API_KEY="$DEEPSEEK_API_KEY" \
  -v ~/.deepseek:/home/codewhale/.deepseek \
  ghcr.io/hmbown/codewhale:v0.8.67
```

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

### Cargo (Linux / macOS)

```bash
cargo install codewhale-cli codewhale-tui --locked
```

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

### Manual download — platform archives (recommended)

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

| Platform | Archive | Install script |
|---|---|---|
| Linux x64 | `codewhale-linux-x64.tar.gz` | `install.sh` |
| Linux ARM64 | `codewhale-linux-arm64.tar.gz` | `install.sh` |
| macOS x64 | `codewhale-macos-x64.tar.gz` | `install.sh` |
| macOS ARM | `codewhale-macos-arm64.tar.gz` | `install.sh` |
| Windows x64 (installer) | `CodeWhaleSetup.exe` | NSIS setup |
| Windows x64 | `codewhale-windows-x64.zip` | `install.bat` |
| Windows x64 (portable) | `codewhale-windows-x64-portable.zip` | — |

**Unix (Linux / macOS):**
```bash
tar xzf codewhale-<platform>.tar.gz
cd codewhale-<platform>
./install.sh
```

**Windows:**
- For the installer path, run `CodeWhaleSetup.exe`; it installs `codewhale.exe`, `codew.exe`, and `codewhale-tui.exe` 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.

Each platform also has **bare, unarchived** binaries attached below (`codewhale-<platform>`, `codew-<platform>`, and `codewhale-tui-<platform>`) — the npm wrapper and the in-app `codewhale update` download the matched runtime binaries, whereas the `.tar.gz` / `.zip` archives above are the recommended manual download and additionally bundle an install script. 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:

```bash
# 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 v0.8.67

### Added

- The model you select in `/model` is now the operator: fleet workers whose
  task spec and roster profile pin no model inherit the active session route
  instead of a hardcoded `auto` sentinel, matching the pinned operator row in
  `/fleet roster`. Task-level and profile model overrides still win, and
  route receipts record which source applied (`task.model`,
  `agent_profile.model`, or `run.model`).
- Added the `/workflow` command (aliases `/workflows`, `/wf`) as the user
  opt-in to workflow orchestration. Bare `/workflow` orchestrates the current
  work — the model synthesizes the objective from the conversation context;
  `/workflow <objective>` narrows the run; `/workflow status [run_id]` and
  `/workflow cancel <run_id>` relay typed run receipts without starting new
  runs.
- Bare `/goal` with no active goal now declares a goal from the conversation
  context via `create_goal` instead of printing usage; with an active goal it
  remains the status readout, and explicit `/goal <objective>` is unchanged.
- Added the constitution-first setup wizard: a unified `/setup` shell with
  resume, back navigation, and skip-retry state; provider/model readiness
  cards with a custom-provider form and provider-picker detail layout; a
  runtime posture card with preset application and project-override warnings;
  a setup verification report; and transactional setup persistence with
  secret redaction and rollback (#3402, #3403, #3404, #3405, #3406, #3410,
  #3411).
- Added a structured user-global constitution with a deterministic renderer,
  prompt-block injection, guided principle authoring with preview and preset
  save, and a `/constitution` manager command as the primary constitution
  management surface, with file state shown in setup and actions surfaced in
  diagnostics (#3793, #3806, #3811).
- Added model-assisted constitution and fleet-profile drafting behind an
  explicit ratify gate, with untrusted-draft provenance recorded so
  model-authored text is never applied silently. Updating users keep their
  existing constitution unchanged, and a localized constitution checkpoint is
  required after update (#3794).
- Added the Hotbar route editor v1 with route-switch slot actions and support
  for custom model routes, plus a configured-provider route manager for
  `/provider` and `/model` with a missing-auth handoff into provider key
  entry (#2066, #3830, #3831).
- Added auto-discovery of `.codewhale/rules/` and `.claude/rules/`
  directories as project context, with a total byte-budget cap on the
  assembled rules block. Contributed by maple (@yekern).
- Exposed `context_input_budget_for_route` from the engine so external
  integrations can reuse route budget math. Contributed by hexin
  (@h3c-hexin).
- Added GUI config persistence to the runtime API. Contributed by @gaord.
- Added a website localization matrix with a locale registry and drift
  checks. Harvested from #3763 by @idling11 (#3090).
- Added `doctor` detection of half-applied setup state, and startup milestone
  tracing for boot-performance diagnosis.
- Added a v0.8.67 computer-use dogfood prompt that covers the Cursor-terminal
  QA flow, headless gates, setup, sub-agent completion, Fleet, Workflow, model
  pricing, and release evidence collection.
- Fleet: local worker memory usage is now reported, including retained memory
  while a task is in Running status. Contributed by @cyq1017 (#3901).
- Website: community hub, constitution thesis page and constitution-centered
  homepage, models page generated from the provider registry, docs dark mode
  and full SEO metadata/sitemap coverage, terminal player for real
  constitution traces, and a live star badge and version.
- Added Meituan LongCat as a first-class OpenAI-compatible provider
  (`longcat`, with `long-cat`, `meituan-longcat`, and `meituan` aliases),
  `LONGCAT_API_KEY` discovery, the `LongCat-2.0` default model, provider
  picker wiring, model completions, provider docs, and web provider facts.
- Fleet: added per-provider setup cards (Persistence, Constitution, Hotbar,
  Tools/MCP, Remote Runtime) with a unified setup catalog and provider-specific
  credential links. Provider setup progress is persisted transactionally with
  rollback guards, Codex OAuth is kept out of provider key storage, and a
  headless QA contract verifies setup readiness across providers.
- Fleet: added Fleet starter profiles with role-aware loadouts (scout→Fast,
  manager→Inherit, etc.), `/fleet setup` profile-authoring wizard, Fleet
  effective-permission recording, and route intent-source tracking.
- Fleet: added 'operator' as a built-in Fleet roster member — the preferred
  helm Fleet slot for workflow coordination. Operator plans, routes, reviews
  outputs, and calls other Fleet slots as needed. This is a roster role, not a
  separate app mode. The full Operation/Operate-mode architecture is deferred
  to 0.9.0.
- Workflow: declarative workflows now run through the production driver, the
  workflow tool is wired to sub-agent dispatch, public Workflow surfaces are
  renamed, and typed workflow-run and status receipts are emitted for
  debugging and verification.
- Added provider-agnostic Fleet rosters and loadouts: provider-specific
  subagent limits, launch concurrency, and admission caps are derived from
  config without hardcoding any single provider.
- Added Workflow runtime foundations: the internal JS authoring/runtime crates
  compile and replay example workflows. 0.8.67 ships the `/workflow` opt-in,
  production-driver dispatch path, sub-agent task handoff, and typed run/status
  receipts; richer authoring UX and the full TUI run view remain tracked for
  v0.8.68 (#2974, #4038).

### Changed

- Clarified the Fleet coordination hierarchy and made roles carry real
  doctrine: the **operator** (the session's `/model` selection) runs the
  operation and assigns managers to workflows; a **manager** is the middle
  manager of exactly one workflow. The built-in **reviewer** is now explicitly
  adversarial (assume the change is broken, try to refute it), and the review
  sub-agent intro adopts the same framing. Built-in `manager`/`operator`/
  `reviewer` roster members now ship role `instructions` that flow into worker
  prompts on both the Fleet task-spec and agent/workflow `profile:` spawn
  paths; custom profiles override them via the same `instructions` field.
- Removed the decorative Fleet vocabulary that never routed differently:
  the `tool-heavy` slot and the `strong`/`balanced`/`deep-reasoning`/`code`/
  `review`/`tool-heavy` loadout tiers. `inherit` (the operator's route) and
  `fast` (the provider's faster class) remain; retired names in existing
  configs keep parsing (as custom labels) with identical auto routing, and
  the `/fleet setup` model-class step now offers only the real choices.
- Raised the default subagent concurrency for high-throughput fanout:
  `max_subagents` default 20 → 64 (config ceiling 128) and the queued+running
  admission cap 200 → 1024. Users on metered plans who want the old behavior
  can set `max_subagents = 20` in config.toml.
- Renamed the internal `whaleflow` subsystem to `workflow` across the
  workspace: the `codewhale-whaleflow`/`codewhale-whaleflow-js` crates become
  `codewhale-workflow`/`codewhale-workflow-js`, Rust identifiers and JS bridge
  symbols are renamed, the `CODEWHALE_WHALEFLOW_JS_*` environment variables
  become `CODEWHALE_WORKFLOW_JS_*`, and the authoring/RFC docs move to
  `WORKFLOW_AUTHORING.md` and `WORKFLOW_EXTERNAL_MEMORY.md`. Historical
  changelog and retro-ledger entries keep the old name as a record.
- Documented the Homebrew rollout strategy and added a distribution-channel
  check to the release checklist. Harvested from #3760 by @idling11 (#3489).
- Paused Linux RISC-V prebuilt release and nightly artifacts because
  `rquickjs-sys` 0.12.0 does not ship `riscv64gc-unknown-linux-gnu` bindings;
  installers, docs, and update paths now treat RISC-V as unsupported until
  upstream bindings or a bindgen-enabled build lands.
- Made the approval prompt calm, compact, and honest, and centered the
  first-run follow-up on the constitution; first-run onboarding now hands off
  into the setup wizard, and the language picker offers every shipped locale
  (#3929).
- Startup performance: boot janitors and store scans no longer block the
  first frame, `@mention` completion no longer re-walks the workspace per
  keystroke, and idle offline-queue clones and duplicate tool-output hashing
  were eliminated.
- Clarified the misleading "Ctrl+B backgrounds this command" shell wording
  (#3859) and the hotbar help shortcuts. Docs contribution by Chanhyo Jung
  (@roian6).
- Documented the enforced repo-law invariants, the constitution flow, and the
  `/fleet setup` profile-authoring wizard; aligned `permissions.toml` action
  docs. Docs contribution by @greyfreedom.
- Bumped web dependencies: wrangler 4.103.0 → 4.107.0, mermaid 11.15.0 →
  11.16.0, vitest 4.1.8 → 4.1.9 (@dependabot).
- Backfilled v0.8.67 regression coverage across sub-agent completion, budget
  exhaustion, delegate ordering, provider onboarding, setup scroll, model
  catalog pricing, Fleet routing, and Workflow gates (#4076).
- Split the large TUI debug command group and palette/theme internals into
  smaller modules without changing user-visible behavior (#4078, #4081).

### Fixed

- Fixed the goal sidebar elapsed timer so completed and blocked goals freeze
  their "completed in {elapsed}" readout instead of ticking forever. Goal state
  now records a `finished_at` instant that both sidebar render paths and the
  engine snapshot clamp elapsed against; `/goal resume` clears the freeze and
  the timer ticks again.
- Fixed paused goals silently un-freezing their sidebar timer: usage keeps
  accruing while paused, and the next goal snapshot used to clear the frozen
  instant. Paused goals now stay frozen until an explicit resume.
- Fixed durable `/goal` progress accounting so usage and continuation updates
  release the shared SQLite connection before re-reading the updated goal,
  unblocking resumed goal loops and full workspace release tests.
- Fixed a scheduled-automation race where deleting an automation while its
  run was being enqueued left the already-created task running untracked;
  the run record is now persisted unconditionally.
- Removed `panic = "abort"` from the release profile: it disabled unwinding
  and broke the panic supervision that keeps one failing tool call from
  taking down the whole session. The `lto`/`strip`/`codegen-units` size and
  speed tuning is unchanged.
- Fixed session save/load to persist and restore the active model provider
  across restarts. Previously sessions created under one provider (e.g.
  DeepSeek) would silently load under a different active provider. Provider,
  subagent limits, fallback chain, context window, and reasoning effort are now
  restored from saved session metadata, with `"deepseek"` as the default for
  legacy sessions.
- Raised the streamed model-response idle timeout and matched the TUI stall
  watchdog to the configured stream budget so long reasoning pauses are not
  recovered as stalled turns (#2487, #3998).
- Fixed Codex OAuth/sub-agent release diagnostics so `auth list` reports an
  active Codex OAuth file, Responses API child requests encode inherited tool
  names safely, rate-limited child requests checkpoint as resumable provider
  interruptions, and failure records surface the real Responses API error
  (#3884).
- Fixed fresh launch/setup testing with an explicit `CODEWHALE_HOME` so
  config, settings, theme prefs, and doctor legacy-state diagnostics do not
  inherit unrelated ambient `~/.deepseek` files (#4001, #4002).
- Sub-agent state now persists to `.codewhale/` instead of the lingering
  pre-rebrand `.deepseek/` path (#3864). Contributed by Stime (@yekern).
- `/plugin enable|disable` now persists across restarts (#3918), and the
  plugin command is hidden from the root slash menu and kept canonical after
  the scanner merge. Contributed by Nightt (@nightt5879).
- `/config ask-rules` now shows ask rule actions with improved diagnostics,
  with file-rule action precedence under test. Contributed by @greyfreedom.
- Fleet/sub-agents: enforced an absolute recursion-depth ceiling and widened
  task-id entropy, gave each atomic state write a unique temp path, kept
  sub-agent tool catalogs in parent parity (#3836), and made the Agents
  sidebar reconcile sub-agent completion and cancellation live (#3837).
- Fixed apply_patch mangling newlines, defaulted fuzz to 3, and made writes
  atomic; fixed compaction to preserve pins on emergency compaction, harden
  the summary fallback, and count image tokens; corrected backtrack boundary,
  checkpoint clear ordering, prune guard, and durable rename.
- Fixed the SSE client to flush the final frame, join multi-line data fields,
  and stop corrupting multibyte UTF-8 split across network reads.
- Kept review-only turns read-only, aliased `auto` mode to the agent policy,
  showed the mode-derived safety policy in status (contributed by @cyq1017),
  and stopped the durable-review floor from holding routine YOLO work
  (#3883).
- Fixed self-update to prefer exact binary release assets. Contributed by
  @LI-Jialu.
- UI polish: stopped constitution and fleet-profile model drafts from
  freezing the event loop, scoped the context-menu backdrop to the popup
  rect, stacked model-picker panes on narrow modals, unified display-width
  helpers on one contract (#3924), removed misleading success toasts,
  issue-number leaks, and dead-end empty states, and repaired the onboarding
  trust and api-key keys.
- Fixed the onboarding Trust step so plain Enter no longer silently grants
  workspace trust; users must choose the explicit trust or exit keys.
- Fixed same-root skill-name collisions being silently shadowed; duplicate
  normalized skill names now warn while keeping discovery deterministic
  (#3919).
- Normalized discovered skill names, removed unenforced trust copy, and
  surfaced the gated constitution override in prompts.
- Fixed a parallel `subagent::` suite flake where one test's process-wide
  `Retry-After` pause could strand unrelated budget-capped workers for the
  full stale window; requests now re-poll the global pause in bounded slices
  and the rate-limit test clears the window on drop.
- Sub-agent and Fleet reliability now fail empty, step-limited, and
  budget-exhausted children with explicit diagnostics instead of silent
  `Completed (no output)` success; budget exhaustion preserves partial output,
  `worktree: true` discovers one-level nested repos from harness directories,
  and completion-before-start delegate events recover into named rows instead
  of ellipsis-only identities (#4050, #4051, #4052, #4053).
- Goal-mode writing and research tasks can complete with
  `verification.status = "not_applicable"` without triggering continuation
  loops (#4054).
- First-run onboarding routes API keys through the selected provider, setup
  wizard bodies scroll with PageUp/PageDown, shipped locale packs are back to
  `en.json` parity with zh-Hant explicitly partial, stable feature flags stay
  out of Experimental, and model/provider rows include current LongCat and
  sourced-pricing hints (#4056, #4057, #4058, #4062, #4063).
- Running tool rows animate while a lone foreground tool is active, and
  workflow receipts render run/status/failure cards instead of one-line or
  null-success output (#4059).
- Model-facing turn metadata now includes a compact git workspace snapshot and
  escalates context pressure at the same thresholds as the TUI, helping agents
  narrow scope or compact before truncation (#4071, #4073).
- Successful child sub-agent completions inline the child's `EVIDENCE` block
  before the completion sentinel, so parents can cite child findings without
  re-running tools (#4072).
- Deferred tools hydrate and execute in the same batch when the original
  arguments are valid, and `[tools].always_load` now keeps configured MCP tools
  active instead of forcing the first-call retry. Thanks @SparkofSpike for the
  hot-path MCP report (#4074, #4027).
- New commit-range co-author checks reject bot/tool trailers on newly pushed
  commits; historical release-range cleanup remains a separate maintenance
  concern (#4075).
- Fixed fuzzy `edit_file` matching so matches that begin with multibyte UTF-8
  characters, including CJK text, advance on character boundaries instead of
  panicking. Contributed by Nightt (@nightt5879), reported by Taixin Guo
  (@taixinguo) (#3971, #4045).
- Fixed Unix dispatcher/TUI output under early-closing pipes such as
  `codewhale doctor | head` by restoring the default `SIGPIPE` handler before
  printing and propagating signal exits quietly. Contributed by @aznikline,
  reported by @BrathonBai (#4030, #4043).
- Suppressed dead_code warnings in the unused plugin registry module and
  fixed formatting across the command-group files. Contributed by Paulo Aboim
  Pinto (@aboimpinto).
- Pointed the website Community nav link at the community hub.

### Security

- MCP client hardening: closed an SSE-endpoint SSRF, bounded the HTTP
  response body via Content-Length instead of a streaming read, bounded stdio
  line reads to prevent OOM denial of service, fixed a dead timeout, and
  removed an unbounded buffer.
- Made execpolicy deny/trust rules segment-aware, closing a command-chaining
  bypass.
- Closed repo-law and safety-floor bypasses found by adversarial review:
  protected invariants are now enforced as mechanism, the destroyer gap in
  the safety floor is closed, a catalog-present tool with no execution path
  now fails closed, `web_run` open/click is classified as destructive, and
  the allow-list gained wildcard and case handling.
- Refused symlinked rules directories to prevent workspace escape via
  discovered rules. Contributed by maple (@yekern).
- Bounded Fleet sub-agent worker output so fanout cannot exhaust TUI memory
  (#3882), and preserved event headroom for progress. Contributed in part by
  @cyq1017.
- Added an untrusted constitution-draft gate with authoring provenance so
  model-drafted constitutions require explicit human ratification.

### Removed

- Removed unused model-registry helpers. Harvested from #3872 by @cyq1017.
- Removed unused request-tuning metadata. Harvested from #3871 by @cyq1017.
- Removed dead fleet task helpers (#3894 by @cyq1017), the unused
  approval-cache container (#3845) and localization QA metadata (both by
  @nightt5879), the dormant tab collaboration subsystem (#3838), the legacy
  flash auto-router (#3839), the stale project_doc loader (#3840), ignored
  mock LLM placeholders (#3841), dead model-catalog helpers (#3842), the
  unused execpolicy amend module, and dead MCP/client retry helpers.
- Retired the deprecated `WHALE.md` context fallback (#3798).

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](https://github.com/Hmbown/CodeWhale/blob/main/CHANGELOG.md) for full notes and [docs/CHANGELOG_ARCHIVE.md](https://github.com/Hmbown/CodeWhale/blob/main/docs/CHANGELOG_ARCHIVE.md) for older releases.