Qwen Code v0.21.1

<!-- qwen-release-notes:v1 --> ## Highlights _See the complete change list below._ ## Breaking Changes No known breaking changes. ## Complete Change List ### Features - feat(core): Align GenAI content telemetry fields ([#7667](https://github.com/QwenLM/qwen-code/pull/7667)) by @doudouOUC - feat(core): add Goal v3 runtime orchestration ([#7664](https://github.com/QwenLM/qwen-code/pull/7664)) by @qqqys - feat(triage): stop in-agent CI polling, finalize evidence and approval after CI completes ([#7693](https://github.com/QwenLM/qwen-code/pull/7693)) by @wenshao - feat(serve): expose workspace Channel management API ([#7637](https://github.com/QwenLM/qwen-code/pull/7637)) by @qqqys - feat(web-shell): add read-only GitHub pull requests panel ([#7683](https://github.com/QwenLM/qwen-code/pull/7683)) by @wenshao - Adds a retrieval-only context_search tool for external corpora, configured by administrators without automatic recall or write capabilities. ([#7586](https://github.com/QwenLM/qwen-code/pull/7586)) by @doudouOUC - Introduces the qwen review comment-status subcommand to quickly triage existing inline comments and reduce API calls during review runs. ([#7690](https://github.com/QwenLM/qwen-code/pull/7690)) by @wenshao - Enforces a strict write contract for review submissions and adds a tripwire to detect and flag any bypass attempts via terminal commands. ([#7691](https://github.com/QwenLM/qwen-code/pull/7691)) by @wenshao - Enables hot-reloading of workspace trust changes in the running daemon, applying new policies immediately without requiring a process restart. ([#7268](https://github.com/QwenLM/qwen-code/pull/7268)) by @doudouOUC - Adds retryInitialDelayMs and retryMaxDelayMs settings to configure stream-side rate-limit retry delays for better provider compatibility. ([#7674](https://github.com/QwenLM/qwen-code/pull/7674)) by @jay666mnj - Updates the /stats command and Session tab to display generation timing metrics including TTFT, duration, output tokens, and TPS. ([#7677](https://github.com/QwenLM/qwen-code/pull/7677)) by @destire-mio - Adds a GitHub channel adapter that polls notifications and responds to mentions by posting comments using a signal-based wakeup architecture. ([#7632](https://github.com/QwenLM/qwen-code/pull/7632)) by @OrbitZore - Adds optional retryInitialDelayMs and retryMaxDelayMs settings to configure SSE stream rate-limit retry delays for specific provider quota windows. ([#7666](https://github.com/QwenLM/qwen-code/pull/7666)) by @hogeheer499-commits - Introduces an overridable default-disabled state for skills, allowing soft defaults that yield to explicit enablement while hard disables remain absolute. ([#7357](https://github.com/QwenLM/qwen-code/pull/7357)) by @samuelhsin - feat(webui): add workspace Channel management hook ([#7728](https://github.com/QwenLM/qwen-code/pull/7728)) by @qqqys - feat(review): redefine medium effort as a balanced verified pass ([#7733](https://github.com/QwenLM/qwen-code/pull/7733)) by @wenshao - feat(review): mutation-test the tests in the test-coverage pass (Agent 5) ([#7735](https://github.com/QwenLM/qwen-code/pull/7735)) by @wenshao - feat(review): borrow maintainer review lenses into the agent briefs ([#7736](https://github.com/QwenLM/qwen-code/pull/7736)) by @wenshao - feat(web-shell): persist terminal history pagination errors ([#7709](https://github.com/QwenLM/qwen-code/pull/7709)) by @PratikWayase - feat(triage): add sandboxed /verify deep-verification lane ([#7710](https://github.com/QwenLM/qwen-code/pull/7710)) by @wenshao - feat(review): give the verifier a probe capability — run a runnable claim, don't just read it ([#7756](https://github.com/QwenLM/qwen-code/pull/7756)) by @wenshao - feat(core): add model grade selection for subagent spawn (#7685) ([#7702](https://github.com/QwenLM/qwen-code/pull/7702)) by @yiliang114 - feat(dingtalk): support outbound image delivery ([#7698](https://github.com/QwenLM/qwen-code/pull/7698)) by @qqqys - feat(web-shell): allow widening sidebar up to half the window width ([#7778](https://github.com/QwenLM/qwen-code/pull/7778)) by @wenshao - feat(core): add Goal v3 worker tools ([#7729](https://github.com/QwenLM/qwen-code/pull/7729)) by @qqqys - feat(review): script-lint as a deterministic gate — compose-review reads the report, no agent ([#7751](https://github.com/QwenLM/qwen-code/pull/7751)) by @wenshao - feat(web-shell): add monitor task details ([#7817](https://github.com/QwenLM/qwen-code/pull/7817)) by @ytahdn - feat(web-shell): Scope voice to composer workspace ([#7754](https://github.com/QwenLM/qwen-code/pull/7754)) by @doudouOUC - feat(hooks): Add submitted prompt provenance ([#7762](https://github.com/QwenLM/qwen-code/pull/7762)) by @doudouOUC - feat(autofix): retry deterministic rejection once ([#7796](https://github.com/QwenLM/qwen-code/pull/7796)) by @qqqys - feat(web-shell): add Channel management page ([#7793](https://github.com/QwenLM/qwen-code/pull/7793)) by @qqqys - feat(acp): add session-scoped runtime MCP ([#7847](https://github.com/QwenLM/qwen-code/pull/7847)) by @qqqys - feat(core): persist and replay Goal v3 state ([#7815](https://github.com/QwenLM/qwen-code/pull/7815)) by @qqqys - feat(core): add full-resolution image zoom tool ([#7809](https://github.com/QwenLM/qwen-code/pull/7809)) by @qqqys - feat(web-shell): add composer footer renderer ([#7856](https://github.com/QwenLM/qwen-code/pull/7856)) by @dreamWB - feat: Gate session writer lease behind opt-in ([#7894](https://github.com/QwenLM/qwen-code/pull/7894)) by @doudouOUC - feat(ci): Deduplicate E2E failure issues by commenting on existing issue ([#7792](https://github.com/QwenLM/qwen-code/pull/7792)) by @yiliang114 - feat(triage): add revert-pattern high-risk path detection ([#7414](https://github.com/QwenLM/qwen-code/pull/7414)) by @yiliang114 - feat(web-shell): add native workspace folder picker ([#7849](https://github.com/QwenLM/qwen-code/pull/7849)) by @qqqys - feat(web-shell): add Channel configuration flows ([#7893](https://github.com/QwenLM/qwen-code/pull/7893)) by @qqqys - feat(core): Add ARMS session user ID ([#7921](https://github.com/QwenLM/qwen-code/pull/7921)) by @doudouOUC - feat(channels): expose loop tools in daemon sessions ([#7891](https://github.com/QwenLM/qwen-code/pull/7891)) by @qqqys - feat(web-shell): honor voice hold mode ([#7839](https://github.com/QwenLM/qwen-code/pull/7839)) by @callmeYe - feat(web-shell): manage Channel pairing requests ([#7909](https://github.com/QwenLM/qwen-code/pull/7909)) by @qqqys - feat(channels): dispatch GitHub notifications by reason ([#7826](https://github.com/QwenLM/qwen-code/pull/7826)) by @yiliang114 - feat(web-shell): add git branch picker, commit dialog, and create PR flow ([#7731](https://github.com/QwenLM/qwen-code/pull/7731)) by @wenshao - feat(triage): surface the sandboxed lanes on the CI path ([#7917](https://github.com/QwenLM/qwen-code/pull/7917)) by @wenshao - feat(web-shell): suggest BTW for side questions ([#7935](https://github.com/QwenLM/qwen-code/pull/7935)) by @carffuca - feat(triage): make the verify report readable in Chinese ([#7918](https://github.com/QwenLM/qwen-code/pull/7918)) by @wenshao ### Bug Fixes - fix(cli): measure insight days and hours in local time everywhere ([#7670](https://github.com/QwenLM/qwen-code/pull/7670)) by @ComplexSimply - fix(ci): don't fail triage cleanup when there is nothing to clean ([#7688](https://github.com/QwenLM/qwen-code/pull/7688)) by @wenshao - fix(ci): update qwen in the runner's active npm prefix ([#7689](https://github.com/QwenLM/qwen-code/pull/7689)) by @yiliang114 - fix(acp): sweep review worktree leases at the end of each prompt turn ([#7694](https://github.com/QwenLM/qwen-code/pull/7694)) by @wenshao - fix(core): write a status sidecar so models stop misreading quiet background shells ([#7669](https://github.com/QwenLM/qwen-code/pull/7669)) by @ComplexSimply - fix(core): exit_plan_mode returns guidance error from execute() instead of permission deny ([#7673](https://github.com/QwenLM/qwen-code/pull/7673)) by @qwen-code-dev-bot - fix(core): tell the model when the user manually exits plan mode ([#7682](https://github.com/QwenLM/qwen-code/pull/7682)) by @zjunothing - fix(core): give plugins from the same repository distinct extension ids ([#7676](https://github.com/QwenLM/qwen-code/pull/7676)) by @zjunothing - fix(core): allow reading saved plan files without a confirmation prompt ([#7678](https://github.com/QwenLM/qwen-code/pull/7678)) by @zjunothing - fix(cli): clear stale retry error when agent auto-recovers mid-turn ([#7681](https://github.com/QwenLM/qwen-code/pull/7681)) by @qwen-code-dev-bot - fix(mcp): harden OAuth callback handling ([#7510](https://github.com/QwenLM/qwen-code/pull/7510)) by @gauravyad86 - fix(web-shell): add :focus-visible outline to GitHub PR list rows ([#7704](https://github.com/QwenLM/qwen-code/pull/7704)) by @wenshao - fix(triage): resolve stage comment ids by marker at patch time, harden model injection ([#7703](https://github.com/QwenLM/qwen-code/pull/7703)) by @wenshao - fix(triage): resolve finalize PRs from the open-PR list, not the commit association ([#7706](https://github.com/QwenLM/qwen-code/pull/7706)) by @wenshao - Fixes inline math recognition to correctly handle single-character expressions, escaped dollars, and code spans across prose and tables. ([#7701](https://github.com/QwenLM/qwen-code/pull/7701)) by @CubeLander - Fixes desktop file size display to correctly render terabyte and larger values instead of showing undefined for huge files. ([#7623](https://github.com/QwenLM/qwen-code/pull/7623)) by @chinesepowered - Ensures const-derived enums are correctly stringified during OpenAPI 3.0 conversion to meet strict schema requirements. ([#7547](https://github.com/QwenLM/qwen-code/pull/7547)) by @chinesepowered - Fixes schema conversion to properly handle nested objects even when property names match JSON Schema constraint keywords like maximum. ([#7546](https://github.com/QwenLM/qwen-code/pull/7546)) by @chinesepowered - Silences xterm.js parser diagnostic logs in headless shell terminals to prevent error messages from leaking when commands emit invalid ANSI escape sequences. ([#7663](https://github.com/QwenLM/qwen-code/pull/7663)) by @mvanhorn - Hardens the review skill by declaring the optional host argument in CommentStatusArgs and improving error handling for authentication or network failures. ([#7708](https://github.com/QwenLM/qwen-code/pull/7708)) by @wenshao - Ensures StopFailure hooks fire correctly when loop detection triggers an early return during the streaming loop instead of only on API errors. ([#7592](https://github.com/QwenLM/qwen-code/pull/7592)) by @qwen-code-dev-bot - Enables the Changes and History dialogs for worktree sessions in the Web Shell by threading the correct git working directory through the entire stack. ([#7695](https://github.com/QwenLM/qwen-code/pull/7695)) by @wenshao - Strictly parses the heatmapDays query parameter to reject malformed values like negative numbers or decimals, falling back to the default instead of clamping invalid input. ([#7218](https://github.com/QwenLM/qwen-code/pull/7218)) by @VectorPeak - Raises default live journal caps to 10,000 events and 8 MiB while exposing --max-journal-events and --max-journal-bytes flags for configuration. ([#7715](https://github.com/QwenLM/qwen-code/pull/7715)) by @wenshao - Allows pinning and grouping secondary workspace sessions in the Web Shell sidebar without requiring the workspace to be locked first. ([#7716](https://github.com/QwenLM/qwen-code/pull/7716)) by @wenshao - Corrects the review submit-gate audit to avoid falsely claiming a bypass when detecting valid same-account writes from other sources like triage bots. ([#7718](https://github.com/QwenLM/qwen-code/pull/7718)) by @wenshao - Fixes QQ Bot session restoration by returning the input session ID from AcpBridge and ensuring session patching runs even when validation fails. ([#7722](https://github.com/QwenLM/qwen-code/pull/7722)) by @Eric-GoodBoy-Tech - fix(core): prevent updates to extension-provided agents ([#7245](https://github.com/QwenLM/qwen-code/pull/7245)) by @destire-mio - fix(core): fall back to system rg when bundled ripgrep cannot run ([#7203](https://github.com/QwenLM/qwen-code/pull/7203)) by @harjothkhara - fix(core): avoid required tools in DashScope thinking ([#7661](https://github.com/QwenLM/qwen-code/pull/7661)) by @hogeheer499-commits - fix(channels): use username as senderId in GitHub adapter to fix allowlist gate ([#7727](https://github.com/QwenLM/qwen-code/pull/7727)) by @OrbitZore - fix(web-shell): parse 256-color and truecolor SGR sequences in parseAnsi ([#7620](https://github.com/QwenLM/qwen-code/pull/7620)) by @chinesepowered - fix(triage): only the bot's own approval counts as already approved ([#7737](https://github.com/QwenLM/qwen-code/pull/7737)) by @wenshao - fix(core): stop humanReadableCron naming intervals that never happen ([#7529](https://github.com/QwenLM/qwen-code/pull/7529)) by @chinesepowered - fix(cli): keep IME cursor aligned after footer updates ([#7711](https://github.com/QwenLM/qwen-code/pull/7711)) by @water-in-stone - fix(core): redact the plan argument from history after an approved exit_plan_mode ([#7197](https://github.com/QwenLM/qwen-code/pull/7197)) by @zjunothing - fix(review): correct the borrowed lenses and vacuous-test severity (follow-up to #7735/#7736) ([#7746](https://github.com/QwenLM/qwen-code/pull/7746)) by @wenshao - fix(core): reliably deliver manual plan-exit notices ([#7744](https://github.com/QwenLM/qwen-code/pull/7744)) by @doudouOUC - fix(review): recover bilingual register from the live PR when the plan omits the Han flag ([#7739](https://github.com/QwenLM/qwen-code/pull/7739)) by @wenshao - fix(cli): complete repeated skill slash commands ([#7720](https://github.com/QwenLM/qwen-code/pull/7720)) by @Sparkle6979 - fix(cli): handle escaped dollars around inline math ([#7741](https://github.com/QwenLM/qwen-code/pull/7741)) by @CubeLander - fix(cli): show tool descriptions in multi-tool compact summaries ([#7589](https://github.com/QwenLM/qwen-code/pull/7589)) by @ovochouovo - fix(ci): rename triage status marker to avoid duplicate-guard collision ([#7723](https://github.com/QwenLM/qwen-code/pull/7723)) by @yiliang114 - fix(core): route id-less continuation chunks to a colliding tool-call opener's slot ([#6981](https://github.com/QwenLM/qwen-code/pull/6981)) by @he-yufeng - fix(autofix): answer every review thread, resolve the ones actually fixed ([#7758](https://github.com/QwenLM/qwen-code/pull/7758)) by @wenshao - fix(core): treat properties as a name map in toOpenAPI30 ([#7760](https://github.com/QwenLM/qwen-code/pull/7760)) by @chinesepowered - fix(core): keep leading whitespace in gitignore patterns ([#7763](https://github.com/QwenLM/qwen-code/pull/7763)) by @chinesepowered - fix(core): stop trailing slash from anchoring nested gitignore patterns ([#7764](https://github.com/QwenLM/qwen-code/pull/7764)) by @chinesepowered - fix(core): keep the model name when a model id carries a variant tag ([#7766](https://github.com/QwenLM/qwen-code/pull/7766)) by @chinesepowered - fix(weixin): create the account credential file already private ([#7726](https://github.com/QwenLM/qwen-code/pull/7726)) by @chinesepowered - fix(core): stop rewriting backslash escapes in gitignore patterns ([#7765](https://github.com/QwenLM/qwen-code/pull/7765)) by @chinesepowered - fix(core): scope the timeout veto to the fragment it appears in ([#7776](https://github.com/QwenLM/qwen-code/pull/7776)) by @chinesepowered - fix(core): decline sed patterns whose bracket expression starts with ] ([#7775](https://github.com/QwenLM/qwen-code/pull/7775)) by @chinesepowered - fix(web-shell): stabilize mobile voice input ([#7806](https://github.com/QwenLM/qwen-code/pull/7806)) by @ytahdn - fix(test): give every E2E case a clean directory again ([#7811](https://github.com/QwenLM/qwen-code/pull/7811)) by @wenshao - fix(web-shell): allow shell commands in new tasks without a session ([#7724](https://github.com/QwenLM/qwen-code/pull/7724)) by @wenshao - fix(triage): carry the /verify lane's hardening across to /tmux ([#7753](https://github.com/QwenLM/qwen-code/pull/7753)) by @wenshao - fix(web-shell): render task notifications as system messages ([#7822](https://github.com/QwenLM/qwen-code/pull/7822)) by @ytahdn - fix(ci): add default bash shell to container jobs in qwen-triage ([#7838](https://github.com/QwenLM/qwen-code/pull/7838)) by @qwen-code-dev-bot - fix(web-shell): preserve pasted text in composer ([#7824](https://github.com/QwenLM/qwen-code/pull/7824)) by @ytahdn - fix(ci): add git safe.directory for container jobs ([#7843](https://github.com/QwenLM/qwen-code/pull/7843)) by @qwen-code-dev-bot - fix(ci): add --init to container jobs to reap zombie processes ([#7848](https://github.com/QwenLM/qwen-code/pull/7848)) by @qwen-code-dev-bot - fix(core): wait for output stream flush before settling background shells ([#7833](https://github.com/QwenLM/qwen-code/pull/7833)) by @ComplexSimply - fix(scripts): retry model calls and surface degraded release notes ([#7535](https://github.com/QwenLM/qwen-code/pull/7535)) by @he-yufeng - fix(ci): prevent worktree cleanup from failing the verify job ([#7857](https://github.com/QwenLM/qwen-code/pull/7857)) by @qwen-code-dev-bot - fix(release): publish all channel packages, not just channel-base ([#7845](https://github.com/QwenLM/qwen-code/pull/7845)) by @yiliang114 - fix(scripts): harden retry classification and preserve deadline error context ([#7854](https://github.com/QwenLM/qwen-code/pull/7854)) by @yiliang114 - fix(core): fast-fail permanent quota-exhaustion 429s instead of silent retry ([#7842](https://github.com/QwenLM/qwen-code/pull/7842)) by @yiliang114 - fix(web-shell): isolate history and session drafts ([#7810](https://github.com/QwenLM/qwen-code/pull/7810)) by @ytahdn - fix(triage): retry a transient npm ci before blaming the PR for it ([#7884](https://github.com/QwenLM/qwen-code/pull/7884)) by @wenshao - fix(webui): fall back to history_truncated marker recordId for transcript pagination anchor ([#7829](https://github.com/QwenLM/qwen-code/pull/7829)) by @wenshao - fix(review): give the review retry the remaining time budget ([#7852](https://github.com/QwenLM/qwen-code/pull/7852)) by @wenshao - fix(integration): configure Docker sandbox networking for submitted-prompt provenance test (#7879) ([#7881](https://github.com/QwenLM/qwen-code/pull/7881)) by @qwen-code-dev-bot - fix(cli): report a genuine $0.00 cost instead of N/A ([#7784](https://github.com/QwenLM/qwen-code/pull/7784)) by @chinesepowered - fix(core): reject socks5h and socks4a proxy URLs ([#7786](https://github.com/QwenLM/qwen-code/pull/7786)) by @chinesepowered - fix(core): correct the character classes in checkContentLoop ([#7788](https://github.com/QwenLM/qwen-code/pull/7788)) by @chinesepowered - fix(web-shell): make /copy with a bare index work ([#7789](https://github.com/QwenLM/qwen-code/pull/7789)) by @chinesepowered - fix(core): decline combined sed flags where -i is not last ([#7790](https://github.com/QwenLM/qwen-code/pull/7790)) by @chinesepowered - fix(core): pass the Grep pattern behind -e so a leading dash is not an option ([#7863](https://github.com/QwenLM/qwen-code/pull/7863)) by @chinesepowered - fix(core): cap a bare Error's message like every other getErrorMessage path ([#7865](https://github.com/QwenLM/qwen-code/pull/7865)) by @chinesepowered - fix(cli): make wrapToVisualLines count zero-width characters like its sibling ([#7873](https://github.com/QwenLM/qwen-code/pull/7873)) by @chinesepowered - fix(core): charge the separator and ellipsis to the preview budget ([#7874](https://github.com/QwenLM/qwen-code/pull/7874)) by @chinesepowered - fix(cli): do not count a partial trailing line when re-opening a split fence ([#7875](https://github.com/QwenLM/qwen-code/pull/7875)) by @chinesepowered - fix(cli): make /copy <message> <index> select the code block ([#7883](https://github.com/QwenLM/qwen-code/pull/7883)) by @chinesepowered - fix(test): Restore first-output benchmark measurement validity and correct its artifact schema ([#7820](https://github.com/QwenLM/qwen-code/pull/7820)) by @doudouOUC - fix(daemon): harden Todo Stop Guard continuations ([#7821](https://github.com/QwenLM/qwen-code/pull/7821)) by @doudouOUC - fix(core): read the stash reflog from the common git dir ([#7774](https://github.com/QwenLM/qwen-code/pull/7774)) by @chinesepowered - fix(core): keep Draft 4 boolean exclusive bounds in toOpenAPI30 ([#7782](https://github.com/QwenLM/qwen-code/pull/7782)) by @chinesepowered - fix(core): apply maxDepth to flat-format memory imports ([#7851](https://github.com/QwenLM/qwen-code/pull/7851)) by @chinesepowered - fix(core): render a thought part's reasoning instead of the boolean flag ([#7866](https://github.com/QwenLM/qwen-code/pull/7866)) by @chinesepowered - fix(core): bridge tool-result images for text-only models ([#7484](https://github.com/QwenLM/qwen-code/pull/7484)) by @LaZzyMan - fix(cli): patch ink to clear staticNode on indirect subtree removal ([#7816](https://github.com/QwenLM/qwen-code/pull/7816)) by @chiga0 - fix(core): auto-retry transient network errors during API calls ([#7898](https://github.com/QwenLM/qwen-code/pull/7898)) by @chiga0 - fix(cli): hide stale sticky todos from previous turns ([#7900](https://github.com/QwenLM/qwen-code/pull/7900)) by @chiga0 - fix(serve): Release managed session writer locks on shutdown ([#7812](https://github.com/QwenLM/qwen-code/pull/7812)) by @doudouOUC - fix(cli): add polling fallback for git branch name display ([#7830](https://github.com/QwenLM/qwen-code/pull/7830)) by @qwen-code-dev-bot - fix(cli): remove redundant 'Read file' prefix from @mention tool card ([#7902](https://github.com/QwenLM/qwen-code/pull/7902)) by @qwen-code-dev-bot - fix(safe-mode): preserve caller-supplied top-tier MCP servers ([#7827](https://github.com/QwenLM/qwen-code/pull/7827)) by @VitaliBabkin - fix(ci): keep the post-merge E2E signal on main alive ([#7795](https://github.com/QwenLM/qwen-code/pull/7795)) by @wenshao - fix(core): short-circuit the flush wait when the output stream cannot flush ([#7905](https://github.com/QwenLM/qwen-code/pull/7905)) by @ComplexSimply - fix(core): track quotes inside a command substitution in splitCommands ([#7870](https://github.com/QwenLM/qwen-code/pull/7870)) by @chinesepowered - fix(core): count the per-server and per-tool always-allow outcomes as approvals ([#7869](https://github.com/QwenLM/qwen-code/pull/7869)) by @chinesepowered - fix(web-shell): report intended workspace to host when starting a new chat ([#7910](https://github.com/QwenLM/qwen-code/pull/7910)) by @wenshao - fix(cli): default to virtualized terminal history ([#5738](https://github.com/QwenLM/qwen-code/pull/5738)) by @ZevGit - fix(ci): restore workspace ownership in cleanup to prevent EACCES ([#7931](https://github.com/QwenLM/qwen-code/pull/7931)) by @qwen-code-dev-bot - fix(review): recover the resolved effort when --effort is not re-threaded ([#7855](https://github.com/QwenLM/qwen-code/pull/7855)) by @wenshao - fix(scripts): slim release-note model prompts and log request timing ([#7941](https://github.com/QwenLM/qwen-code/pull/7941)) by @yiliang114 - fix(release): pin channel-base dep to exact version during release bump ([#7953](https://github.com/QwenLM/qwen-code/pull/7953)) by @yiliang114 - fix(triage): make the build-process guard diagnosable and zombie-aware ([#7858](https://github.com/QwenLM/qwen-code/pull/7858)) by @wenshao - fix(ci): give each job its own proxy wrapper directory ([#7951](https://github.com/QwenLM/qwen-code/pull/7951)) by @wenshao ### Performance - perf(cli): cache GitHub PR list in the daemon route with a 60s TTL ([#7705](https://github.com/QwenLM/qwen-code/pull/7705)) by @wenshao - perf(core): keep the volatile auto-memory section last in the system prompt ([#7651](https://github.com/QwenLM/qwen-code/pull/7651)) by @DragonnZhang - perf(web-shell): paint the composer git chip before git status completes ([#7680](https://github.com/QwenLM/qwen-code/pull/7680)) by @wenshao - perf(core): Lazy-load first-use dependencies ([#7686](https://github.com/QwenLM/qwen-code/pull/7686)) by @doudouOUC - perf(cli): replace comment-json settings parser ([#7747](https://github.com/QwenLM/qwen-code/pull/7747)) by @doudouOUC - perf(acp): Preload providers after session creation ([#7767](https://github.com/QwenLM/qwen-code/pull/7767)) by @doudouOUC - perf(core): add early Anthropic cache breakpoint on the stable system prefix ([#7912](https://github.com/QwenLM/qwen-code/pull/7912)) by @DragonnZhang - perf(ci): cut the E2E suite from ~40min to ~24min ([#7798](https://github.com/QwenLM/qwen-code/pull/7798)) by @wenshao ### Documentation - docs(channels): Document loops and proactive delivery ([#7628](https://github.com/QwenLM/qwen-code/pull/7628)) by @wenshao ### Internal Changes - refactor(autofix): extract review verification runner ([#7644](https://github.com/QwenLM/qwen-code/pull/7644)) by @qqqys - test(web-shell): capture the git-mode new-branch sub-state in the visuals suite ([#7672](https://github.com/QwenLM/qwen-code/pull/7672)) by @wenshao - Refactors system prompt assembly into a layered builder to explicitly manage stable, context, and volatile instruction layers. ([#7707](https://github.com/QwenLM/qwen-code/pull/7707)) by @DragonnZhang - Adds regression tests to verify that restored-session transcript pagination correctly URL-encodes boundaries and preserves state during retry attempts. ([#7657](https://github.com/QwenLM/qwen-code/pull/7657)) by @jay666mnj - test(cli): cover bottom-stuck virtualized list behavior ([#7652](https://github.com/QwenLM/qwen-code/pull/7652)) by @jay666mnj - ci: keep the critical-audit gate honest when npm cannot answer ([#7743](https://github.com/QwenLM/qwen-code/pull/7743)) by @wenshao - test(integration): deflake tool-control permission cases ([#7725](https://github.com/QwenLM/qwen-code/pull/7725)) by @yiliang114 - revert: drop the stale-base un-park recovery (#7602) ([#7640](https://github.com/QwenLM/qwen-code/pull/7640)) by @wenshao - test(serve): Add first-output latency benchmark ([#7761](https://github.com/QwenLM/qwen-code/pull/7761)) by @doudouOUC - test(channels): run the feishu, weixin and qqbot suites in CI ([#7853](https://github.com/QwenLM/qwen-code/pull/7853)) by @chinesepowered ## New Contributors - @jay666mnj made their first contribution in [#7674](https://github.com/QwenLM/qwen-code/pull/7674) - @harjothkhara made their first contribution in [#7203](https://github.com/QwenLM/qwen-code/pull/7203) - @PratikWayase made their first contribution in [#7709](https://github.com/QwenLM/qwen-code/pull/7709) - @Sparkle6979 made their first contribution in [#7720](https://github.com/QwenLM/qwen-code/pull/7720) - @VitaliBabkin made their first contribution in [#7827](https://github.com/QwenLM/qwen-code/pull/7827) **Full Changelog**: https://github.com/QwenLM/qwen-code/compare/v0.21.0...v0.21.1