Qwen Code v0.20.1-nightly.20260724.7d17c44a3 预发布

<!-- Release notes generated using configuration in .github/release.yml at release/v0.20.1-nightly.20260724.7d17c44a3 --> ## What's Changed * test(telemetry): Cover daemon metrics init ordering and document metricReader asymmetry by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7456 * perf(startup): Load undici lazily behind package-local dynamic imports by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7455 * fix(autofix): keep a still-red check visible until its head is judged by @wenshao in https://github.com/QwenLM/qwen-code/pull/7438 * feat(web-shell): add workspace selector button with add/switch dropdown in composer toolbar by @water-in-stone in https://github.com/QwenLM/qwen-code/pull/7390 * fix: normalize empty working_dir to unset when isolation:worktree is set on AgentTool by @mvanhorn in https://github.com/QwenLM/qwen-code/pull/7403 * fix(core): Advertise completed task revival by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/7454 * fix(cli): narrow update-check error classification by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7431 * fix(web-shell): isolate component styles from host CSS by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7466 * feat(web-shell): show subagent sessions in detail panel by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7380 * perf(web-shell): optimize long session rendering by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7408 * feat(web-shell): add rendered file previews by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7467 * feat(core): add memory recall delivery telemetry by @ZijianZhang989 in https://github.com/QwenLM/qwen-code/pull/7393 * fix(web-shell): polish embedded shell interactions by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7477 * fix(release): exclude mobile-mcp from core version bump by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7474 * fix(ci): inject actual model name into triage signature by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7475 * fix(core): align stepped day fields with Vixie cron semantics by @destire-mio in https://github.com/QwenLM/qwen-code/pull/7464 * feat(core): keep completed background agents resident by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/7426 * fix(acp-bridge): close prompt-terminal follow-ups from the PR #7400 self-review by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7453 * fix(core): strip Qwen-internal daemon secrets from agent-spawned child env by @chinesepowered in https://github.com/QwenLM/qwen-code/pull/7256 * docs(core): align JSDoc @param names with actual function signatures by @ovochouovo in https://github.com/QwenLM/qwen-code/pull/7492 * feat(serve): support forced MCP reconnects by @BZ-D in https://github.com/QwenLM/qwen-code/pull/7488 * fix(cli): insert newline on Shift+Enter and stop streaming thinking-block flicker by @chiga0 in https://github.com/QwenLM/qwen-code/pull/7397 * fix(web-shell): open singleton subagent details by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7495 * fix(web-shell): avoid redundant Git status requests by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7496 * fix(agent): ignore empty working_dir placeholders by @Truraly in https://github.com/QwenLM/qwen-code/pull/7343 * feat(prompts): allow overriding core identity via QWEN_SYSTEM_IDENTITY_MD by @zjgzx1988 in https://github.com/QwenLM/qwen-code/pull/7478 * fix(cli): yield to single-slot background agents by @hogeheer499-commits in https://github.com/QwenLM/qwen-code/pull/7258 * docs(autofix): require evidenced pre-commit verification, not a bare "verified" by @wenshao in https://github.com/QwenLM/qwen-code/pull/7486 * feat(autofix): stop a PR that fails to push for N rounds in a row by @wenshao in https://github.com/QwenLM/qwen-code/pull/7482 * feat(core): restore background agent roster by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/7459 * feat(cli): support custom skill directories via settings by @qwen-code-dev-bot in https://github.com/QwenLM/qwen-code/pull/7395 * fix(core): add image modality support for qwen3.8-max and kimi-k3 models by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7491 * fix(dingtalk): preserve non-bot mention context by @qwen-code-dev-bot in https://github.com/QwenLM/qwen-code/pull/7473 * fix(core): harden the usage salvage around session deletion by @zjunothing in https://github.com/QwenLM/qwen-code/pull/7425 * fix(core): make fork subagents discoverable by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/7460 * test(core): Cover Shell truncation without an artifact by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7470 * fix(ci): autofix route checks existing labels on non-trigger label events by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7481 * feat(cli): preserve semantic text when copying VP selections by @chiga0 in https://github.com/QwenLM/qwen-code/pull/7286 * test(core): stub the registry methods agent.ts actually calls by @chinesepowered in https://github.com/QwenLM/qwen-code/pull/7538 * perf(startup): lazy-load Google GenAI SDK on first use by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7512 * fix(vscode): use file picker image paths for vision input by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7493 * fix(cli): open the actual serve fallback port by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7501 * fix(ci): don't let one failing scenario sink the whole visual preview by @wenshao in https://github.com/QwenLM/qwen-code/pull/7511 * feat(web-shell): add selective Shadow DOM isolation by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7551 * feat(web-shell): add renderChatHeader slot for custom session header by @yuanyuanAli in https://github.com/QwenLM/qwen-code/pull/7553 * fix(cli): say review coverage gaps in the author's units, not chunk ids by @wenshao in https://github.com/QwenLM/qwen-code/pull/7550 * fix(autofix): retry a skipped-Prepare instead of stranding the PR terminal by @wenshao in https://github.com/QwenLM/qwen-code/pull/7490 * fix(cli): keep role codenames and brief paths out of the posted review body by @wenshao in https://github.com/QwenLM/qwen-code/pull/7560 * fix(autofix): retry an agent timeout instead of advancing past its feedback by @wenshao in https://github.com/QwenLM/qwen-code/pull/7563 * feat(serve): add workspace-level generation by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7552 * fix(web-shell): expose managedId in artifact open requests by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7570 * feat(serve): persist workspace channel configuration by @qqqys in https://github.com/QwenLM/qwen-code/pull/7514 * fix(sdk-python): require canonical form in validate_session_id by @chinesepowered in https://github.com/QwenLM/qwen-code/pull/7532 * fix(web-shell): sync background agent status by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7561 * feat(core): propagate trusted daemon invocation context by @callmeYe in https://github.com/QwenLM/qwen-code/pull/7279 * fix(feishu): await stream cancels in media download teardown by @chinesepowered in https://github.com/QwenLM/qwen-code/pull/7465 * fix(autofix): make the review-address report comment fully bilingual by @wenshao in https://github.com/QwenLM/qwen-code/pull/7569 * feat(cli): post the review body bilingually when the PR description is Chinese by @wenshao in https://github.com/QwenLM/qwen-code/pull/7564 * feat(autofix): auto-rerun a check that died on infrastructure, once by @wenshao in https://github.com/QwenLM/qwen-code/pull/7562 * fix(serve): detect stale SSE cursors across daemon restarts via epoch token; preserve turn attribution and surface compaction failures in replay by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7458 * feat(core): Align GenAI telemetry with ARMS by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7536 * fix(serve): avoid TOCTOU race dropping live sessions from list response by @yiliang114 with @Copilot in https://github.com/QwenLM/qwen-code/pull/7556 * fix(cli): prevent monitor notifications after task_stop by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7573 * Fix(cli): use npm view for update check instead of update-notifier (#7515) by @dtometzki in https://github.com/QwenLM/qwen-code/pull/7528 * fix(core): avoid empty transcript history pages by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7582 * fix(cli): surface unhandled rejections and render errors instead of swallowing them by @chiga0 in https://github.com/QwenLM/qwen-code/pull/7406 * fix(web-shell): isolate slash command plugin pages by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7581 * feat(web-shell): add workspace agent management by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7572 * fix(sdk-python): validate max_tool_calls and max_subagent_depth as integers by @chinesepowered in https://github.com/QwenLM/qwen-code/pull/7548 * test(autofix): single-source the infra-signature list from the workflow by @wenshao in https://github.com/QwenLM/qwen-code/pull/7565 * fix(cli): correct queued message display style and ordering by @chiga0 in https://github.com/QwenLM/qwen-code/pull/7381 * perf(cli): Defer ACP telemetry initialization by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7558 * fix(core): persist usage for tool-only subagent rounds by @DragonnZhang in https://github.com/QwenLM/qwen-code/pull/7557 * fix(core): retry requests when providers require thinking by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7534 * feat(autofix): update a stale base when the gate rejects a behind-main fix by @wenshao in https://github.com/QwenLM/qwen-code/pull/7595 * fix(core): record auto-memory index reads in FileReadCache by @han-dreamer in https://github.com/QwenLM/qwen-code/pull/7468 * fix(web-shell): initialize workspace selector from ID by @patrick-andstar in https://github.com/QwenLM/qwen-code/pull/7518 * fix(core): reject nested background requests by @patrick-andstar in https://github.com/QwenLM/qwen-code/pull/7593 * fix(web-shell): prevent React measure detail OOM by @ytahdn in https://github.com/QwenLM/qwen-code/pull/7596 * fix(core): jsonl write([]) should leave an empty file, not a stray newline by @chinesepowered in https://github.com/QwenLM/qwen-code/pull/7533 * fix(core): strip daemon secrets from hook and tool-discovery child env by @chinesepowered in https://github.com/QwenLM/qwen-code/pull/7527 * fix(core): preserve disabled reasoning effort by @WladmirJunior in https://github.com/QwenLM/qwen-code/pull/7541 * feat(sdk-java): Add daemon transport by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/7463 * fix(web-shell): render a plain textarea composer on touch devices by @ComplexSimply in https://github.com/QwenLM/qwen-code/pull/7587 * fix(core): treat backslash as literal inside single quotes in splitCommands by @chinesepowered in https://github.com/QwenLM/qwen-code/pull/7526 * fix(cli): stop treating version-manager npm shims as npm-cli.js by @nerdalytics in https://github.com/QwenLM/qwen-code/pull/7545 * fix(core,cli): isolate teammate leader turns from agent context by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/7576 * feat(autofix): auto-update a PR red only from a stale, since-fixed base by @wenshao in https://github.com/QwenLM/qwen-code/pull/7554 * test: deflake write_file content assertion in tool-control E2E test by @yiliang114 with @Copilot in https://github.com/QwenLM/qwen-code/pull/7613 * fix(sdk-java): let a terminal continuation start the next prompt by @wenshao in https://github.com/QwenLM/qwen-code/pull/7615 * feat(autofix): auto-recover a PR parked on a stale base by @wenshao in https://github.com/QwenLM/qwen-code/pull/7602 * fix(channels): route Telegram replies to forum topics by @hogeheer499-commits in https://github.com/QwenLM/qwen-code/pull/7612 * fix(acp): hide discontinued OAuth model for other auth types by @hogeheer499-commits in https://github.com/QwenLM/qwen-code/pull/7522 * test(sdk-java): widen the SSE idle-watchdog margins in the slow-line test by @wenshao in https://github.com/QwenLM/qwen-code/pull/7617 * feat(core): add Goal v3 state protocol by @qqqys in https://github.com/QwenLM/qwen-code/pull/7517 * feat(daemon): add explicit channel delivery by @BenGuanRan in https://github.com/QwenLM/qwen-code/pull/7388 * feat(serve): scope channel lifecycle to workspace runtimes by @qqqys in https://github.com/QwenLM/qwen-code/pull/7577 ## New Contributors * @ovochouovo made their first contribution in https://github.com/QwenLM/qwen-code/pull/7492 * @Truraly made their first contribution in https://github.com/QwenLM/qwen-code/pull/7343 * @zjgzx1988 made their first contribution in https://github.com/QwenLM/qwen-code/pull/7478 * @hogeheer499-commits made their first contribution in https://github.com/QwenLM/qwen-code/pull/7258 * @dtometzki made their first contribution in https://github.com/QwenLM/qwen-code/pull/7528 * @patrick-andstar made their first contribution in https://github.com/QwenLM/qwen-code/pull/7518 * @WladmirJunior made their first contribution in https://github.com/QwenLM/qwen-code/pull/7541 * @nerdalytics made their first contribution in https://github.com/QwenLM/qwen-code/pull/7545 **Full Changelog**: https://github.com/QwenLM/qwen-code/compare/v0.20.1...v0.20.1-nightly.20260724.7d17c44a3