From e27321cb2369f8709110a7e4fa6b5b445a1ae1fc Mon Sep 17 00:00:00 2001 From: Administrator Date: Mon, 29 Jun 2026 14:40:26 +0800 Subject: [PATCH 01/16] feat: enhance public task markdown preview --- .../src/pages/PublicScheduledTaskPage.tsx | 164 +++++++++++++++++- 1 file changed, 161 insertions(+), 3 deletions(-) diff --git a/frontend-web/src/pages/PublicScheduledTaskPage.tsx b/frontend-web/src/pages/PublicScheduledTaskPage.tsx index 76ec60ec..79244de9 100644 --- a/frontend-web/src/pages/PublicScheduledTaskPage.tsx +++ b/frontend-web/src/pages/PublicScheduledTaskPage.tsx @@ -2,6 +2,7 @@ import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query"; import { + BookOpen, Check, ChevronLeft, ChevronRight, @@ -46,7 +47,9 @@ import { getBackendBaseURL } from "@/core/config"; import { usePublicBackground } from "@/core/embed/background"; import { downloadMarkdownAsDocx } from "@/core/scheduled-tasks/markdown-docx"; import { reasoningPlugins } from "@/core/streamdown"; +import { createMarkdownAnchorResolver } from "@/core/utils/markdown"; import { cn } from "@/lib/utils"; +import "@/components/wiki/wiki.css"; type PublicTask = { task_id: string; @@ -264,6 +267,153 @@ function stripTimestampSuffix(title: string): string { return cleaned.trim() || title; } +function stripMarkdownToText(markdown: string): string { + return markdown + .replace(/```[\s\S]*?```/g, " ") + .replace(/~~~[\s\S]*?~~~/g, " ") + .replace(/!\[[^\]]*\]\([^)]+\)/g, " ") + .replace(/\[([^\]]+)\]\([^)]+\)/g, "$1") + .replace(/^#{1,6}\s+/gm, "") + .replace(/^[>\-*+\d.)\s]+/gm, "") + .replace(/[`*_~|]/g, "") + .replace(/<[^>]+>/g, " ") + .replace(/\s+/g, " ") + .trim(); +} + +function getMarkdownExcerpt(markdown: string): string { + const text = stripMarkdownToText(markdown); + if (!text) return ""; + return text.length > 180 ? `${text.slice(0, 180)}...` : text; +} + +function countReadableWords(markdown: string): number { + const text = stripMarkdownToText(markdown); + if (!text) return 0; + const cjk = text.match(/[\u4e00-\u9fff]/g)?.length ?? 0; + const words = text.replace(/[\u4e00-\u9fff]/g, " ").match(/[A-Za-z0-9]+/g)?.length ?? 0; + return cjk + words; +} + +function selectorForId(id: string): string { + const css = window.CSS; + if (css?.escape) return `#${css.escape(id)}`; + return `[id="${id.replace(/\\/g, "\\\\").replace(/"/g, '\\"')}"]`; +} + +function KnowledgeMarkdownView({ + content, + title, + subtitle, + className, +}: { + content: string; + title?: string; + subtitle?: string; + className?: string; + isLoading?: boolean; + remarkPlugins?: unknown; + rehypePlugins?: unknown; +}) { + const rootRef = useRef(null); + const anchorResolver = useMemo(() => createMarkdownAnchorResolver(content), [content]); + const headings = useMemo( + () => anchorResolver.headings.filter((item) => item.level >= 1 && item.level <= 3), + [anchorResolver], + ); + const excerpt = useMemo(() => getMarkdownExcerpt(content), [content]); + const sectionCount = headings.length; + const wordCount = useMemo(() => countReadableWords(content), [content]); + + const scrollToHeading = useCallback((id: string) => { + const root = rootRef.current; + const target = + root?.querySelector(selectorForId(id)) ?? document.getElementById(id); + target?.scrollIntoView({ behavior: "smooth", block: "start" }); + }, []); + + return ( +
+
+
+
+ + + 知识梳理 + + {subtitle && {subtitle}} +
+ {title && ( +

+ {title} +

+ )} + {excerpt && ( +

+ {excerpt} +

+ )} +
+ + {sectionCount || 1} 个章节 + + + 约 {wordCount} 字 + +
+
+
+ +
+
+ + +
+ ); +} + const BODY_TAB_VALUE = "__body__"; const HTML_TAB_VALUE = "__html__"; const RUN_LIST_COLLAPSED_WIDTH = 56; @@ -688,9 +838,10 @@ function PublicFullscreenPreviewContent({ preview }: { preview: PublicFullscreen } if (preview.mode === "markdown") { return ( - ) : ( - Date: Mon, 29 Jun 2026 14:40:26 +0800 Subject: [PATCH 02/16] feat: enhance public task markdown preview --- .../src/pages/PublicScheduledTaskPage.tsx | 164 +++++++++++++++++- 1 file changed, 161 insertions(+), 3 deletions(-) diff --git a/frontend-web/src/pages/PublicScheduledTaskPage.tsx b/frontend-web/src/pages/PublicScheduledTaskPage.tsx index 76ec60ec..79244de9 100644 --- a/frontend-web/src/pages/PublicScheduledTaskPage.tsx +++ b/frontend-web/src/pages/PublicScheduledTaskPage.tsx @@ -2,6 +2,7 @@ import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query"; import { + BookOpen, Check, ChevronLeft, ChevronRight, @@ -46,7 +47,9 @@ import { getBackendBaseURL } from "@/core/config"; import { usePublicBackground } from "@/core/embed/background"; import { downloadMarkdownAsDocx } from "@/core/scheduled-tasks/markdown-docx"; import { reasoningPlugins } from "@/core/streamdown"; +import { createMarkdownAnchorResolver } from "@/core/utils/markdown"; import { cn } from "@/lib/utils"; +import "@/components/wiki/wiki.css"; type PublicTask = { task_id: string; @@ -264,6 +267,153 @@ function stripTimestampSuffix(title: string): string { return cleaned.trim() || title; } +function stripMarkdownToText(markdown: string): string { + return markdown + .replace(/```[\s\S]*?```/g, " ") + .replace(/~~~[\s\S]*?~~~/g, " ") + .replace(/!\[[^\]]*\]\([^)]+\)/g, " ") + .replace(/\[([^\]]+)\]\([^)]+\)/g, "$1") + .replace(/^#{1,6}\s+/gm, "") + .replace(/^[>\-*+\d.)\s]+/gm, "") + .replace(/[`*_~|]/g, "") + .replace(/<[^>]+>/g, " ") + .replace(/\s+/g, " ") + .trim(); +} + +function getMarkdownExcerpt(markdown: string): string { + const text = stripMarkdownToText(markdown); + if (!text) return ""; + return text.length > 180 ? `${text.slice(0, 180)}...` : text; +} + +function countReadableWords(markdown: string): number { + const text = stripMarkdownToText(markdown); + if (!text) return 0; + const cjk = text.match(/[\u4e00-\u9fff]/g)?.length ?? 0; + const words = text.replace(/[\u4e00-\u9fff]/g, " ").match(/[A-Za-z0-9]+/g)?.length ?? 0; + return cjk + words; +} + +function selectorForId(id: string): string { + const css = window.CSS; + if (css?.escape) return `#${css.escape(id)}`; + return `[id="${id.replace(/\\/g, "\\\\").replace(/"/g, '\\"')}"]`; +} + +function KnowledgeMarkdownView({ + content, + title, + subtitle, + className, +}: { + content: string; + title?: string; + subtitle?: string; + className?: string; + isLoading?: boolean; + remarkPlugins?: unknown; + rehypePlugins?: unknown; +}) { + const rootRef = useRef(null); + const anchorResolver = useMemo(() => createMarkdownAnchorResolver(content), [content]); + const headings = useMemo( + () => anchorResolver.headings.filter((item) => item.level >= 1 && item.level <= 3), + [anchorResolver], + ); + const excerpt = useMemo(() => getMarkdownExcerpt(content), [content]); + const sectionCount = headings.length; + const wordCount = useMemo(() => countReadableWords(content), [content]); + + const scrollToHeading = useCallback((id: string) => { + const root = rootRef.current; + const target = + root?.querySelector(selectorForId(id)) ?? document.getElementById(id); + target?.scrollIntoView({ behavior: "smooth", block: "start" }); + }, []); + + return ( +
+
+
+
+ + + 知识梳理 + + {subtitle && {subtitle}} +
+ {title && ( +

+ {title} +

+ )} + {excerpt && ( +

+ {excerpt} +

+ )} +
+ + {sectionCount || 1} 个章节 + + + 约 {wordCount} 字 + +
+
+
+ +
+
+ + +
+ ); +} + const BODY_TAB_VALUE = "__body__"; const HTML_TAB_VALUE = "__html__"; const RUN_LIST_COLLAPSED_WIDTH = 56; @@ -688,9 +838,10 @@ function PublicFullscreenPreviewContent({ preview }: { preview: PublicFullscreen } if (preview.mode === "markdown") { return ( - ) : ( - Date: Mon, 29 Jun 2026 15:09:27 +0800 Subject: [PATCH 03/16] feat: improve agent skill selection --- .../agents/skill-transfer-selector.tsx | 41 ++++++-- .../settings/skill-settings-page.tsx | 25 +++-- frontend-web/src/pages/NewAgentPage.tsx | 94 +++---------------- 3 files changed, 57 insertions(+), 103 deletions(-) diff --git a/frontend-web/src/components/workspace/agents/skill-transfer-selector.tsx b/frontend-web/src/components/workspace/agents/skill-transfer-selector.tsx index c9ba7372..f1ec5e02 100644 --- a/frontend-web/src/components/workspace/agents/skill-transfer-selector.tsx +++ b/frontend-web/src/components/workspace/agents/skill-transfer-selector.tsx @@ -45,11 +45,32 @@ function skillLabel(skill: Skill) { function matchesQuery(skill: Skill, q: string) { if (!q) return true; - return ( - skill.name.toLowerCase().includes(q) || - (skill.name_zh ?? "").toLowerCase().includes(q) || - (skill.description ?? "").toLowerCase().includes(q) - ); + const haystack = [ + skill.name, + skill.name_zh, + skill.description, + skill.detail, + skill.category, + skill.license, + skill.source, + skill.state, + skill.owner_username, + skill.created_at, + skill.last_activity_at, + skill.use_count != null ? `使用 ${skill.use_count}` : null, + skill.view_count != null ? `查看 ${skill.view_count}` : null, + skill.patch_count != null ? `变迁 ${skill.patch_count}` : null, + ...(skill.tags ?? []).flatMap((tag) => [ + tag.name, + tag.description, + tag.color, + tag.usage_count != null ? `标签使用 ${tag.usage_count}` : null, + ]), + ] + .filter((value): value is string => typeof value === "string" && value.length > 0) + .join(" ") + .toLowerCase(); + return haystack.includes(q); } export function SkillTransferSelector({ @@ -159,7 +180,7 @@ export function SkillTransferSelector({
{ setSearch(e.currentTarget.value); @@ -239,7 +260,7 @@ export function SkillTransferSelector({ {/* Two-column transfer panes */}
diff --git a/frontend-web/src/components/workspace/settings/skill-settings-page.tsx b/frontend-web/src/components/workspace/settings/skill-settings-page.tsx index 67db57a4..4f6a3e7d 100644 --- a/frontend-web/src/components/workspace/settings/skill-settings-page.tsx +++ b/frontend-web/src/components/workspace/settings/skill-settings-page.tsx @@ -470,11 +470,11 @@ function SkillSettingsList({ onClose }: { onClose?: () => void }) { - @@ -497,11 +497,16 @@ function SkillSettingsList({ onClose }: { onClose?: () => void }) { 名称 Z-A - - } - /> - {publisher && ( -
+
+ } + /> + {!isLoading && !error && ( +
+ 共 {orderedSkills.length} 个技能 +
+ )} + {publisher && ( +
筛选发布者：
- {enabledSkills.length > 0 && ( -
- - setSkillSearch(e.currentTarget.value)} - className="h-8 pl-7 text-xs" - /> -
- )} -
- {isSkillsLoading ? ( -

技能加载中...

- ) : enabledSkills.length === 0 ? ( -

- 当前没有可用技能。 -

- ) : filteredSkills.length === 0 ? ( -

- 没有匹配的技能。 -

- ) : ( -
- {filteredSkills.map((skill) => { - const checked = selectedSkills.includes(skill.name); - return ( -
- - -
- ); - })} -
- )} -
+ Date: Mon, 29 Jun 2026 16:09:51 +0800 Subject: [PATCH 04/16] =?UTF-8?q?=E5=B9=B6=E5=8F=91=E7=89=88?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- frontend-web/src/components/wiki/wiki.css | 109 ++++++++++++ frontend-web/src/core/system-settings/api.ts | 4 + frontend-web/src/core/threads/hooks.ts | 56 +++++-- .../pages/ModelConcurrencySettingsPage.tsx | 157 ++++++++++++------ .../src/pages/PublicScheduledTaskPage.tsx | 51 +++++- .../app/gateway/routers/system_settings.py | 35 +++- .../backend/app/gateway/services.py | 73 +++++++- .../deerflow/config/system_settings.py | 10 ++ .../harness/deerflow/runtime/runs/manager.py | 19 ++- .../runtime/runs/model_concurrency.py | 39 +++++ .../backend/tests/test_model_concurrency.py | 50 ++++++ 11 files changed, 530 insertions(+), 73 deletions(-) create mode 100644 offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/model_concurrency.py diff --git a/frontend-web/src/components/wiki/wiki.css b/frontend-web/src/components/wiki/wiki.css index 1abca062..fe3b3965 100644 --- a/frontend-web/src/components/wiki/wiki.css +++ b/frontend-web/src/components/wiki/wiki.css @@ -111,6 +111,115 @@ border-color: #eef1f5; } +.public-report-markdown { + color: #334155; + font-feature-settings: "kern"; + text-rendering: optimizeLegibility; +} + +.public-report-markdown h1 { + font-size: 30px; + letter-spacing: 0; + margin-bottom: 24px; +} + +.public-report-markdown h2 { + font-size: 22px; + margin-top: 42px; + margin-bottom: 16px; + padding-bottom: 10px; +} + +.public-report-markdown h3 { + font-size: 18px; + margin-top: 30px; + margin-bottom: 12px; +} + +.public-report-markdown p, +.public-report-markdown li { + font-size: 16px; + line-height: 1.9; +} + +.public-report-markdown p { + margin-bottom: 16px; +} + +.public-report-markdown ul, +.public-report-markdown ol { + margin: 14px 0 18px; + padding-left: 26px; +} + +.public-report-markdown ul { + list-style-type: disc; +} + +.public-report-markdown ol { + list-style-type: decimal; +} + +.public-report-markdown li { + padding-left: 2px; +} + +.public-report-markdown code { + border-radius: 6px; + background: #f1f5f9; + padding: 2px 6px; + color: #be123c; + font-size: 0.92em; +} + +.public-report-markdown pre { + margin: 20px 0; + border-radius: 8px; + background: #0f172a; + padding: 16px; + overflow-x: auto; +} + +.public-report-markdown pre code { + background: transparent; + padding: 0; + color: inherit; +} + +.public-report-markdown table { + display: block; + overflow-x: auto; + border-radius: 8px; + border: 1px solid #e8ecf1; +} + +.public-report-markdown th, +.public-report-markdown td { + min-width: 120px; + vertical-align: top; +} + +.public-report-markdown blockquote { + margin: 20px 0; +} + +.dark .public-report-markdown { + color: #cbd5e1; +} + +.dark .public-report-markdown code { + background: #1e293b; + color: #f9a8d4; +} + +.dark .public-report-markdown pre { + background: #020617; +} + +.dark .public-report-markdown table { + border-color: #334155; +} + /* Anchor target offset so headings clear the floating "目录" pill. */ .wiki-markdown h1, .wiki-markdown h2, diff --git a/frontend-web/src/core/system-settings/api.ts b/frontend-web/src/core/system-settings/api.ts index dacfa808..cb3ebfcc 100644 --- a/frontend-web/src/core/system-settings/api.ts +++ b/frontend-web/src/core/system-settings/api.ts @@ -449,12 +449,16 @@ export interface ModelConcurrencySettings { enabled: boolean; default_limit: number; model_limits: Record; + user_default_limit: number; + user_model_limits: Record; } export interface ModelConcurrencyUpdate { enabled?: boolean; default_limit?: number; model_limits?: Record; + user_default_limit?: number; + user_model_limits?: Record; } export async function loadSkillCompressionSettings(): Promise { diff --git a/frontend-web/src/core/threads/hooks.ts b/frontend-web/src/core/threads/hooks.ts index 14c9a942..9ef72004 100644 --- a/frontend-web/src/core/threads/hooks.ts +++ b/frontend-web/src/core/threads/hooks.ts @@ -263,11 +263,24 @@ function isModelConcurrencyLimitError(message: string): boolean { return ( /\b429\b/.test(message) || /并发已达到上限/.test(message) || + /并发调用已达到上限/.test(message) || /使用人数较多/.test(message) || /too many requests/i.test(message) ); } +function friendlySendLimitMessage(message: string): string | null { + if (isActiveRunConflictError(message)) { + return "当前对话还在生成回答，请等待本轮回复结束，或先点“停止”再发送新的问题。"; + } + if (isModelConcurrencyLimitError(message)) { + return /当前账号对该模型/.test(message) + ? "当前账号对该模型的并发调用已达到上限，请等待已有回答结束后再试。" + : "当前使用人数较多，请稍等一下再试。"; + } + return null; +} + function buildCurrentUserLabel(): string { const auth = getStoredAuth(); if (!auth) { @@ -604,14 +617,11 @@ export function useThreadStream({ } return; } - // 上一轮还没结束又发了新问题（后端 409）：给一句明确的中文提示，而不是 - // 把原始的英文/HTTP 错误抛给用户。 - if (isActiveRunConflictError(message)) { - toast.info("当前对话还在生成回答，请等待本轮回复结束，或先点「停止」再发送新的问题。"); - return; - } - if (isModelConcurrencyLimitError(message)) { - toast.info("当前使用人数较多，请稍等一下再试。"); + // 上一轮还没结束又发了新问题（409）或并发达到上限（429）： + // 给一句明确的中文提示，而不是把原始 HTTP 错误抛给用户。 + const friendly = friendlySendLimitMessage(message); + if (friendly) { + toast.info(friendly); return; } toast.error(message); @@ -1008,11 +1018,31 @@ export function useThreadStream({ }, ); void queryClient.invalidateQueries({ queryKey: ["threads", "search"] }); - } catch (error) { - setOptimisticMessages([]); - setIsUploading(false); - throw error; - } finally { + } catch (error) { + const message = + error instanceof Error ? error.message : String(error); + const friendly = friendlySendLimitMessage(message); + if (friendly) { + setOptimisticMessages((messages) => { + const human = messages.find((item) => item.type === "human"); + const kept = human ? [human] : []; + return [ + ...kept, + { + type: "ai", + id: `opt-ai-limit-${Date.now()}`, + content: friendly, + additional_kwargs: { element: "task" }, + } as Message, + ]; + }); + setIsUploading(false); + return; + } + setOptimisticMessages([]); + setIsUploading(false); + throw error; + } finally { sendInFlightRef.current = false; } }, diff --git a/frontend-web/src/pages/ModelConcurrencySettingsPage.tsx b/frontend-web/src/pages/ModelConcurrencySettingsPage.tsx index d8c59569..ed1a5d88 100644 --- a/frontend-web/src/pages/ModelConcurrencySettingsPage.tsx +++ b/frontend-web/src/pages/ModelConcurrencySettingsPage.tsx @@ -13,6 +13,21 @@ import { useUpdateModelConcurrencySettings, } from "@/core/system-settings/hooks"; +function cleanLimitMap(values: Record, label: string): Record | null { + const cleaned: Record = {}; + for (const [name, rawValue] of Object.entries(values)) { + const trimmed = rawValue.trim(); + if (!trimmed) continue; + const parsed = Math.round(Number(trimmed)); + if (!Number.isFinite(parsed) || parsed < 1) { + toast.error(`${label}「${name}」必须是大于 0 的整数`); + return null; + } + cleaned[name] = Math.min(parsed, 10000); + } + return cleaned; +} + export default function ModelConcurrencySettingsPage({ embedded = false, }: { @@ -24,13 +39,16 @@ export default function ModelConcurrencySettingsPage({ const [enabled, setEnabled] = useState(false); const [defaultLimit, setDefaultLimit] = useState(10); + const [userDefaultLimit, setUserDefaultLimit] = useState(0); const [modelLimits, setModelLimits] = useState>({}); + const [userModelLimits, setUserModelLimits] = useState>({}); const initializedRef = useRef(false); useEffect(() => { if (!data || initializedRef.current) return; setEnabled(data.enabled); setDefaultLimit(data.default_limit); + setUserDefaultLimit(data.user_default_limit ?? 0); setModelLimits( Object.fromEntries( Object.entries(data.model_limits ?? {}).map(([name, limit]) => [ @@ -39,39 +57,45 @@ export default function ModelConcurrencySettingsPage({ ]), ), ); + setUserModelLimits( + Object.fromEntries( + Object.entries(data.user_model_limits ?? {}).map(([name, limit]) => [ + name, + String(limit), + ]), + ), + ); initializedRef.current = true; }, [data]); const modelNames = useMemo(() => models.map((model) => model.name), [models]); - const handleModelLimitChange = (modelName: string, value: string) => { - setModelLimits((prev) => ({ ...prev, [modelName]: value })); - }; - const handleSave = async () => { const nextDefault = Math.max(1, Math.min(Math.round(defaultLimit) || 10, 10000)); - const cleaned: Record = {}; - for (const [name, rawValue] of Object.entries(modelLimits)) { - const trimmed = rawValue.trim(); - if (!trimmed) continue; - const parsed = Math.round(Number(trimmed)); - if (!Number.isFinite(parsed) || parsed < 1) { - toast.error(`模型 ${name} 的并发上限必须是大于 0 的整数`); - return; - } - cleaned[name] = Math.min(parsed, 10000); - } + const nextUserDefault = Math.max(0, Math.min(Math.round(userDefaultLimit) || 0, 10000)); + const cleanedModelLimits = cleanLimitMap(modelLimits, "模型总上限"); + if (!cleanedModelLimits) return; + const cleanedUserModelLimits = cleanLimitMap(userModelLimits, "单用户上限"); + if (!cleanedUserModelLimits) return; try { await update.mutateAsync({ enabled, default_limit: nextDefault, - model_limits: cleaned, + model_limits: cleanedModelLimits, + user_default_limit: nextUserDefault, + user_model_limits: cleanedUserModelLimits, }); setDefaultLimit(nextDefault); + setUserDefaultLimit(nextUserDefault); setModelLimits( Object.fromEntries( - Object.entries(cleaned).map(([name, limit]) => [name, String(limit)]), + Object.entries(cleanedModelLimits).map(([name, limit]) => [name, String(limit)]), + ), + ); + setUserModelLimits( + Object.fromEntries( + Object.entries(cleanedUserModelLimits).map(([name, limit]) => [name, String(limit)]), ), ); toast.success("已保存"); @@ -85,7 +109,7 @@ export default function ModelConcurrencySettingsPage({ className={ embedded ? "flex w-full flex-col gap-6" - : "mx-auto flex h-full min-h-0 w-full max-w-3xl flex-col gap-6 overflow-y-auto px-6 py-10" + : "mx-auto flex h-full min-h-0 w-full max-w-4xl flex-col gap-6 overflow-y-auto px-6 py-10" } >
@@ -93,7 +117,7 @@ export default function ModelConcurrencySettingsPage({ 模型并发限制

- 按模型限制同时进行中的问答数。达到上限后，新请求会收到“当前使用人数较多，请稍等一下再试”的提示。 + 按模型控制系统总并发，也可以限制单个用户对某个模型同时发起的调用数。达到上限时会在发送问题后立即提示，问题会保留在对话中。

@@ -102,42 +126,59 @@ export default function ModelConcurrencySettingsPage({ ) : error ? ( -
+
加载失败：{error.message}
) : ( -
+
开启限制
- 关闭后，不再按模型并发数拦截问答请求。 + 关闭后，不再按模型或用户并发数拦截问答请求。
-
-

默认上限

-

- 未单独配置的模型，使用这个并发上限。 -

- setDefaultLimit(Number(e.target.value))} - className="w-32" - /> +
+
+

模型默认总上限

+

+ 未单独配置的模型，使用这个系统总并发上限。 +

+ setDefaultLimit(Number(e.target.value))} + className="w-32" + /> +
+
+

单用户默认上限

+

+ 同一用户对同一模型的默认并发上限。填 0 表示不启用默认用户限制。 +

+ setUserDefaultLimit(Number(e.target.value))} + className="w-32" + /> +

按模型单独设置

- 留空表示使用默认上限。 + 留空表示使用对应默认值；单用户上限填空且默认值为 0 时，不限制该维度。

@@ -150,24 +191,42 @@ export default function ModelConcurrencySettingsPage({ modelNames.map((modelName) => (
{modelName}
- 留空时使用默认上限 {defaultLimit || 10} + 默认总上限 {defaultLimit || 10}；默认单用户上限 {userDefaultLimit || "不限制"}
- handleModelLimitChange(modelName, e.target.value)} - className="w-32" - placeholder={String(defaultLimit || 10)} - /> + +
)) )} diff --git a/frontend-web/src/pages/PublicScheduledTaskPage.tsx b/frontend-web/src/pages/PublicScheduledTaskPage.tsx index 79244de9..190493e1 100644 --- a/frontend-web/src/pages/PublicScheduledTaskPage.tsx +++ b/frontend-web/src/pages/PublicScheduledTaskPage.tsx @@ -321,14 +321,49 @@ function KnowledgeMarkdownView({ () => anchorResolver.headings.filter((item) => item.level >= 1 && item.level <= 3), [anchorResolver], ); + const [activeHeadingId, setActiveHeadingId] = useState(null); const excerpt = useMemo(() => getMarkdownExcerpt(content), [content]); const sectionCount = headings.length; const wordCount = useMemo(() => countReadableWords(content), [content]); + useEffect(() => { + setActiveHeadingId(headings[0]?.id ?? null); + }, [headings]); + + useEffect(() => { + const root = rootRef.current; + if (!root || headings.length === 0) return; + + const targets = headings + .map((item) => root.querySelector(selectorForId(item.id))) + .filter((item): item is HTMLElement => !!item); + if (targets.length === 0) return; + + const observer = new IntersectionObserver( + (entries) => { + const visible = entries + .filter((entry) => entry.isIntersecting) + .sort((a, b) => a.boundingClientRect.top - b.boundingClientRect.top)[0]; + if (visible?.target.id) { + setActiveHeadingId(visible.target.id); + } + }, + { + root: root.closest("[data-radix-scroll-area-viewport]") ?? null, + rootMargin: "-96px 0px -65% 0px", + threshold: [0, 1], + }, + ); + + targets.forEach((target) => observer.observe(target)); + return () => observer.disconnect(); + }, [headings, content]); + const scrollToHeading = useCallback((id: string) => { const root = rootRef.current; const target = root?.querySelector(selectorForId(id)) ?? document.getElementById(id); + setActiveHeadingId(id); target?.scrollIntoView({ behavior: "smooth", block: "start" }); }, []); @@ -336,12 +371,12 @@ function KnowledgeMarkdownView({
-
-
+
+
@@ -368,7 +403,7 @@ function KnowledgeMarkdownView({
-
+
+
+ ); +} diff --git a/frontend-web/src/components/wiki/wiki.css b/frontend-web/src/components/wiki/wiki.css index fe3b3965..5c510c4e 100644 --- a/frontend-web/src/components/wiki/wiki.css +++ b/frontend-web/src/components/wiki/wiki.css @@ -220,6 +220,231 @@ border-color: #334155; } +.report-markdown-shell { + align-items: start; + padding: 2px; +} + +.report-markdown-paper { + border: 1px solid rgba(203, 213, 225, 0.85); + border-radius: 8px; + background: + linear-gradient(180deg, rgba(248, 250, 252, 0.92), rgba(255, 255, 255, 0) 220px), + #ffffff; + box-shadow: + 0 1px 2px rgba(15, 23, 42, 0.04), + 0 18px 44px rgba(15, 23, 42, 0.07); +} + +.report-markdown-header { + border-bottom: 1px solid #e8edf4; + padding: 26px 34px 22px; +} + +.report-markdown-header-compact { + padding-bottom: 18px; +} + +.report-markdown-badge { + gap: 6px; + border: 1px solid #dbe3ef; + background: #ffffff; + color: #475569; +} + +.report-markdown-title { + margin: 14px 0 0; + color: #0f172a; + font-size: 26px; + font-weight: 700; + line-height: 1.35; + letter-spacing: 0; +} + +.report-markdown-excerpt { + margin: 14px 0 0; + max-width: 840px; + color: #64748b; + font-size: 14px; + line-height: 1.8; +} + +.report-markdown-metrics { + display: flex; + flex-wrap: wrap; + gap: 8px; + margin-top: 16px; + color: #64748b; + font-size: 12px; +} + +.report-markdown-metrics span { + border: 1px solid #e2e8f0; + border-radius: 6px; + background: rgba(255, 255, 255, 0.78); + padding: 4px 8px; +} + +.report-markdown-body { + max-width: 920px; + margin: 0 auto; + padding: 32px 42px 44px; +} + +.report-markdown-body h1 { + font-size: 28px; +} + +.report-markdown-body h2 { + position: relative; + margin-top: 44px; + border-bottom-color: #e5edf7; + font-size: 22px; +} + +.report-markdown-body h2::before { + content: ""; + position: absolute; + left: -16px; + top: 0.34em; + width: 4px; + height: 1.05em; + border-radius: 999px; + background: #2563eb; +} + +.report-markdown-body h3 { + color: #1e293b; + font-size: 18px; +} + +.report-markdown-body p, +.report-markdown-body li { + font-size: 16px; + line-height: 1.92; +} + +.report-markdown-body blockquote { + border-left-color: #2563eb; + background: #f8fbff; +} + +.report-markdown-body table { + box-shadow: 0 1px 0 rgba(15, 23, 42, 0.04); +} + +.report-markdown-toc { + position: sticky; + top: 16px; + max-height: calc(100vh - 32px); + overflow: hidden; + border: 1px solid rgba(203, 213, 225, 0.82); + border-radius: 8px; + background: rgba(255, 255, 255, 0.92); + padding: 14px; + box-shadow: 0 12px 34px rgba(15, 23, 42, 0.06); + backdrop-filter: blur(10px); +} + +.report-markdown-toc-list { + max-height: calc(100vh - 128px); + overflow: auto; + padding-right: 4px; +} + +.report-markdown-toc-item { + display: block; + width: 100%; + overflow: hidden; + text-overflow: ellipsis; + white-space: nowrap; + border-left: 2px solid transparent; + border-radius: 6px; + padding: 7px 8px; + color: #64748b; + font-size: 12px; + line-height: 1.4; + text-align: left; + transition: background 160ms ease, border-color 160ms ease, color 160ms ease; +} + +.report-markdown-toc-item:hover { + background: #f1f5f9; + color: #0f172a; +} + +.report-markdown-toc-item-active { + border-left-color: #2563eb; + background: #eff6ff; + color: #1d4ed8; + font-weight: 600; +} + +.report-markdown-toc-level-2 { + padding-left: 18px; +} + +.report-markdown-toc-level-3 { + padding-left: 30px; + color: #94a3b8; +} + +.dark .report-markdown-paper { + border-color: #1f2937; + background: + linear-gradient(180deg, rgba(15, 23, 42, 0.9), rgba(2, 6, 23, 0) 220px), + #020617; + box-shadow: 0 18px 44px rgba(0, 0, 0, 0.24); +} + +.dark .report-markdown-header { + border-bottom-color: #1e293b; +} + +.dark .report-markdown-badge, +.dark .report-markdown-metrics span { + border-color: #334155; + background: #0f172a; + color: #cbd5e1; +} + +.dark .report-markdown-title { + color: #f8fafc; +} + +.dark .report-markdown-excerpt, +.dark .report-markdown-metrics { + color: #94a3b8; +} + +.dark .report-markdown-body h3 { + color: #e2e8f0; +} + +.dark .report-markdown-body h2::before { + background: #60a5fa; +} + +.dark .report-markdown-body blockquote { + background: #0f172a; + border-left-color: #60a5fa; +} + +.dark .report-markdown-toc { + border-color: #1f2937; + background: rgba(2, 6, 23, 0.88); +} + +.dark .report-markdown-toc-item:hover { + background: #0f172a; + color: #f8fafc; +} + +.dark .report-markdown-toc-item-active { + background: rgba(37, 99, 235, 0.16); + color: #93c5fd; +} + /* Anchor target offset so headings clear the floating "目录" pill. */ .wiki-markdown h1, .wiki-markdown h2, diff --git a/frontend-web/src/pages/PublicScheduledTaskPage.tsx b/frontend-web/src/pages/PublicScheduledTaskPage.tsx index 190493e1..fdd5de51 100644 --- a/frontend-web/src/pages/PublicScheduledTaskPage.tsx +++ b/frontend-web/src/pages/PublicScheduledTaskPage.tsx @@ -24,6 +24,11 @@ import { toast } from "sonner"; import { Badge } from "@/components/ui/badge"; import { Button } from "@/components/ui/button"; import { Card, CardContent, CardHeader, CardTitle } from "@/components/ui/card"; +import { + Collapsible, + CollapsibleContent, + CollapsibleTrigger, +} from "@/components/ui/collapsible"; import { Dialog, DialogContent, @@ -43,13 +48,13 @@ import { } from "@/components/ui/tooltip"; import { MarkdownContent } from "@/components/workspace/messages/markdown-content"; import { HtmlPreviewFrame } from "@/components/html-preview-frame"; +import { ReportMarkdownView } from "@/components/wiki/ReportMarkdownView"; import { getBackendBaseURL } from "@/core/config"; import { usePublicBackground } from "@/core/embed/background"; import { downloadMarkdownAsDocx } from "@/core/scheduled-tasks/markdown-docx"; import { reasoningPlugins } from "@/core/streamdown"; import { createMarkdownAnchorResolver } from "@/core/utils/markdown"; import { cn } from "@/lib/utils"; -import "@/components/wiki/wiki.css"; type PublicTask = { task_id: string; @@ -875,13 +880,10 @@ function PublicFullscreenPreviewContent({ preview }: { preview: PublicFullscreen } if (preview.mode === "markdown") { return ( - ); } @@ -889,13 +891,10 @@ function PublicFullscreenPreviewContent({ preview }: { preview: PublicFullscreen const content = preview.file.content || "无文件内容"; if (isMarkdownFile(preview.file)) { return ( - ); } @@ -920,6 +919,7 @@ function publicFullscreenHref(taskId: string, runId: string, tab: ContentTab): s function RunDetailPane({ run, taskId }: { run: PublicRun; taskId: string }) { const queryClient = useQueryClient(); const [fullscreenPreview, setFullscreenPreview] = useState(null); + const [runMetaOpen, setRunMetaOpen] = useState(false); // HTML page run: render a sandboxed iframe preview instead of the MD tabs. const isHtmlRun = @@ -1159,7 +1159,29 @@ function RunDetailPane({ run, taskId }: { run: PublicRun; taskId: string }) { )} -

+ + + + + +

+
计划：{formatDate(run.scheduled_for)}
+
开始：{formatDate(run.started_at)}
+
完成：{formatDate(run.finished_at)}
+
+ + +

开始：{formatDate(run.started_at)} · 完成：{formatDate(run.finished_at)}

@@ -1311,7 +1333,7 @@ function RunDetailPane({ run, taskId }: { run: PublicRun; taskId: string }) { 无内容。点击「编辑」开始添加。
) : ( - )} diff --git a/frontend-web/src/pages/ScheduledTaskRunDetailPage.tsx b/frontend-web/src/pages/ScheduledTaskRunDetailPage.tsx index 9002fbcb..125c2252 100644 --- a/frontend-web/src/pages/ScheduledTaskRunDetailPage.tsx +++ b/frontend-web/src/pages/ScheduledTaskRunDetailPage.tsx @@ -20,6 +20,7 @@ import { } from "@/components/workspace/workspace-container"; import { MarkdownContent } from "@/components/workspace/messages/markdown-content"; import { HtmlPreviewFrame } from "@/components/html-preview-frame"; +import { ReportMarkdownView } from "@/components/wiki/ReportMarkdownView"; import { downloadScheduledTaskRun, downloadScheduledTaskRunFile, @@ -91,11 +92,12 @@ function RenderPreview({ file }: { file: ScheduledTaskResultFileContent }) { } if (isMarkdownFile(file)) { return ( - ); } @@ -114,11 +116,11 @@ function FullscreenPreviewContent({ preview }: { preview: FullscreenPreview }) { } if (preview.mode === "markdown") { return ( - ); } @@ -653,11 +655,12 @@ export default function ScheduledTaskRunDetailPage() {
{result.format === "markdown" || result.format === "mixed" ? ( - ) : ( 

From be06cac2235dc610540d7819951c478b5277273b Mon Sep 17 00:00:00 2001
From: Rocky Kwok 
Date: Mon, 29 Jun 2026 17:12:18 +0800
Subject: [PATCH 07/16] fix: reset code

---
 .../workspace/agents/agent-selector.tsx       | 68 +++++++++----------
 .../plugin-management/domain-agent.test.mjs   | 22 ------
 frontend-web/src/pages/IframeChatPage.tsx     | 61 +++++------------
 3 files changed, 50 insertions(+), 101 deletions(-)

diff --git a/frontend-web/src/components/workspace/agents/agent-selector.tsx b/frontend-web/src/components/workspace/agents/agent-selector.tsx
index 9c031ac2..67afc60c 100644
--- a/frontend-web/src/components/workspace/agents/agent-selector.tsx
+++ b/frontend-web/src/components/workspace/agents/agent-selector.tsx
@@ -20,7 +20,6 @@ import {
 } from "@/components/ui/dialog";
 import { PromptInputButton } from "@/components/ai-elements/prompt-input";
 import { AgentManageDialog } from "@/components/workspace/agents/agent-manage-dialog";
-import { Tooltip } from "@/components/workspace/tooltip";
 import { useAgentSelector, useAgentsPaginated } from "@/core/agents";
 import type { Agent } from "@/core/agents";
 import { getStoredAuth } from "@/core/auth";
@@ -87,7 +86,6 @@ export function AgentSelector({
   };
 
   const triggerLabel = selectedAgent ? desensitize(selectedAgent.name) : "选择 Agent";
-  const iconTooltipLabel = selectedAgent ? triggerLabel : "未选择";
   const iconMode = triggerVariant === "icon";
 
   return (
@@ -95,40 +93,38 @@ export function AgentSelector({
       
         
           {iconMode ? (
-            
-              
-            
+            
           ) : (
              {
-  assert.deepEqual(
-    resolveIncomingIframeSelection({
-      pageUrl: "https://another-unknown.example.com/article",
-      mappings: [
-        {
-          id: "m1",
-          domain: "foo.example.com",
-          agentId: "agent-1",
-          enabled: true,
-        },
-      ],
-      availableAgentIds: ["agent-1", "agent-2"],
-      manualSelectedAgentId: null,
-    }),
-    {
-      selectedAgentId: null,
-      forceDefaultAssistant: true,
-    },
-  );
 });
\ No newline at end of file
diff --git a/frontend-web/src/pages/IframeChatPage.tsx b/frontend-web/src/pages/IframeChatPage.tsx
index 6a6aadae..e1088dfe 100644
--- a/frontend-web/src/pages/IframeChatPage.tsx
+++ b/frontend-web/src/pages/IframeChatPage.tsx
@@ -83,13 +83,6 @@ type BrainyChatInterruptMessage = {
   };
   requestId?: string;
 };
-type BrainyUrlContextMessage = {
-  source?: string;
-  type?: string;
-  payload?: {
-    url?: string;
-  };
-};
 type PendingIframeSubmit = {
   message: PromptInputMessage;
   externalContext?: Record;
@@ -436,33 +429,9 @@ export default function IframeChatPage() {
     [thread],
   );
 
-  const applyIncomingUrlSelection = useCallback(
-    (pageUrl: string | null | undefined) => {
-      const nextSelection = resolveIncomingIframeSelection({
-        pageUrl,
-        mappings: pluginManagementSettings?.domainAgentMappings ?? [],
-        availableAgentIds: selectorLoading
-          ? undefined
-          : selectorAgents.map((agent) => agent.id),
-        manualSelectedAgentId,
-      });
-      setSelectedAgentId(nextSelection.selectedAgentId);
-      setForceDefaultAssistant(nextSelection.forceDefaultAssistant);
-      return nextSelection;
-    },
-    [
-      manualSelectedAgentId,
-      pluginManagementSettings?.domainAgentMappings,
-      selectorAgents,
-      selectorLoading,
-    ],
-  );
-
   useEffect(() => {
     const handleMessage = (
-      event: MessageEvent<
-        BrainyChatSubmitMessage | BrainyChatInterruptMessage | BrainyUrlContextMessage
-      >,
+      event: MessageEvent,
     ) => {
       const data = event.data;
       if (data?.source !== "brainy-extension") return;
@@ -470,12 +439,6 @@ export default function IframeChatPage() {
         void handleInterrupt(data as BrainyChatInterruptMessage);
         return;
       }
-      if (data?.type === "brainy:context-url") {
-        applyIncomingUrlSelection(
-          (data as BrainyUrlContextMessage).payload?.url ?? null,
-        );
-        return;
-      }
       if (data?.type !== "brainy:chat-submit") return;
       const submitMessage = data as BrainyChatSubmitMessage;
       const { submitText, files, iframeContext, additionalKwargs } =
@@ -483,6 +446,17 @@ export default function IframeChatPage() {
 
       if (!submitText && files.length === 0) return;
 
+      const nextSelection = resolveIncomingIframeSelection({
+        pageUrl: submitMessage.payload?.pageContext?.url,
+        mappings: pluginManagementSettings?.domainAgentMappings ?? [],
+        availableAgentIds: selectorLoading
+          ? undefined
+          : selectorAgents.map((agent) => agent.id),
+        manualSelectedAgentId,
+      });
+      setSelectedAgentId(nextSelection.selectedAgentId);
+      setForceDefaultAssistant(nextSelection.forceDefaultAssistant);
+
       console.log("收到插件消息", {
         text: submitMessage.payload?.text,
         prompt: submitMessage.payload?.prompt,
@@ -499,8 +473,8 @@ export default function IframeChatPage() {
         message: { text: submitText, files },
         externalContext: iframeContext,
         options: additionalKwargs ? { additionalKwargs } : undefined,
-        targetSelectedAgentId: selectedAgentId,
-        targetForceDefaultAssistant: forceDefaultAssistant,
+        targetSelectedAgentId: nextSelection.selectedAgentId,
+        targetForceDefaultAssistant: nextSelection.forceDefaultAssistant,
       });
     };
 
@@ -509,10 +483,11 @@ export default function IframeChatPage() {
       window.removeEventListener("message", handleMessage);
     };
   }, [
-    applyIncomingUrlSelection,
-    forceDefaultAssistant,
     handleInterrupt,
-    selectedAgentId,
+    manualSelectedAgentId,
+    pluginManagementSettings?.domainAgentMappings,
+    selectorAgents,
+    selectorLoading,
   ]);
 
   const isLandingState = isNewThread && !hasStartedConversation;

From 794aab10078b7ab35bee25d193353c73c304562d Mon Sep 17 00:00:00 2001
From: Rocky Kwok 
Date: Mon, 29 Jun 2026 17:14:01 +0800
Subject: [PATCH 08/16] fix: recode

---
 .../workspace/agents/agent-selector.tsx       | 68 +++++++++--------
 .../plugin-management/domain-agent.test.mjs   | 22 ++++++
 frontend-web/src/pages/IframeChatPage.tsx     | 73 +++++++++++++------
 3 files changed, 107 insertions(+), 56 deletions(-)

diff --git a/frontend-web/src/components/workspace/agents/agent-selector.tsx b/frontend-web/src/components/workspace/agents/agent-selector.tsx
index 67afc60c..9c031ac2 100644
--- a/frontend-web/src/components/workspace/agents/agent-selector.tsx
+++ b/frontend-web/src/components/workspace/agents/agent-selector.tsx
@@ -20,6 +20,7 @@ import {
 } from "@/components/ui/dialog";
 import { PromptInputButton } from "@/components/ai-elements/prompt-input";
 import { AgentManageDialog } from "@/components/workspace/agents/agent-manage-dialog";
+import { Tooltip } from "@/components/workspace/tooltip";
 import { useAgentSelector, useAgentsPaginated } from "@/core/agents";
 import type { Agent } from "@/core/agents";
 import { getStoredAuth } from "@/core/auth";
@@ -86,6 +87,7 @@ export function AgentSelector({
   };
 
   const triggerLabel = selectedAgent ? desensitize(selectedAgent.name) : "选择 Agent";
+  const iconTooltipLabel = selectedAgent ? triggerLabel : "未选择";
   const iconMode = triggerVariant === "icon";
 
   return (
@@ -93,38 +95,40 @@ export function AgentSelector({
       
         
           {iconMode ? (
-            
+            
+              
+            
           ) : (
              {
+  assert.deepEqual(
+    resolveIncomingIframeSelection({
+      pageUrl: "https://another-unknown.example.com/article",
+      mappings: [
+        {
+          id: "m1",
+          domain: "foo.example.com",
+          agentId: "agent-1",
+          enabled: true,
+        },
+      ],
+      availableAgentIds: ["agent-1", "agent-2"],
+      manualSelectedAgentId: null,
+    }),
+    {
+      selectedAgentId: null,
+      forceDefaultAssistant: true,
+    },
+  );
 });
\ No newline at end of file
diff --git a/frontend-web/src/pages/IframeChatPage.tsx b/frontend-web/src/pages/IframeChatPage.tsx
index e1088dfe..6a6aadae 100644
--- a/frontend-web/src/pages/IframeChatPage.tsx
+++ b/frontend-web/src/pages/IframeChatPage.tsx
@@ -83,6 +83,13 @@ type BrainyChatInterruptMessage = {
   };
   requestId?: string;
 };
+type BrainyUrlContextMessage = {
+  source?: string;
+  type?: string;
+  payload?: {
+    url?: string;
+  };
+};
 type PendingIframeSubmit = {
   message: PromptInputMessage;
   externalContext?: Record;
@@ -429,25 +436,10 @@ export default function IframeChatPage() {
     [thread],
   );
 
-  useEffect(() => {
-    const handleMessage = (
-      event: MessageEvent,
-    ) => {
-      const data = event.data;
-      if (data?.source !== "brainy-extension") return;
-      if (data?.type === "brainy:chat-interrupt") {
-        void handleInterrupt(data as BrainyChatInterruptMessage);
-        return;
-      }
-      if (data?.type !== "brainy:chat-submit") return;
-      const submitMessage = data as BrainyChatSubmitMessage;
-      const { submitText, files, iframeContext, additionalKwargs } =
-        parseBrainyChatSubmitMessage(submitMessage);
-
-      if (!submitText && files.length === 0) return;
-
+  const applyIncomingUrlSelection = useCallback(
+    (pageUrl: string | null | undefined) => {
       const nextSelection = resolveIncomingIframeSelection({
-        pageUrl: submitMessage.payload?.pageContext?.url,
+        pageUrl,
         mappings: pluginManagementSettings?.domainAgentMappings ?? [],
         availableAgentIds: selectorLoading
           ? undefined
@@ -456,6 +448,40 @@ export default function IframeChatPage() {
       });
       setSelectedAgentId(nextSelection.selectedAgentId);
       setForceDefaultAssistant(nextSelection.forceDefaultAssistant);
+      return nextSelection;
+    },
+    [
+      manualSelectedAgentId,
+      pluginManagementSettings?.domainAgentMappings,
+      selectorAgents,
+      selectorLoading,
+    ],
+  );
+
+  useEffect(() => {
+    const handleMessage = (
+      event: MessageEvent<
+        BrainyChatSubmitMessage | BrainyChatInterruptMessage | BrainyUrlContextMessage
+      >,
+    ) => {
+      const data = event.data;
+      if (data?.source !== "brainy-extension") return;
+      if (data?.type === "brainy:chat-interrupt") {
+        void handleInterrupt(data as BrainyChatInterruptMessage);
+        return;
+      }
+      if (data?.type === "brainy:context-url") {
+        applyIncomingUrlSelection(
+          (data as BrainyUrlContextMessage).payload?.url ?? null,
+        );
+        return;
+      }
+      if (data?.type !== "brainy:chat-submit") return;
+      const submitMessage = data as BrainyChatSubmitMessage;
+      const { submitText, files, iframeContext, additionalKwargs } =
+        parseBrainyChatSubmitMessage(submitMessage);
+
+      if (!submitText && files.length === 0) return;
 
       console.log("收到插件消息", {
         text: submitMessage.payload?.text,
@@ -473,8 +499,8 @@ export default function IframeChatPage() {
         message: { text: submitText, files },
         externalContext: iframeContext,
         options: additionalKwargs ? { additionalKwargs } : undefined,
-        targetSelectedAgentId: nextSelection.selectedAgentId,
-        targetForceDefaultAssistant: nextSelection.forceDefaultAssistant,
+        targetSelectedAgentId: selectedAgentId,
+        targetForceDefaultAssistant: forceDefaultAssistant,
       });
     };
 
@@ -483,11 +509,10 @@ export default function IframeChatPage() {
       window.removeEventListener("message", handleMessage);
     };
   }, [
+    applyIncomingUrlSelection,
+    forceDefaultAssistant,
     handleInterrupt,
-    manualSelectedAgentId,
-    pluginManagementSettings?.domainAgentMappings,
-    selectorAgents,
-    selectorLoading,
+    selectedAgentId,
   ]);
 
   const isLandingState = isNewThread && !hasStartedConversation;

From 5a420189b1e5c2ea187eb577c5e3d72ca72f002a Mon Sep 17 00:00:00 2001
From: Rocky Kwok 
Date: Mon, 29 Jun 2026 17:35:36 +0800
Subject: [PATCH 09/16] =?UTF-8?q?fix:=20iframe=E5=BA=95=E9=83=A8=E9=80=89?=
 =?UTF-8?q?=E6=8B=A9=E6=99=BA=E8=83=BD=E4=BD=93?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../workspace/agents/agent-selector.tsx       | 84 ++++++++++---------
 1 file changed, 44 insertions(+), 40 deletions(-)

diff --git a/frontend-web/src/components/workspace/agents/agent-selector.tsx b/frontend-web/src/components/workspace/agents/agent-selector.tsx
index 9c031ac2..0f954f10 100644
--- a/frontend-web/src/components/workspace/agents/agent-selector.tsx
+++ b/frontend-web/src/components/workspace/agents/agent-selector.tsx
@@ -93,43 +93,47 @@ export function AgentSelector({
   return (
     <>
       
-        
-          {iconMode ? (
-            
-              
-            
-          ) : (
+        {iconMode ? (
+          
+            

+              
+                
+              
+            

+          
+        ) : (
+          
             
               )}
             
-          )}
-        
+          
+        )}
         
           选择 Agent
           
@@ -235,4 +239,4 @@ export function AgentSelector({
       )}
     
   );
-}
\ No newline at end of file
+}

From b4206d6421e3088746224e03026560a973a40258 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=90=89=E9=91=AB=E5=8D=87?= 
Date: Tue, 30 Jun 2026 11:25:44 +0800
Subject: [PATCH 10/16] Add Redis-backed concurrency gate

---
 .../backend/app/gateway/app.py                |  20 +-
 .../backend/app/gateway/deps.py               |  19 +-
 .../app/gateway/routers/admin_active_runs.py  |  47 ++-
 .../backend/app/gateway/services.py           | 154 ++++++--
 offline-backend-20260512/backend/config.yaml  |  20 +
 .../harness/deerflow/config/app_config.py     |  10 +-
 .../deerflow/config/concurrency_config.py     |  35 ++
 .../harness/deerflow/runtime/__init__.py      |  10 +
 .../harness/deerflow/runtime/runs/__init__.py |   6 +
 .../deerflow/runtime/runs/concurrency_gate.py | 369 ++++++++++++++++++
 .../harness/deerflow/runtime/runs/manager.py  |   3 +-
 .../backend/packages/harness/pyproject.toml   |   7 +-
 12 files changed, 654 insertions(+), 46 deletions(-)
 create mode 100644 offline-backend-20260512/backend/packages/harness/deerflow/config/concurrency_config.py
 create mode 100644 offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/concurrency_gate.py

diff --git a/offline-backend-20260512/backend/app/gateway/app.py b/offline-backend-20260512/backend/app/gateway/app.py
index fe20b1ce..ff972426 100644
--- a/offline-backend-20260512/backend/app/gateway/app.py
+++ b/offline-backend-20260512/backend/app/gateway/app.py
@@ -853,8 +853,9 @@ async def _concurrency_sampler_loop(app: FastAPI) -> None:
         logger.info("Concurrency sampler disabled by CONCURRENCY_SAMPLE_ENABLED")
         return
 
-    store = getattr(app.state, "concurrency_sample_store", None)
-    run_mgr = getattr(app.state, "run_manager", None)
+    store = getattr(app.state, "concurrency_sample_store", None)
+    run_mgr = getattr(app.state, "run_manager", None)
+    concurrency_gate = getattr(app.state, "concurrency_gate", None)
     if store is None or run_mgr is None:
         logger.info("Concurrency sampler not started (no store / run manager — memory backend?)")
         return
@@ -875,11 +876,16 @@ async def _concurrency_sampler_loop(app: FastAPI) -> None:
             try:
                 # Runtime toggle (admin can turn sampling off without a restart).
                 # mtime-cached read, so this is cheap every tick.
-                from deerflow.config.system_settings import get_concurrency_monitor_settings
-
-                if get_concurrency_monitor_settings().enabled:
-                    active = await run_mgr.list_active()
-                    await store.record(len(active))
+                from deerflow.config.system_settings import get_concurrency_monitor_settings
+
+                if get_concurrency_monitor_settings().enabled:
+                    active_count = None
+                    if concurrency_gate is not None:
+                        active_count = await concurrency_gate.count_active()
+                    if active_count is None:
+                        active = await run_mgr.list_active()
+                        active_count = len(active)
+                    await store.record(active_count)
                     ticks += 1
                     if ticks % purge_every_ticks == 0:
                         from datetime import UTC, datetime, timedelta
diff --git a/offline-backend-20260512/backend/app/gateway/deps.py b/offline-backend-20260512/backend/app/gateway/deps.py
index 49db1366..fd158208 100644
--- a/offline-backend-20260512/backend/app/gateway/deps.py
+++ b/offline-backend-20260512/backend/app/gateway/deps.py
@@ -367,8 +367,11 @@ async def langgraph_runtime(app: FastAPI) -> AsyncGenerator[None, None]:
         run_events_config = getattr(config, "run_events", None)
         app.state.run_event_store = make_run_event_store(run_events_config)
 
-        # RunManager with store backing for persistence
-        app.state.run_manager = RunManager(store=app.state.run_store)
+        # RunManager with store backing for persistence
+        app.state.run_manager = RunManager(store=app.state.run_store)
+        from deerflow.runtime import make_redis_concurrency_gate
+
+        app.state.concurrency_gate = make_redis_concurrency_gate(config)
 
         try:
             yield
@@ -382,7 +385,10 @@ async def langgraph_runtime(app: FastAPI) -> AsyncGenerator[None, None]:
 
                 await queue.stop()
                 set_auto_ingest_queue(None)
-            await close_engine()
+            await close_engine()
+            gate = getattr(app.state, "concurrency_gate", None)
+            if gate is not None:
+                await gate.aclose()
 
 
 # ---------------------------------------------------------------------------
@@ -408,7 +414,12 @@ get_run_manager: Callable[[Request], RunManager] = _require("run_manager", "Run
 get_checkpointer: Callable[[Request], Checkpointer] = _require("checkpointer", "Checkpointer")
 get_run_event_store: Callable[[Request], RunEventStore] = _require("run_event_store", "Run event store")
 get_feedback_repo: Callable[[Request], FeedbackRepository] = _require("feedback_repo", "Feedback")
-get_run_store: Callable[[Request], RunStore] = _require("run_store", "Run store")
+get_run_store: Callable[[Request], RunStore] = _require("run_store", "Run store")
+
+
+def get_concurrency_gate(request: Request):
+    """Return the optional cross-worker concurrency gate."""
+    return getattr(request.app.state, "concurrency_gate", None)
 
 
 def get_store(request: Request):
diff --git a/offline-backend-20260512/backend/app/gateway/routers/admin_active_runs.py b/offline-backend-20260512/backend/app/gateway/routers/admin_active_runs.py
index 352f0af9..a602a398 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/admin_active_runs.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/admin_active_runs.py
@@ -22,7 +22,7 @@ from fastapi import APIRouter, Query, Request
 from pydantic import BaseModel
 from sqlalchemy import select
 
-from app.gateway.deps import get_concurrency_sample_store, get_run_manager
+from app.gateway.deps import get_concurrency_gate, get_concurrency_sample_store, get_run_manager
 from app.gateway.routers.admin_users import _require_admin
 from deerflow.config.system_settings import load_system_settings, save_system_settings
 from deerflow.persistence.agents.model import AgentRow
@@ -149,6 +149,15 @@ def _elapsed_seconds(created_at: str, now: datetime) -> int:
     return max(0, int(delta))
 
 
+def _elapsed_seconds_from_epoch(created_at: float, now: datetime) -> int:
+    try:
+        if created_at <= 0:
+            return 0
+        return max(0, int(now.timestamp() - float(created_at)))
+    except Exception:
+        return 0
+
+
 def _kind_for(metadata: dict | None) -> str:
     """Coarse human label for the run's origin, from its thread metadata."""
     meta = metadata or {}
@@ -218,14 +227,46 @@ async def list_active_runs(request: Request) -> ActiveRunsResponse:
     """List every in-flight run with who/what is driving it (admin only)."""
     await _require_admin(request)
     run_mgr = get_run_manager(request)
-    records = await run_mgr.list_active()
+    gate = get_concurrency_gate(request)
+    leases = await gate.list_active() if gate is not None else None
+    records = [] if leases is not None else await run_mgr.list_active()
     now = datetime.now(UTC)
-    maps = await _enrich(records)
+    enrich_source = leases if leases is not None else records
+    maps = await _enrich(enrich_source)
     thread_meta = maps["thread_meta"]
     user_emails = maps["user_emails"]
     agent_names = maps["agent_names"]
 
     items: list[ActiveRunItem] = []
+    if leases is not None:
+        for r in leases:
+            meta_row = thread_meta.get(r.thread_id)
+            user_id = r.user_id or (meta_row.user_id if meta_row else "") or ""
+            metadata = (meta_row.metadata_json if meta_row else None) or {}
+            assistant_id = r.assistant_id or ""
+            if assistant_id in _DEFAULT_ASSISTANT_IDS:
+                agent_name = "榛樿鍔╂墜"
+            else:
+                agent_name = agent_names.get(assistant_id) or assistant_id
+            items.append(
+                ActiveRunItem(
+                    run_id=r.run_id,
+                    thread_id=r.thread_id,
+                    status="running",
+                    created_at=datetime.fromtimestamp(r.created_at, tz=UTC).isoformat() if r.created_at else "",
+                    elapsed_seconds=_elapsed_seconds_from_epoch(r.created_at, now),
+                    user_id=user_id,
+                    user_name=user_emails.get(user_id) or user_id or "鏈煡鐢ㄦ埛",
+                    agent_id=assistant_id,
+                    agent_name=agent_name,
+                    model_name=r.model_name,
+                    thread_title=(meta_row.display_name if meta_row else "") or "",
+                    kind=_kind_for(metadata),
+                    multitask_strategy="reject",
+                )
+            )
+        return ActiveRunsResponse(count=len(items), server_time=now.isoformat(), runs=items)
+
     for r in records:
         meta_row = thread_meta.get(r.thread_id)
         user_id = (meta_row.user_id if meta_row else "") or ""
diff --git a/offline-backend-20260512/backend/app/gateway/services.py b/offline-backend-20260512/backend/app/gateway/services.py
index 090cb654..8331a565 100644
--- a/offline-backend-20260512/backend/app/gateway/services.py
+++ b/offline-backend-20260512/backend/app/gateway/services.py
@@ -12,13 +12,14 @@ import json
 import logging
 import re
 import time
+import uuid
 from collections.abc import Mapping
 from typing import Any
 
 from fastapi import HTTPException, Request
 from langchain_core.messages import HumanMessage
 
-from app.gateway.deps import get_run_context, get_run_manager, get_stream_bridge
+from app.gateway.deps import get_concurrency_gate, get_run_context, get_run_manager, get_stream_bridge
 from app.gateway.utils import sanitize_log_param
 from deerflow.config import get_app_config
 from deerflow.config.agents_config import load_agent_config, validate_agent_id
@@ -41,6 +42,45 @@ from deerflow.runtime.user_context import get_effective_user_id
 logger = logging.getLogger(__name__)
 
 
+async def _run_agent_with_concurrency_lease(
+    *,
+    concurrency_gate: Any | None,
+    lease_acquired: bool,
+    lease_metadata: dict[str, str],
+    heartbeat_interval_seconds: int,
+    run_agent_kwargs: dict[str, Any],
+) -> None:
+    """Run the agent while refreshing and finally releasing an optional lease."""
+
+    heartbeat_task: asyncio.Task | None = None
+
+    async def _heartbeat_loop() -> None:
+        if concurrency_gate is None or not lease_acquired:
+            return
+        interval = max(5, int(heartbeat_interval_seconds or 20))
+        while True:
+            await asyncio.sleep(interval)
+            await concurrency_gate.heartbeat(**lease_metadata)
+
+    if concurrency_gate is not None and lease_acquired:
+        heartbeat_task = asyncio.create_task(
+            _heartbeat_loop(),
+            name=f"concurrency-heartbeat:{lease_metadata.get('run_id')}",
+        )
+
+    try:
+        await run_agent(**run_agent_kwargs)
+    finally:
+        if heartbeat_task is not None:
+            heartbeat_task.cancel()
+            try:
+                await heartbeat_task
+            except asyncio.CancelledError:
+                pass
+        if concurrency_gate is not None and lease_acquired:
+            await concurrency_gate.release(**lease_metadata)
+
+
 # ---------------------------------------------------------------------------
 # SSE formatting
 # ---------------------------------------------------------------------------
@@ -737,6 +777,7 @@ async def start_run(
     bridge = get_stream_bridge(request)
     run_mgr = get_run_manager(request)
     run_ctx = get_run_context(request)
+    concurrency_gate = get_concurrency_gate(request)
 
     disconnect = DisconnectMode.cancel if body.on_disconnect == "cancel" else DisconnectMode.continue_
 
@@ -754,11 +795,46 @@ async def start_run(
     effective_user_id = get_effective_user_id()
     model_concurrency_limit = _get_model_concurrency_limit(effective_model_name)
     user_model_concurrency_limit = _get_user_model_concurrency_limit(effective_model_name)
+    run_id = str(uuid.uuid4())
 
     run_metadata = dict(body.metadata or {})
     run_metadata["model_name"] = effective_model_name
     run_metadata["user_id"] = effective_user_id
 
+    redis_lease_acquired = False
+    redis_gate_degraded = True
+    lease_metadata = {
+        "run_id": run_id,
+        "thread_id": thread_id,
+        "model_name": effective_model_name,
+        "user_id": effective_user_id,
+    }
+    if concurrency_gate is not None:
+        acquire = await concurrency_gate.try_acquire(
+            run_id=run_id,
+            thread_id=thread_id,
+            model_name=effective_model_name,
+            user_id=effective_user_id,
+            assistant_id=body.assistant_id or "",
+            model_limit=model_concurrency_limit,
+            user_model_limit=user_model_concurrency_limit,
+            reject_thread=body.multitask_strategy == "reject",
+        )
+        redis_lease_acquired = acquire.acquired
+        redis_gate_degraded = acquire.degraded
+        if acquire.rejected:
+            if acquire.reason == "thread":
+                raise HTTPException(
+                    status_code=409,
+                    detail="Current thread already has an active run. Please wait or stop it before sending another message.",
+                )
+            if acquire.reason == "user_model":
+                raise HTTPException(
+                    status_code=429,
+                    detail="This account has reached the concurrency limit for the selected model. Please try again later.",
+                )
+            raise HTTPException(status_code=429, detail="The selected model is busy. Please try again later.")
+
     try:
         record = await run_mgr.create_or_reject(
             thread_id,
@@ -767,24 +843,35 @@ async def start_run(
             metadata=run_metadata,
             kwargs={"input": body.input, "config": body.config},
             multitask_strategy=body.multitask_strategy,
-            model_concurrency_limit=(effective_model_name, model_concurrency_limit) if model_concurrency_limit is not None else None,
-            user_model_concurrency_limit=(effective_model_name, effective_user_id, user_model_concurrency_limit) if user_model_concurrency_limit is not None else None,
+            model_concurrency_limit=(effective_model_name, model_concurrency_limit) if redis_gate_degraded and model_concurrency_limit is not None else None,
+            user_model_concurrency_limit=(effective_model_name, effective_user_id, user_model_concurrency_limit) if redis_gate_degraded and user_model_concurrency_limit is not None else None,
+            run_id=run_id,
         )
     except ModelConcurrencyLimitError as exc:
+        if concurrency_gate is not None and redis_lease_acquired:
+            await concurrency_gate.release(**lease_metadata)
         if getattr(exc, "scope", "") == "user_model":
-            detail = "当前账号对该模型的并发调用已达到上限，请等待已有回答结束后再试。"
+            detail = "This account has reached the concurrency limit for the selected model. Please try again later."
         else:
-            detail = "当前使用人数较多，请稍等一下再试。"
+            detail = "The selected model is busy. Please try again later."
         raise HTTPException(status_code=429, detail=detail) from exc
     except ConflictError as exc:
+        if concurrency_gate is not None and redis_lease_acquired:
+            await concurrency_gate.release(**lease_metadata)
         # 同一会话上一轮回答还没结束，又提交了新问题 → 拒绝。给用户中文提示，
         # 而不是把内部英文「Thread ... already has an active run」直接抛到前端。
         raise HTTPException(
             status_code=409,
-            detail="当前对话还在生成回答，请等待本轮回复结束，或先点「停止」再发送新的问题。",
+            detail="Current thread already has an active run. Please wait or stop it before sending another message.",
         ) from exc
     except UnsupportedStrategyError as exc:
+        if concurrency_gate is not None and redis_lease_acquired:
+            await concurrency_gate.release(**lease_metadata)
         raise HTTPException(status_code=501, detail=str(exc)) from exc
+    except Exception:
+        if concurrency_gate is not None and redis_lease_acquired:
+            await concurrency_gate.release(**lease_metadata)
+        raise
     t_post_create = time.perf_counter()
 
     # Upsert thread metadata so the thread appears in /threads/search,
@@ -822,7 +909,12 @@ async def start_run(
         logger.warning("Failed to upsert thread_meta for %s (non-fatal)", sanitize_log_param(thread_id))
     t_post_thread_upsert = time.perf_counter()
 
-    agent_factory = resolve_agent_factory(body.assistant_id)
+    try:
+        agent_factory = resolve_agent_factory(body.assistant_id)
+    except Exception:
+        if concurrency_gate is not None and redis_lease_acquired:
+            await concurrency_gate.release(**lease_metadata)
+        raise
     graph_input = normalize_input(body.input)
 
     # Admin-configurable prompt prefix: when the caller opted in via
@@ -860,24 +952,38 @@ async def start_run(
     stream_modes = normalize_stream_modes(body.stream_mode)
 
     t_pre_task = time.perf_counter()
-    task = asyncio.create_task(
-        run_agent(
-            bridge,
-            run_mgr,
-            record,
-            ctx=run_ctx,
-            agent_factory=agent_factory,
-            graph_input=graph_input,
-            config=config,
-            stream_modes=stream_modes,
-            stream_subgraphs=body.stream_subgraphs,
-            interrupt_before=body.interrupt_before,
-            interrupt_after=body.interrupt_after,
-            # 透传 SDK ``Command``：resume / update / goto。Pregel 会据此从最近
-            # checkpoint 处恢复，而不是把 graph_input 当新一轮输入跑。
-            command=getattr(body, "command", None),
+    heartbeat_interval_seconds = 20
+    if concurrency_gate is not None:
+        heartbeat_interval_seconds = int(
+            getattr(getattr(concurrency_gate, "config", None), "heartbeat_interval_seconds", 20) or 20
         )
-    )
+    try:
+        task = asyncio.create_task(
+            _run_agent_with_concurrency_lease(
+                concurrency_gate=concurrency_gate,
+                lease_acquired=redis_lease_acquired,
+                lease_metadata=lease_metadata,
+                heartbeat_interval_seconds=heartbeat_interval_seconds,
+                run_agent_kwargs={
+                    "bridge": bridge,
+                    "run_manager": run_mgr,
+                    "record": record,
+                    "ctx": run_ctx,
+                    "agent_factory": agent_factory,
+                    "graph_input": graph_input,
+                    "config": config,
+                    "stream_modes": stream_modes,
+                    "stream_subgraphs": body.stream_subgraphs,
+                    "interrupt_before": body.interrupt_before,
+                    "interrupt_after": body.interrupt_after,
+                    "command": getattr(body, "command", None),
+                },
+            )
+        )
+    except Exception:
+        if concurrency_gate is not None and redis_lease_acquired:
+            await concurrency_gate.release(**lease_metadata)
+        raise
     record.task = task
 
     # Title sync is handled by worker.py's finally block which reads the
diff --git a/offline-backend-20260512/backend/config.yaml b/offline-backend-20260512/backend/config.yaml
index 5d9d9a69..1b01c238 100644
--- a/offline-backend-20260512/backend/config.yaml
+++ b/offline-backend-20260512/backend/config.yaml
@@ -9,6 +9,26 @@ log_level: info
 gateway:
   port: 8001
 
+# Optional Redis-backed global concurrency coordination.
+# Leave disabled to keep the existing in-process behavior.
+# When enabled, Redis operations are short-timeout and fail-open by default:
+# if Redis is slow or unavailable, user Q&A continues with local-only limits.
+concurrency:
+  redis:
+    enabled: false
+    url: redis://127.0.0.1:6379/0
+    password: null
+    namespace: deerflow:concurrency
+    socket_connect_timeout_ms: 80
+    socket_timeout_ms: 80
+    acquire_timeout_ms: 120
+    operation_timeout_ms: 80
+    circuit_breaker_failures: 3
+    circuit_breaker_cooldown_seconds: 30
+    lease_ttl_seconds: 90
+    heartbeat_interval_seconds: 20
+    fail_open: true
+
 models:
   - name: deepseek-chat
     display_name: DeepSeek Chat
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/config/app_config.py b/offline-backend-20260512/backend/packages/harness/deerflow/config/app_config.py
index 58d09b4b..8deaeb3c 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/config/app_config.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/config/app_config.py
@@ -11,8 +11,9 @@ from pydantic import BaseModel, ConfigDict, Field
 from deerflow.config.acp_config import ACPAgentConfig, load_acp_config_from_dict
 from deerflow.config.agents_api_config import AgentsApiConfig, load_agents_api_config_from_dict
 from deerflow.config.browser_context_config import BrowserContextConfig
-from deerflow.config.checkpointer_config import CheckpointerConfig, load_checkpointer_config_from_dict
-from deerflow.config.database_config import DatabaseConfig
+from deerflow.config.checkpointer_config import CheckpointerConfig, load_checkpointer_config_from_dict
+from deerflow.config.concurrency_config import ConcurrencyConfig
+from deerflow.config.database_config import DatabaseConfig
 from deerflow.config.extensions_config import ExtensionsConfig
 from deerflow.config.gateway_config import GatewayConfig
 from deerflow.config.guardrails_config import GuardrailsConfig, load_guardrails_config_from_dict
@@ -137,8 +138,9 @@ class AppConfig(BaseModel):
     gateway: GatewayConfig = Field(default_factory=GatewayConfig, description="Gateway FastAPI listening port (used by intra-process loopback)")
     run_events: RunEventsConfig = Field(default_factory=RunEventsConfig, description="Run event storage configuration")
     checkpointer: CheckpointerConfig | None = Field(default=None, description="Checkpointer configuration")
-    stream_bridge: StreamBridgeConfig | None = Field(default=None, description="Stream bridge configuration")
-    browser_context: BrowserContextConfig = Field(default_factory=BrowserContextConfig, description="Optional browser page capture integration")
+    stream_bridge: StreamBridgeConfig | None = Field(default=None, description="Stream bridge configuration")
+    concurrency: ConcurrencyConfig = Field(default_factory=ConcurrencyConfig, description="Optional cross-worker concurrency coordination")
+    browser_context: BrowserContextConfig = Field(default_factory=BrowserContextConfig, description="Optional browser page capture integration")
     knowledge: KnowledgeConfig = Field(default_factory=KnowledgeConfig, description="Product knowledge base (obsidian-wiki compatible vault) configuration")
     knowledge_ingest: KnowledgeIngestConfig = Field(default_factory=KnowledgeIngestConfig, description="External knowledge-ingest endpoint configuration")
     task_deeplink: TaskDeeplinkConfig = Field(default_factory=TaskDeeplinkConfig, description="任务深链(rwfx)目的树闸门与跳转地址配置(前端经 /api/public/config/task-deeplink 读取)")
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/config/concurrency_config.py b/offline-backend-20260512/backend/packages/harness/deerflow/config/concurrency_config.py
new file mode 100644
index 00000000..d741fda1
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/config/concurrency_config.py
@@ -0,0 +1,35 @@
+"""Runtime concurrency coordination configuration."""
+
+from __future__ import annotations
+
+from pydantic import BaseModel, Field
+
+
+class RedisConcurrencyConfig(BaseModel):
+    """Optional Redis-backed cross-worker concurrency coordination."""
+
+    enabled: bool = Field(default=False, description="Enable Redis-backed global run concurrency coordination.")
+    url: str | None = Field(default=None, description="Redis URL, for example redis://127.0.0.1:6379/0.")
+    password: str | None = Field(
+        default=None,
+        description="Optional Redis password. You may also include credentials directly in url.",
+    )
+    namespace: str = Field(default="deerflow:concurrency", description="Redis key prefix for concurrency leases.")
+    socket_connect_timeout_ms: int = Field(default=80, ge=1, le=10_000)
+    socket_timeout_ms: int = Field(default=80, ge=1, le=10_000)
+    acquire_timeout_ms: int = Field(default=120, ge=1, le=10_000)
+    operation_timeout_ms: int = Field(default=80, ge=1, le=10_000)
+    circuit_breaker_failures: int = Field(default=3, ge=1, le=100)
+    circuit_breaker_cooldown_seconds: int = Field(default=30, ge=1, le=3600)
+    lease_ttl_seconds: int = Field(default=90, ge=10, le=3600)
+    heartbeat_interval_seconds: int = Field(default=20, ge=5, le=600)
+    fail_open: bool = Field(
+        default=True,
+        description="Allow runs to proceed with local-only limits when Redis is unavailable or slow.",
+    )
+
+
+class ConcurrencyConfig(BaseModel):
+    """Top-level concurrency coordination config."""
+
+    redis: RedisConcurrencyConfig = Field(default_factory=RedisConcurrencyConfig)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/__init__.py b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/__init__.py
index 63ea0860..d0bfc8e4 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/__init__.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/__init__.py
@@ -7,14 +7,19 @@ directly from ``deerflow.runtime``.
 
 from .checkpointer import checkpointer_context, get_checkpointer, make_checkpointer, reset_checkpointer
 from .runs import (
+    AcquireDecision,
+    AcquireResult,
+    ActiveLease,
     ConflictError,
     DisconnectMode,
     ModelConcurrencyLimitError,
+    RedisConcurrencyGate,
     RunContext,
     RunManager,
     RunRecord,
     RunStatus,
     UnsupportedStrategyError,
+    make_redis_concurrency_gate,
     run_agent,
 )
 from .serialization import serialize, serialize_channel_values, serialize_lc_object, serialize_messages_tuple
@@ -28,14 +33,19 @@ __all__ = [
     "make_checkpointer",
     "reset_checkpointer",
     # runs
+    "AcquireDecision",
+    "AcquireResult",
+    "ActiveLease",
     "ConflictError",
     "DisconnectMode",
     "ModelConcurrencyLimitError",
+    "RedisConcurrencyGate",
     "RunContext",
     "RunManager",
     "RunRecord",
     "RunStatus",
     "UnsupportedStrategyError",
+    "make_redis_concurrency_gate",
     "run_agent",
     # serialization
     "serialize",
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/__init__.py b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/__init__.py
index 318b58c5..d6382f97 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/__init__.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/__init__.py
@@ -2,16 +2,22 @@
 
 from .manager import ConflictError, ModelConcurrencyLimitError, RunManager, RunRecord, UnsupportedStrategyError
 from .schemas import DisconnectMode, RunStatus
+from .concurrency_gate import AcquireDecision, AcquireResult, ActiveLease, RedisConcurrencyGate, make_redis_concurrency_gate
 from .worker import RunContext, run_agent
 
 __all__ = [
+    "AcquireDecision",
+    "AcquireResult",
+    "ActiveLease",
     "ConflictError",
     "DisconnectMode",
     "ModelConcurrencyLimitError",
+    "RedisConcurrencyGate",
     "RunContext",
     "RunManager",
     "RunRecord",
     "RunStatus",
     "UnsupportedStrategyError",
+    "make_redis_concurrency_gate",
     "run_agent",
 ]
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/concurrency_gate.py b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/concurrency_gate.py
new file mode 100644
index 00000000..b2bd9e2f
--- /dev/null
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/concurrency_gate.py
@@ -0,0 +1,369 @@
+"""Optional cross-worker run concurrency gates.
+
+The Redis gate is deliberately fail-open by default: Redis coordinates global
+limits when it is fast and healthy, but it must not block the chat path when it
+is slow or unavailable.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import logging
+import time
+from dataclasses import dataclass
+from enum import Enum
+from typing import Any
+
+from deerflow.config.concurrency_config import RedisConcurrencyConfig
+
+logger = logging.getLogger(__name__)
+
+
+class AcquireDecision(str, Enum):
+    ACQUIRED = "acquired"
+    REJECTED = "rejected"
+    DEGRADED = "degraded"
+
+
+@dataclass(frozen=True)
+class AcquireResult:
+    decision: AcquireDecision
+    reason: str = ""
+    source: str = "redis"
+
+    @property
+    def acquired(self) -> bool:
+        return self.decision == AcquireDecision.ACQUIRED
+
+    @property
+    def rejected(self) -> bool:
+        return self.decision == AcquireDecision.REJECTED
+
+    @property
+    def degraded(self) -> bool:
+        return self.decision == AcquireDecision.DEGRADED
+
+
+@dataclass(frozen=True)
+class ActiveLease:
+    run_id: str
+    thread_id: str = ""
+    model_name: str = ""
+    user_id: str = ""
+    assistant_id: str = ""
+    created_at: float = 0.0
+    expires_at: float = 0.0
+
+
+class CircuitBreaker:
+    """Tiny in-process circuit breaker for Redis failures."""
+
+    def __init__(self, *, failure_threshold: int, cooldown_seconds: int) -> None:
+        self._failure_threshold = max(1, int(failure_threshold))
+        self._cooldown_seconds = max(1, int(cooldown_seconds))
+        self._failures = 0
+        self._opened_until = 0.0
+
+    def allow(self) -> bool:
+        return time.monotonic() >= self._opened_until
+
+    def record_success(self) -> None:
+        self._failures = 0
+        self._opened_until = 0.0
+
+    def record_failure(self) -> None:
+        self._failures += 1
+        if self._failures >= self._failure_threshold:
+            self._opened_until = time.monotonic() + self._cooldown_seconds
+
+    @property
+    def open(self) -> bool:
+        return not self.allow()
+
+
+class RedisConcurrencyGate:
+    """Redis lease-based global concurrency limiter."""
+
+    _ACQUIRE_SCRIPT = """
+local now = tonumber(ARGV[1])
+local expires_at = tonumber(ARGV[2])
+local run_id = ARGV[3]
+local model_limit = tonumber(ARGV[4])
+local user_model_limit = tonumber(ARGV[5])
+local reject_thread = tonumber(ARGV[6])
+local payload = ARGV[7]
+local ttl_ms = tonumber(ARGV[8])
+
+for i = 1, 4 do
+  redis.call('ZREMRANGEBYSCORE', KEYS[i], '-inf', now)
+end
+
+if reject_thread == 1 and redis.call('ZCARD', KEYS[4]) > 0 then
+  return {'rejected', 'thread'}
+end
+if model_limit > 0 and redis.call('ZCARD', KEYS[2]) >= model_limit then
+  return {'rejected', 'model'}
+end
+if user_model_limit > 0 and redis.call('ZCARD', KEYS[3]) >= user_model_limit then
+  return {'rejected', 'user_model'}
+end
+
+redis.call('ZADD', KEYS[1], expires_at, run_id)
+redis.call('ZADD', KEYS[2], expires_at, run_id)
+redis.call('ZADD', KEYS[3], expires_at, run_id)
+redis.call('ZADD', KEYS[4], expires_at, run_id)
+redis.call('SET', KEYS[5], payload, 'PX', ttl_ms)
+return {'acquired', ''}
+"""
+
+    def __init__(self, config: RedisConcurrencyConfig) -> None:
+        self.config = config
+        self._client: Any | None = None
+        self._breaker = CircuitBreaker(
+            failure_threshold=config.circuit_breaker_failures,
+            cooldown_seconds=config.circuit_breaker_cooldown_seconds,
+        )
+
+    @property
+    def enabled(self) -> bool:
+        return bool(self.config.enabled and self.config.url)
+
+    def health(self) -> dict[str, Any]:
+        return {
+            "enabled": self.enabled,
+            "source": "redis",
+            "circuit_open": self._breaker.open,
+            "fail_open": self.config.fail_open,
+        }
+
+    async def aclose(self) -> None:
+        client = self._client
+        self._client = None
+        if client is not None:
+            try:
+                await client.aclose()
+            except Exception:
+                logger.debug("Failed to close Redis concurrency client", exc_info=True)
+
+    def _ns(self) -> str:
+        return self.config.namespace.strip().rstrip(":") or "deerflow:concurrency"
+
+    def _keys(self, *, run_id: str, model_name: str, user_id: str, thread_id: str) -> list[str]:
+        ns = self._ns()
+        return [
+            f"{ns}:all",
+            f"{ns}:model:{model_name}",
+            f"{ns}:user_model:{model_name}:{user_id}",
+            f"{ns}:thread:{thread_id}",
+            f"{ns}:run:{run_id}",
+        ]
+
+    async def _redis(self) -> Any:
+        if self._client is not None:
+            return self._client
+        try:
+            from redis.asyncio import Redis
+        except Exception as exc:  # pragma: no cover - exercised when dependency is absent
+            raise RuntimeError("redis package is not installed") from exc
+
+        timeout = max(0.001, self.config.socket_timeout_ms / 1000)
+        connect_timeout = max(0.001, self.config.socket_connect_timeout_ms / 1000)
+        self._client = Redis.from_url(
+            self.config.url,
+            password=self.config.password,
+            socket_timeout=timeout,
+            socket_connect_timeout=connect_timeout,
+            decode_responses=True,
+        )
+        return self._client
+
+    async def _with_timeout(self, coro, timeout_ms: int) -> Any:
+        return await asyncio.wait_for(coro, timeout=max(0.001, timeout_ms / 1000))
+
+    def _degraded(self, reason: str) -> AcquireResult:
+        return AcquireResult(AcquireDecision.DEGRADED, reason=reason)
+
+    def _on_failure(self, action: str, exc: BaseException) -> None:
+        self._breaker.record_failure()
+        logger.warning(
+            "Redis concurrency %s failed; fail_open=%s circuit_open=%s: %s",
+            action,
+            self.config.fail_open,
+            self._breaker.open,
+            exc,
+        )
+
+    async def try_acquire(
+        self,
+        *,
+        run_id: str,
+        thread_id: str,
+        model_name: str,
+        user_id: str,
+        assistant_id: str = "",
+        model_limit: int | None = None,
+        user_model_limit: int | None = None,
+        reject_thread: bool = True,
+    ) -> AcquireResult:
+        if not self.enabled:
+            return self._degraded("disabled")
+        if not self._breaker.allow():
+            return self._degraded("circuit_open")
+
+        now = time.time()
+        expires_at = now + self.config.lease_ttl_seconds
+        payload = {
+            "run_id": run_id,
+            "thread_id": thread_id,
+            "model_name": model_name,
+            "user_id": user_id,
+            "assistant_id": assistant_id,
+            "created_at": now,
+            "expires_at": expires_at,
+        }
+        import json
+
+        keys = self._keys(run_id=run_id, model_name=model_name, user_id=user_id, thread_id=thread_id)
+        ttl_ms = int(self.config.lease_ttl_seconds * 1000)
+        try:
+            client = await self._redis()
+            raw = await self._with_timeout(
+                client.eval(
+                    self._ACQUIRE_SCRIPT,
+                    len(keys),
+                    *keys,
+                    str(now),
+                    str(expires_at),
+                    run_id,
+                    str(max(0, int(model_limit or 0))),
+                    str(max(0, int(user_model_limit or 0))),
+                    "1" if reject_thread else "0",
+                    json.dumps(payload, ensure_ascii=False),
+                    str(ttl_ms),
+                ),
+                self.config.acquire_timeout_ms,
+            )
+            self._breaker.record_success()
+        except Exception as exc:
+            self._on_failure("acquire", exc)
+            if self.config.fail_open:
+                return self._degraded(type(exc).__name__)
+            raise
+
+        status = raw[0] if isinstance(raw, list) and raw else ""
+        reason = raw[1] if isinstance(raw, list) and len(raw) > 1 else ""
+        if status == "acquired":
+            return AcquireResult(AcquireDecision.ACQUIRED)
+        return AcquireResult(AcquireDecision.REJECTED, reason=str(reason or "limit"))
+
+    async def heartbeat(
+        self,
+        *,
+        run_id: str,
+        thread_id: str,
+        model_name: str,
+        user_id: str,
+    ) -> None:
+        if not self.enabled or not self._breaker.allow():
+            return
+        now = time.time()
+        expires_at = now + self.config.lease_ttl_seconds
+        keys = self._keys(run_id=run_id, model_name=model_name, user_id=user_id, thread_id=thread_id)
+        try:
+            client = await self._redis()
+            pipe = client.pipeline(transaction=False)
+            for key in keys[:4]:
+                pipe.zadd(key, {run_id: expires_at})
+            pipe.expire(keys[4], self.config.lease_ttl_seconds)
+            await self._with_timeout(pipe.execute(), self.config.operation_timeout_ms)
+            self._breaker.record_success()
+        except Exception as exc:
+            self._on_failure("heartbeat", exc)
+
+    async def release(
+        self,
+        *,
+        run_id: str,
+        thread_id: str,
+        model_name: str,
+        user_id: str,
+    ) -> None:
+        if not self.enabled or not self._breaker.allow():
+            return
+        keys = self._keys(run_id=run_id, model_name=model_name, user_id=user_id, thread_id=thread_id)
+        try:
+            client = await self._redis()
+            pipe = client.pipeline(transaction=False)
+            for key in keys[:4]:
+                pipe.zrem(key, run_id)
+            pipe.delete(keys[4])
+            await self._with_timeout(pipe.execute(), self.config.operation_timeout_ms)
+            self._breaker.record_success()
+        except Exception as exc:
+            self._on_failure("release", exc)
+
+    async def count_active(self) -> int | None:
+        if not self.enabled or not self._breaker.allow():
+            return None
+        key = f"{self._ns()}:all"
+        now = time.time()
+        try:
+            client = await self._redis()
+            pipe = client.pipeline(transaction=False)
+            pipe.zremrangebyscore(key, "-inf", now)
+            pipe.zcard(key)
+            result = await self._with_timeout(pipe.execute(), self.config.operation_timeout_ms)
+            self._breaker.record_success()
+            return int(result[-1] or 0)
+        except Exception as exc:
+            self._on_failure("count", exc)
+            return None
+
+    async def list_active(self, *, limit: int = 1000) -> list[ActiveLease] | None:
+        if not self.enabled or not self._breaker.allow():
+            return None
+        import json
+
+        all_key = f"{self._ns()}:all"
+        now = time.time()
+        try:
+            client = await self._redis()
+            await self._with_timeout(client.zremrangebyscore(all_key, "-inf", now), self.config.operation_timeout_ms)
+            run_ids = await self._with_timeout(client.zrange(all_key, 0, max(0, limit - 1)), self.config.operation_timeout_ms)
+            if not run_ids:
+                self._breaker.record_success()
+                return []
+            keys = [f"{self._ns()}:run:{run_id}" for run_id in run_ids]
+            payloads = await self._with_timeout(client.mget(keys), self.config.operation_timeout_ms)
+            leases: list[ActiveLease] = []
+            for run_id, payload in zip(run_ids, payloads, strict=False):
+                if not payload:
+                    continue
+                try:
+                    data = json.loads(payload)
+                except Exception:
+                    continue
+                leases.append(
+                    ActiveLease(
+                        run_id=str(data.get("run_id") or run_id),
+                        thread_id=str(data.get("thread_id") or ""),
+                        model_name=str(data.get("model_name") or ""),
+                        user_id=str(data.get("user_id") or ""),
+                        assistant_id=str(data.get("assistant_id") or ""),
+                        created_at=float(data.get("created_at") or 0),
+                        expires_at=float(data.get("expires_at") or 0),
+                    )
+                )
+            self._breaker.record_success()
+            return leases
+        except Exception as exc:
+            self._on_failure("list", exc)
+            return None
+
+
+def make_redis_concurrency_gate(config: Any) -> RedisConcurrencyGate | None:
+    concurrency = getattr(config, "concurrency", None)
+    redis_config = getattr(concurrency, "redis", None)
+    if redis_config is None or not getattr(redis_config, "enabled", False) or not getattr(redis_config, "url", None):
+        return None
+    return RedisConcurrencyGate(redis_config)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/manager.py b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/manager.py
index d6842657..edef5296 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/manager.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/manager.py
@@ -219,6 +219,7 @@ class RunManager:
         multitask_strategy: str = "reject",
         model_concurrency_limit: tuple[str, int] | None = None,
         user_model_concurrency_limit: tuple[str, str, int] | None = None,
+        run_id: str | None = None,
     ) -> RunRecord:
         """Atomically check for inflight runs and create a new one.
 
@@ -229,7 +230,7 @@ class RunManager:
         This method holds the lock across both the check and the insert,
         eliminating the TOCTOU race in separate ``has_inflight`` + ``create``.
         """
-        run_id = str(uuid.uuid4())
+        run_id = run_id or str(uuid.uuid4())
         now = _now_iso()
 
         _supported_strategies = ("reject", "interrupt", "rollback")
diff --git a/offline-backend-20260512/backend/packages/harness/pyproject.toml b/offline-backend-20260512/backend/packages/harness/pyproject.toml
index b8b5ff94..2c47c1a5 100644
--- a/offline-backend-20260512/backend/packages/harness/pyproject.toml
+++ b/offline-backend-20260512/backend/packages/harness/pyproject.toml
@@ -23,9 +23,10 @@ dependencies = [
     "markdownify>=1.2.2",
     "markitdown[all,xlsx]>=0.0.1a2",
     "pydantic>=2.12.5",
-    "pyyaml>=6.0.3",
-    "readabilipy>=0.3.0",
-    "tavily-python>=0.7.17",
+    "pyyaml>=6.0.3",
+    "readabilipy>=0.3.0",
+    "redis>=5.0.0",
+    "tavily-python>=0.7.17",
     "firecrawl-py>=1.15.0",
     "tiktoken>=0.8.0",
     "ddgs>=9.10.0",

From c23350f17b4516a85b981980f2c1e4ad3a28b110 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=90=89=E9=91=AB=E5=8D=87?= 
Date: Tue, 30 Jun 2026 14:56:29 +0800
Subject: [PATCH 11/16] fix: isolate user concurrency from internal agent runs

---
 .../gateway/roundtable_inprocess_gateway.py   |  9 ++-
 .../app/gateway/routers/multi_agent.py        |  8 ++-
 .../backend/app/gateway/services.py           | 70 ++++++++++++++++++-
 offline-backend-20260512/backend/config.yaml  |  5 +-
 .../deerflow/config/concurrency_config.py     |  5 +-
 .../deerflow/runtime/runs/concurrency_gate.py | 18 +++--
 .../harness/deerflow/runtime/runs/manager.py  |  1 +
 .../backend/tests/test_model_concurrency.py   | 56 +++++++++++++++
 8 files changed, 162 insertions(+), 10 deletions(-)

diff --git a/offline-backend-20260512/backend/app/gateway/roundtable_inprocess_gateway.py b/offline-backend-20260512/backend/app/gateway/roundtable_inprocess_gateway.py
index fadc3093..c34a7867 100644
--- a/offline-backend-20260512/backend/app/gateway/roundtable_inprocess_gateway.py
+++ b/offline-backend-20260512/backend/app/gateway/roundtable_inprocess_gateway.py
@@ -32,6 +32,7 @@ from app.gateway.roundtable_model_fallback import (
 )
 from app.gateway.roundtable_run_policy import roundtable_run_policy
 from app.gateway.roundtable_seat_skills import append_seat_skill_directive
+from app.gateway.services import RUN_SCOPE_BACKGROUND_CHILD
 from app.gateway.routers._roundtable_seed import (
     COORDINATOR_AGENT_ID,
     DASHBOARD_AGENT_ID,
@@ -346,6 +347,7 @@ async def _run_agent_turn(
     body = RunCreateRequest(
         assistant_id="lead_agent",
         input={"messages": [{"type": "human", "content": [{"type": "text", "text": message}], "additional_kwargs": {}}]},
+        metadata={"run_scope": RUN_SCOPE_BACKGROUND_CHILD},
         config={"recursion_limit": 1000},
         context={
             "agent_name": agent_name,
@@ -374,7 +376,12 @@ async def _run_agent_turn(
     # 落 checkpoint）。额度满时本轮在此 await 排队，等同用户某个在跑轮次释放额度后再开。
     async with _user_background_semaphore(user_id):
         try:
-            record = await start_run(body, thread_id, request)  # type: ignore[arg-type]
+            record = await start_run(  # type: ignore[arg-type]
+                body,
+                thread_id,
+                request,
+                run_scope=RUN_SCOPE_BACKGROUND_CHILD,
+            )
         except Exception as exc:  # noqa: BLE001 — start_run 报错（409 抢占 / DB 抖动等）也要落诊断
             if diag is not None:
                 await diag.record(
diff --git a/offline-backend-20260512/backend/app/gateway/routers/multi_agent.py b/offline-backend-20260512/backend/app/gateway/routers/multi_agent.py
index 83046e24..c3bcfe0a 100644
--- a/offline-backend-20260512/backend/app/gateway/routers/multi_agent.py
+++ b/offline-backend-20260512/backend/app/gateway/routers/multi_agent.py
@@ -65,6 +65,7 @@ from langgraph.checkpoint.base import empty_checkpoint
 from app.gateway.roundtable_diag import record_foreground_diag
 from app.gateway.roundtable_model_fallback import fallback_model_chain, is_llm_error_message, reason_text
 from app.gateway.roundtable_run_policy import merge_skill_stop_names, roundtable_run_policy
+from app.gateway.services import RUN_SCOPE_ORCHESTRATION_CHILD, internal_run_scope_headers
 from deerflow.agents.roundtable_orchestrator.extraction_spec import build_summary_directive
 
 from app.gateway.roundtable_seat_skills import append_seat_skill_directive
@@ -325,7 +326,11 @@ def _auth_headers(request: Request) -> dict[str, str]:
     token = request.headers.get("authorization")
     if not token:
         raise HTTPException(status_code=401, detail="Missing Authorization header")
-    return {"Authorization": token, "Content-Type": "application/json"}
+    return {
+        "Authorization": token,
+        "Content-Type": "application/json",
+        **internal_run_scope_headers(),
+    }
 
 
 async def _create_thread(
@@ -694,6 +699,7 @@ def _run_payload(
                 }
             ]
         },
+        "metadata": {"run_scope": RUN_SCOPE_ORCHESTRATION_CHILD},
         "config": {"recursion_limit": 1000},
         "context": {
             "agent_name": agent_name,
diff --git a/offline-backend-20260512/backend/app/gateway/services.py b/offline-backend-20260512/backend/app/gateway/services.py
index 8331a565..22cdd9e9 100644
--- a/offline-backend-20260512/backend/app/gateway/services.py
+++ b/offline-backend-20260512/backend/app/gateway/services.py
@@ -10,6 +10,7 @@ from __future__ import annotations
 import asyncio
 import json
 import logging
+import os
 import re
 import time
 import uuid
@@ -41,12 +42,68 @@ from deerflow.runtime.user_context import get_effective_user_id
 
 logger = logging.getLogger(__name__)
 
+RUN_SCOPE_EXTERNAL_ROOT = "external_root"
+RUN_SCOPE_ORCHESTRATION_CHILD = "orchestration_child"
+RUN_SCOPE_BACKGROUND_CHILD = "background_child"
+_INTERNAL_RUN_SCOPES = {RUN_SCOPE_ORCHESTRATION_CHILD, RUN_SCOPE_BACKGROUND_CHILD}
+_ALL_RUN_SCOPES = {RUN_SCOPE_EXTERNAL_ROOT, *_INTERNAL_RUN_SCOPES}
+_INTERNAL_RUN_SCOPE_HEADER = "x-deerflow-internal-run-scope-token"
+_INTERNAL_RUN_SCOPE_TOKEN = (
+    os.getenv("DEERFLOW_INTERNAL_RUN_SCOPE_TOKEN", "").strip()
+    or uuid.uuid4().hex
+)
+
+
+def internal_run_scope_headers() -> dict[str, str]:
+    """Headers used by same-process loopback calls to mark internal child runs."""
+
+    return {_INTERNAL_RUN_SCOPE_HEADER: _INTERNAL_RUN_SCOPE_TOKEN}
+
+
+def _metadata_run_scope(body: Any) -> str | None:
+    metadata = getattr(body, "metadata", None) or {}
+    if not isinstance(metadata, dict):
+        return None
+    raw = str(metadata.get("run_scope") or "").strip()
+    return raw or None
+
+
+def _request_is_loopback(request: Request) -> bool:
+    client = getattr(request, "client", None)
+    host = str(getattr(client, "host", "") or "").strip()
+    return host in {"127.0.0.1", "::1", "localhost"}
+
+
+def _has_internal_run_scope_header(request: Request) -> bool:
+    headers = getattr(request, "headers", None)
+    token = headers.get(_INTERNAL_RUN_SCOPE_HEADER) if headers is not None else None
+    return str(token or "") == _INTERNAL_RUN_SCOPE_TOKEN
+
+
+def _resolve_run_scope(body: Any, request: Request, run_scope: str | None) -> str:
+    """Return the trusted run scope used for concurrency accounting."""
+
+    if run_scope:
+        scope = str(run_scope).strip()
+        return scope if scope in _ALL_RUN_SCOPES else RUN_SCOPE_EXTERNAL_ROOT
+
+    scope = _metadata_run_scope(body)
+    if (
+        scope in _INTERNAL_RUN_SCOPES
+        and _request_is_loopback(request)
+        and _has_internal_run_scope_header(request)
+    ):
+        return scope
+    if scope == RUN_SCOPE_EXTERNAL_ROOT:
+        return RUN_SCOPE_EXTERNAL_ROOT
+    return RUN_SCOPE_EXTERNAL_ROOT
+
 
 async def _run_agent_with_concurrency_lease(
     *,
     concurrency_gate: Any | None,
     lease_acquired: bool,
-    lease_metadata: dict[str, str],
+    lease_metadata: dict[str, Any],
     heartbeat_interval_seconds: int,
     run_agent_kwargs: dict[str, Any],
 ) -> None:
@@ -750,6 +807,7 @@ async def start_run(
     request: Request,
     *,
     enforce_access: bool = True,
+    run_scope: str | None = None,
 ) -> RunRecord:
     """Create a RunRecord and launch the background agent task.
 
@@ -793,21 +851,29 @@ async def start_run(
 
     effective_model_name = _resolve_effective_run_model_name(body)
     effective_user_id = get_effective_user_id()
+    effective_run_scope = _resolve_run_scope(body, request, run_scope)
     model_concurrency_limit = _get_model_concurrency_limit(effective_model_name)
-    user_model_concurrency_limit = _get_user_model_concurrency_limit(effective_model_name)
+    user_model_concurrency_limit = (
+        _get_user_model_concurrency_limit(effective_model_name)
+        if effective_run_scope == RUN_SCOPE_EXTERNAL_ROOT
+        else None
+    )
     run_id = str(uuid.uuid4())
 
     run_metadata = dict(body.metadata or {})
     run_metadata["model_name"] = effective_model_name
     run_metadata["user_id"] = effective_user_id
+    run_metadata["run_scope"] = effective_run_scope
 
     redis_lease_acquired = False
     redis_gate_degraded = True
+    user_model_accounted = user_model_concurrency_limit is not None
     lease_metadata = {
         "run_id": run_id,
         "thread_id": thread_id,
         "model_name": effective_model_name,
         "user_id": effective_user_id,
+        "user_model_accounted": user_model_accounted,
     }
     if concurrency_gate is not None:
         acquire = await concurrency_gate.try_acquire(
diff --git a/offline-backend-20260512/backend/config.yaml b/offline-backend-20260512/backend/config.yaml
index 1b01c238..474607d1 100644
--- a/offline-backend-20260512/backend/config.yaml
+++ b/offline-backend-20260512/backend/config.yaml
@@ -18,7 +18,10 @@ concurrency:
     enabled: false
     url: redis://127.0.0.1:6379/0
     password: null
-    namespace: deerflow:concurrency
+    # Prefix with the deployment/system id because Redis may be shared by
+    # multiple projects. Keep this unique per system to avoid cross-project
+    # leases affecting each other.
+    namespace: cmzs_deerflow:concurrency
     socket_connect_timeout_ms: 80
     socket_timeout_ms: 80
     acquire_timeout_ms: 120
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/config/concurrency_config.py b/offline-backend-20260512/backend/packages/harness/deerflow/config/concurrency_config.py
index d741fda1..381e099b 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/config/concurrency_config.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/config/concurrency_config.py
@@ -14,7 +14,10 @@ class RedisConcurrencyConfig(BaseModel):
         default=None,
         description="Optional Redis password. You may also include credentials directly in url.",
     )
-    namespace: str = Field(default="deerflow:concurrency", description="Redis key prefix for concurrency leases.")
+    namespace: str = Field(
+        default="cmzs_deerflow:concurrency",
+        description="Redis key prefix for concurrency leases. Use a deployment/system prefix when Redis is shared.",
+    )
     socket_connect_timeout_ms: int = Field(default=80, ge=1, le=10_000)
     socket_timeout_ms: int = Field(default=80, ge=1, le=10_000)
     acquire_timeout_ms: int = Field(default=120, ge=1, le=10_000)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/concurrency_gate.py b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/concurrency_gate.py
index b2bd9e2f..b483a2e3 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/concurrency_gate.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/concurrency_gate.py
@@ -110,7 +110,9 @@ end
 
 redis.call('ZADD', KEYS[1], expires_at, run_id)
 redis.call('ZADD', KEYS[2], expires_at, run_id)
-redis.call('ZADD', KEYS[3], expires_at, run_id)
+if user_model_limit > 0 then
+  redis.call('ZADD', KEYS[3], expires_at, run_id)
+end
 redis.call('ZADD', KEYS[4], expires_at, run_id)
 redis.call('SET', KEYS[5], payload, 'PX', ttl_ms)
 return {'acquired', ''}
@@ -146,7 +148,7 @@ return {'acquired', ''}
                 logger.debug("Failed to close Redis concurrency client", exc_info=True)
 
     def _ns(self) -> str:
-        return self.config.namespace.strip().rstrip(":") or "deerflow:concurrency"
+        return self.config.namespace.strip().rstrip(":") or "cmzs_deerflow:concurrency"
 
     def _keys(self, *, run_id: str, model_name: str, user_id: str, thread_id: str) -> list[str]:
         ns = self._ns()
@@ -263,6 +265,7 @@ return {'acquired', ''}
         thread_id: str,
         model_name: str,
         user_id: str,
+        user_model_accounted: bool = True,
     ) -> None:
         if not self.enabled or not self._breaker.allow():
             return
@@ -272,7 +275,10 @@ return {'acquired', ''}
         try:
             client = await self._redis()
             pipe = client.pipeline(transaction=False)
-            for key in keys[:4]:
+            refresh_keys = [keys[0], keys[1], keys[3]]
+            if user_model_accounted:
+                refresh_keys.append(keys[2])
+            for key in refresh_keys:
                 pipe.zadd(key, {run_id: expires_at})
             pipe.expire(keys[4], self.config.lease_ttl_seconds)
             await self._with_timeout(pipe.execute(), self.config.operation_timeout_ms)
@@ -287,6 +293,7 @@ return {'acquired', ''}
         thread_id: str,
         model_name: str,
         user_id: str,
+        user_model_accounted: bool = True,
     ) -> None:
         if not self.enabled or not self._breaker.allow():
             return
@@ -294,7 +301,10 @@ return {'acquired', ''}
         try:
             client = await self._redis()
             pipe = client.pipeline(transaction=False)
-            for key in keys[:4]:
+            release_keys = [keys[0], keys[1], keys[3]]
+            if user_model_accounted:
+                release_keys.append(keys[2])
+            for key in release_keys:
                 pipe.zrem(key, run_id)
             pipe.delete(keys[4])
             await self._with_timeout(pipe.execute(), self.config.operation_timeout_ms)
diff --git a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/manager.py b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/manager.py
index edef5296..07daf14c 100644
--- a/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/manager.py
+++ b/offline-backend-20260512/backend/packages/harness/deerflow/runtime/runs/manager.py
@@ -273,6 +273,7 @@ class RunManager:
                     if r.status in (RunStatus.pending, RunStatus.running)
                     and str((r.metadata or {}).get("model_name") or "").strip() == model_name
                     and str((r.metadata or {}).get("user_id") or "").strip() == user_id
+                    and str((r.metadata or {}).get("run_scope") or "external_root").strip() == "external_root"
                 )
                 if active_for_user_model >= max(1, int(limit)):
                     raise ModelConcurrencyLimitError(model_name, limit, scope="user_model", user_id=user_id)
diff --git a/offline-backend-20260512/backend/tests/test_model_concurrency.py b/offline-backend-20260512/backend/tests/test_model_concurrency.py
index 86091e12..7468d9d4 100644
--- a/offline-backend-20260512/backend/tests/test_model_concurrency.py
+++ b/offline-backend-20260512/backend/tests/test_model_concurrency.py
@@ -10,11 +10,17 @@ import pytest
 
 from app.gateway.routers import system_settings as ss_router
 from app.gateway.services import (
+    RUN_SCOPE_EXTERNAL_ROOT,
+    RUN_SCOPE_ORCHESTRATION_CHILD,
     _get_model_concurrency_limit,
     _get_user_model_concurrency_limit,
+    _resolve_run_scope,
     _resolve_effective_run_model_name,
+    internal_run_scope_headers,
 )
 from deerflow.config import system_settings as ss
+from deerflow.config.concurrency_config import RedisConcurrencyConfig
+from deerflow.runtime.runs.concurrency_gate import RedisConcurrencyGate
 from deerflow.runtime.runs.manager import ModelConcurrencyLimitError, RunManager
 
 
@@ -212,6 +218,15 @@ async def test_run_manager_rejects_user_model_limit_atomically(
     assert _get_user_model_concurrency_limit("other") == 3
 
     run_mgr = RunManager()
+    await run_mgr.create_or_reject(
+        "t-child",
+        metadata={
+            "model_name": "glm5",
+            "user_id": "u1",
+            "run_scope": RUN_SCOPE_ORCHESTRATION_CHILD,
+        },
+        user_model_concurrency_limit=("glm5", "u1", 1),
+    )
     await run_mgr.create_or_reject("t1", metadata={"model_name": "glm5", "user_id": "u1"})
     await run_mgr.create_or_reject(
         "t2",
@@ -247,3 +262,44 @@ def test_get_model_concurrency_limit_skips_when_disabled(
     )
     assert _get_model_concurrency_limit("glm5") is None
     assert _get_user_model_concurrency_limit("glm5") is None
+
+
+def test_internal_run_scope_is_only_trusted_for_loopback_or_server_argument() -> None:
+    body = SimpleNamespace(metadata={"run_scope": RUN_SCOPE_ORCHESTRATION_CHILD})
+
+    trusted_loopback_request = SimpleNamespace(
+        client=SimpleNamespace(host="127.0.0.1"),
+        headers=internal_run_scope_headers(),
+    )
+    untrusted_loopback_request = SimpleNamespace(
+        client=SimpleNamespace(host="127.0.0.1"),
+        headers={},
+    )
+    public_request = SimpleNamespace(
+        client=SimpleNamespace(host="10.0.0.8"),
+        headers=internal_run_scope_headers(),
+    )
+
+    assert (
+        _resolve_run_scope(body, trusted_loopback_request, None)
+        == RUN_SCOPE_ORCHESTRATION_CHILD
+    )
+    assert (
+        _resolve_run_scope(body, untrusted_loopback_request, None)
+        == RUN_SCOPE_EXTERNAL_ROOT
+    )
+    assert _resolve_run_scope(body, public_request, None) == RUN_SCOPE_EXTERNAL_ROOT
+    assert (
+        _resolve_run_scope(
+            SimpleNamespace(metadata={}),
+            public_request,
+            RUN_SCOPE_ORCHESTRATION_CHILD,
+        )
+        == RUN_SCOPE_ORCHESTRATION_CHILD
+    )
+
+
+def test_redis_concurrency_namespace_has_system_prefix() -> None:
+    assert RedisConcurrencyConfig().namespace == "cmzs_deerflow:concurrency"
+    gate = RedisConcurrencyGate(RedisConcurrencyConfig(namespace=""))
+    assert gate._ns() == "cmzs_deerflow:concurrency"

From 2b0ba2a68b5e6e1770b9d2fd2406b96aaba6b49b Mon Sep 17 00:00:00 2001
From: Rocky Kwok 
Date: Wed, 1 Jul 2026 10:08:46 +0800
Subject: [PATCH 12/16] =?UTF-8?q?fix:=20=E6=B3=A8=E9=87=8A=E6=8F=92?=
 =?UTF-8?q?=E4=BB=B6=E7=AE=A1=E7=90=86?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../workspace/workspace-nav-menu.tsx          | 27 +++++++++----------
 1 file changed, 13 insertions(+), 14 deletions(-)

diff --git a/frontend-web/src/components/workspace/workspace-nav-menu.tsx b/frontend-web/src/components/workspace/workspace-nav-menu.tsx
index 5df42f84..67b450f3 100644
--- a/frontend-web/src/components/workspace/workspace-nav-menu.tsx
+++ b/frontend-web/src/components/workspace/workspace-nav-menu.tsx
@@ -7,12 +7,11 @@ import {
   BriefcaseIcon,
   CalendarClockIcon,
   ChevronsUpDown,
-  DatabaseZapIcon,
-  GaugeIcon,
-  HistoryIcon,
-  MessageCircleQuestionIcon,
-  PlugIcon,
-  ScrollTextIcon,
+  DatabaseZapIcon,
+  GaugeIcon,
+  HistoryIcon,
+  MessageCircleQuestionIcon,
+  ScrollTextIcon,
   Settings2Icon,
   SettingsIcon,
   SparklesIcon,
@@ -219,14 +218,14 @@ export function WorkspaceNavMenu() {
                     
                     {t.adminTags.title}
                   
-                  {isAdmin && (
-                     navigate("/page/workspace/admin/plugins")}
-                    >
-                      
-                      插件管理
-                    
-                  )}
+                  {/* {isAdmin && (
+                     navigate("/page/workspace/admin/plugins")}
+                    >
+                      
+                      插件管理
+                    
+                  )} */}
                   {isAdmin && (
                      navigate("/page/workspace/admin/positions")}

From f8cb7be18cc236f258603099b6c211e501a3e4b5 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=90=89=E9=91=AB=E5=8D=87?= 
Date: Tue, 21 Jul 2026 14:35:06 +0800
Subject: [PATCH 13/16] chore: add offline backend wheelhouse deployment

---
 offline-backend-20260512/README.md            |  64 +++++++
 offline-backend-20260512/backend/CLAUDE.md    |   1 +
 offline-backend-20260512/backend/README.md    |   1 +
 .../offline-runtime-requirements.txt          |  27 +++
 .../prepare_wheelhouse.sh                     |  99 ++++++++++
 .../start_backend_offline.sh                  | 178 ++++++++++++++++++
 6 files changed, 370 insertions(+)
 create mode 100644 offline-backend-20260512/README.md
 create mode 100644 offline-backend-20260512/offline-runtime-requirements.txt
 create mode 100755 offline-backend-20260512/prepare_wheelhouse.sh
 create mode 100755 offline-backend-20260512/start_backend_offline.sh

diff --git a/offline-backend-20260512/README.md b/offline-backend-20260512/README.md
new file mode 100644
index 00000000..e77ac35a
--- /dev/null
+++ b/offline-backend-20260512/README.md
@@ -0,0 +1,64 @@
+# DeerFlow 离线后端部署补充
+
+本目录提供一套与 `start_backend_offline.sh` 配套的 wheelhouse 流程，用于在联网机器准备 Python 依赖，再将整个发布目录复制到无网络服务器运行。
+
+## 文件
+
+- `offline-runtime-requirements.txt`：启动时补装的固定版本依赖。
+- `prepare_wheelhouse.sh`：联网机器下载 Linux/Python 3.12 wheel，并验证传递依赖完整性。
+- `start_backend_offline.sh`：从本地 wheelhouse 补装依赖后启动 Docker 后端。
+- `wheelhouse/`：下载脚本生成；应与启动脚本一起复制到离线机。
+
+## 联网机器准备
+
+默认目标是后端镜像使用的 CPython 3.12、Linux x86_64：
+
+```bash
+cd offline-backend-20260512
+bash prepare_wheelhouse.sh
+```
+
+脚本固定下载：
+
+- `croniter==6.2.2`
+- `hindsight-client==0.6.2`
+- `redis==5.2.1`
+- 上述包在 Python 3.12/Linux x86_64 下的 19 个传递依赖；版本也已全部锁定在 `offline-runtime-requirements.txt`
+
+国内镜像示例：
+
+```bash
+PIP_INDEX_URL=https://pypi.tuna.tsinghua.edu.cn/simple bash prepare_wheelhouse.sh
+```
+
+ARM64 镜像需要改目标平台：
+
+```bash
+TARGET_PLATFORM=manylinux2014_aarch64 bash prepare_wheelhouse.sh
+```
+
+下载完成后，`wheelhouse/SHA256SUMS` 可用于离线机校验文件完整性。
+
+## 离线机器启动
+
+确保发布目录还包含 Docker 镜像 tar、`.env`、`config.yaml`、`skill_evolution.json` 和 `backend/`，然后运行。`extensions_config.json` 与 `skills/` 会优先从发布根目录读取，不存在时自动回退到 `backend/`：
+
+```bash
+cd offline-backend-20260512
+bash start_backend_offline.sh
+docker logs -f deerflow-backend
+```
+
+只检查目录、配置项、wheel 和 checksum，不操作容器：
+
+```bash
+PREFLIGHT_ONLY=1 bash start_backend_offline.sh
+```
+
+启动脚本只使用 `/app/wheelhouse`，不会访问 PyPI。它会验证 checksum，并在镜像虚拟环境缺包时一次性安装三个固定版本的运行时包。
+
+## Redis 说明
+
+`redis==5.2.1` 只是 Python 客户端，不包含 Redis 服务端。`config.yaml` 中 `concurrency.redis.enabled: false` 时无需部署 Redis；启用后必须另行准备可访问的 Redis 服务或离线 Redis Docker 镜像。
+
+如果 Redis 运行在另一个容器，`redis://127.0.0.1:6379/0` 指向的是后端容器自身，不能直接使用。应将两个容器加入同一个 Docker network，并把 URL 改成类似 `redis://redis:6379/0`，同时为不同环境设置不同的 `namespace`。
diff --git a/offline-backend-20260512/backend/CLAUDE.md b/offline-backend-20260512/backend/CLAUDE.md
index 1cb9bcd3..839e9615 100644
--- a/offline-backend-20260512/backend/CLAUDE.md
+++ b/offline-backend-20260512/backend/CLAUDE.md
@@ -998,6 +998,7 @@ Both the `view_image` tool and `ViewImageMiddleware` are registered whenever `vi
 ## Documentation
 
 See `docs/` directory for detailed documentation:
+- [../README.md](../README.md) - Offline Docker wheelhouse preparation and startup workflow. The pinned versions in `../offline-runtime-requirements.txt` and `../start_backend_offline.sh` must stay synchronized. The download target defaults to CPython 3.12 on Linux x86_64 and verifies dependency closure using wheelhouse only.
 - [CONFIGURATION.md](docs/CONFIGURATION.md) - Configuration options
 - [ARCHITECTURE.md](docs/ARCHITECTURE.md) - Architecture details
 - [API.md](docs/API.md) - API reference
diff --git a/offline-backend-20260512/backend/README.md b/offline-backend-20260512/backend/README.md
index 40e7e03b..2ffcacd6 100644
--- a/offline-backend-20260512/backend/README.md
+++ b/offline-backend-20260512/backend/README.md
@@ -407,6 +407,7 @@ uv run pytest
 - [Configuration Guide](docs/CONFIGURATION.md)
 - [Architecture Details](docs/ARCHITECTURE.md)
 - [API Reference](docs/API.md)
+- [Offline Docker wheelhouse deployment (ZH)](../README.md)
 - [File Upload](docs/FILE_UPLOAD.md)
 - [Path Examples](docs/PATH_EXAMPLES.md)
 - [Context Summarization](docs/summarization.md)
diff --git a/offline-backend-20260512/offline-runtime-requirements.txt b/offline-backend-20260512/offline-runtime-requirements.txt
new file mode 100644
index 00000000..b493ed02
--- /dev/null
+++ b/offline-backend-20260512/offline-runtime-requirements.txt
@@ -0,0 +1,27 @@
+# Direct runtime packages layered into the prebuilt offline backend image.
+# Keep these three pins synchronized with start_backend_offline.sh.
+croniter==6.2.2
+hindsight-client==0.6.2
+redis==5.2.1
+
+# Fully pinned dependency closure for CPython 3.12 on Linux. Keeping the
+# transitive packages here makes a newly generated wheelhouse reproducible.
+aiohappyeyeballs==2.7.1
+aiohttp==3.14.2
+aiohttp-retry==2.9.1
+aiosignal==1.4.0
+annotated-types==0.7.0
+async-timeout==5.0.1
+attrs==26.1.0
+frozenlist==1.8.0
+idna==3.18
+multidict==6.7.1
+propcache==0.5.2
+pydantic==2.13.4
+pydantic-core==2.46.4
+python-dateutil==2.9.0.post0
+six==1.17.0
+typing-extensions==4.16.0
+typing-inspection==0.4.2
+urllib3==2.7.0
+yarl==1.24.5
diff --git a/offline-backend-20260512/prepare_wheelhouse.sh b/offline-backend-20260512/prepare_wheelhouse.sh
new file mode 100755
index 00000000..81cff94a
--- /dev/null
+++ b/offline-backend-20260512/prepare_wheelhouse.sh
@@ -0,0 +1,99 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+PYTHON_BIN="${PYTHON_BIN:-python3}"
+REQUIREMENTS_FILE="${REQUIREMENTS_FILE:-${SCRIPT_DIR}/offline-runtime-requirements.txt}"
+WHEELHOUSE_DIR="${WHEELHOUSE_DIR:-${SCRIPT_DIR}/wheelhouse}"
+PIP_INDEX_URL="${PIP_INDEX_URL:-https://pypi.org/simple}"
+
+# The supplied backend image is CPython 3.12 on Debian x86_64. These values can
+# be overridden when building an offline bundle for a different image/CPU.
+TARGET_PLATFORM="${TARGET_PLATFORM:-manylinux2014_x86_64}"
+TARGET_PYTHON_VERSION="${TARGET_PYTHON_VERSION:-312}"
+TARGET_IMPLEMENTATION="${TARGET_IMPLEMENTATION:-cp}"
+TARGET_ABI="${TARGET_ABI:-cp312}"
+
+die() {
+  echo "ERROR: $*" >&2
+  exit 1
+}
+
+require_cmd() {
+  command -v "$1" >/dev/null 2>&1 || die "command not found: $1"
+}
+
+require_wheel() {
+  local pattern="$1"
+  compgen -G "${WHEELHOUSE_DIR}/${pattern}" >/dev/null || die "wheel not downloaded: ${pattern}"
+}
+
+require_cmd "${PYTHON_BIN}"
+require_cmd sha256sum
+[ -f "${REQUIREMENTS_FILE}" ] || die "requirements file not found: ${REQUIREMENTS_FILE}"
+"${PYTHON_BIN}" -m pip --version >/dev/null 2>&1 || die "pip is not available for ${PYTHON_BIN}"
+
+mkdir -p "${WHEELHOUSE_DIR}"
+
+download_args=(
+  download
+  --dest "${WHEELHOUSE_DIR}"
+  --only-binary=:all:
+  --platform "${TARGET_PLATFORM}"
+  --python-version "${TARGET_PYTHON_VERSION}"
+  --implementation "${TARGET_IMPLEMENTATION}"
+  --abi "${TARGET_ABI}"
+)
+
+echo "Downloading Linux/Python runtime wheels"
+echo "  requirements: ${REQUIREMENTS_FILE}"
+echo "  destination:  ${WHEELHOUSE_DIR}"
+echo "  platform:     ${TARGET_PLATFORM}"
+echo "  Python/ABI:   ${TARGET_PYTHON_VERSION}/${TARGET_ABI}"
+
+"${PYTHON_BIN}" -m pip "${download_args[@]}" \
+  --index-url "${PIP_INDEX_URL}" \
+  --requirement "${REQUIREMENTS_FILE}"
+
+require_wheel 'croniter-6.2.2-*.whl'
+require_wheel 'hindsight_client-0.6.2-*.whl'
+require_wheel 'redis-5.2.1-*.whl'
+
+# Resolve the same requirements a second time using only wheelhouse. This
+# catches a missing transitive dependency before the bundle reaches offline.
+verify_dir="$(mktemp -d)"
+cleanup() {
+  if [ -n "${verify_dir:-}" ] && [ -d "${verify_dir}" ]; then
+    rm -rf -- "${verify_dir}"
+  fi
+}
+trap cleanup EXIT
+
+verify_args=(
+  download
+  --dest "${verify_dir}"
+  --no-index
+  --find-links "${WHEELHOUSE_DIR}"
+  --only-binary=:all:
+  --platform "${TARGET_PLATFORM}"
+  --python-version "${TARGET_PYTHON_VERSION}"
+  --implementation "${TARGET_IMPLEMENTATION}"
+  --abi "${TARGET_ABI}"
+  --requirement "${REQUIREMENTS_FILE}"
+)
+"${PYTHON_BIN}" -m pip "${verify_args[@]}" >/dev/null
+
+shopt -s nullglob
+wheels=("${WHEELHOUSE_DIR}"/*.whl)
+[ "${#wheels[@]}" -gt 0 ] || die "wheelhouse is empty: ${WHEELHOUSE_DIR}"
+(
+  cd "${WHEELHOUSE_DIR}"
+  sha256sum -- *.whl | LC_ALL=C sort > SHA256SUMS
+)
+
+echo
+echo "Wheelhouse is complete for the pinned runtime requirements."
+echo "  wheel count: ${#wheels[@]}"
+echo "  checksums:   ${WHEELHOUSE_DIR}/SHA256SUMS"
+echo "Copy the wheelhouse directory with the deployment bundle."
diff --git a/offline-backend-20260512/start_backend_offline.sh b/offline-backend-20260512/start_backend_offline.sh
new file mode 100755
index 00000000..1952f766
--- /dev/null
+++ b/offline-backend-20260512/start_backend_offline.sh
@@ -0,0 +1,178 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+IMAGE_TAG="${IMAGE_TAG:-deerflow-backend-offline:20260505}"
+CONTAINER_NAME="${CONTAINER_NAME:-deerflow-backend}"
+HOST_PORT="${HOST_PORT:-8001}"
+IMAGE_TAR="${IMAGE_TAR:-${SCRIPT_DIR}/deerflow-backend-offline-image-20260505.tar}"
+CONTAINER_TZ="${CONTAINER_TZ:-${TZ:-Asia/Shanghai}}"
+
+default_extensions_config="${SCRIPT_DIR}/extensions_config.json"
+if [ ! -f "${default_extensions_config}" ]; then
+  default_extensions_config="${SCRIPT_DIR}/backend/extensions_config.json"
+fi
+EXTENSIONS_CONFIG_FILE="${EXTENSIONS_CONFIG_FILE:-${default_extensions_config}}"
+
+default_skills_dir="${SCRIPT_DIR}/skills"
+if [ ! -d "${default_skills_dir}" ]; then
+  default_skills_dir="${SCRIPT_DIR}/backend/skills"
+fi
+SKILLS_DIR="${SKILLS_DIR:-${default_skills_dir}}"
+
+die() {
+  echo "ERROR: $*" >&2
+  exit 1
+}
+
+require_cmd() {
+  command -v "$1" >/dev/null 2>&1 || die "command not found: $1"
+}
+
+require_file() {
+  [ -f "$1" ] || die "required file not found: $1"
+}
+
+require_dir() {
+  [ -d "$1" ] || die "required directory not found: $1"
+}
+
+require_env_key() {
+  local key="$1"
+  grep -Eq "^[[:space:]]*${key}=" "${SCRIPT_DIR}/.env" || die "required .env key not found: ${key}"
+}
+
+require_wheel() {
+  local pattern="$1"
+  compgen -G "${SCRIPT_DIR}/wheelhouse/${pattern}" >/dev/null || die "required wheel not found: wheelhouse/${pattern}"
+}
+
+require_cmd docker
+
+if [ ! -f "${SCRIPT_DIR}/.env" ]; then
+  if [ -f "${SCRIPT_DIR}/.env.example" ]; then
+    cp "${SCRIPT_DIR}/.env.example" "${SCRIPT_DIR}/.env"
+    echo "Generated .env from .env.example. Please edit .env and run this script again."
+    exit 1
+  fi
+  die "missing .env"
+fi
+
+if [ ! -f "${SCRIPT_DIR}/config.yaml" ]; then
+  if [ -f "${SCRIPT_DIR}/config.yaml.example" ]; then
+    cp "${SCRIPT_DIR}/config.yaml.example" "${SCRIPT_DIR}/config.yaml"
+    echo "Generated config.yaml from config.yaml.example. Please edit model/API settings and run this script again."
+    exit 1
+  fi
+  die "missing config.yaml"
+fi
+
+require_env_key AUTH_JWT_SECRET
+require_env_key SCHEDULED_TASKS_ENABLED
+require_env_key ELEMENT_ENABLED
+require_file "${SCRIPT_DIR}/config.yaml"
+require_file "${EXTENSIONS_CONFIG_FILE}"
+require_file "${SCRIPT_DIR}/skill_evolution.json"
+require_file "${SCRIPT_DIR}/backend/pyproject.toml"
+require_file "${SCRIPT_DIR}/backend/app/gateway/app.py"
+require_file "${SCRIPT_DIR}/backend/packages/harness/deerflow/runtime/scheduler/element.py"
+require_file "${SCRIPT_DIR}/backend/langgraph.json"
+require_dir "${SKILLS_DIR}"
+require_dir "${SCRIPT_DIR}/wheelhouse"
+require_dir "${SCRIPT_DIR}/backend/knowledge/knowledge_base"
+require_wheel 'croniter-6.2.2-*.whl'
+require_wheel 'hindsight_client-0.6.2-*.whl'
+require_wheel 'redis-5.2.1-*.whl'
+
+if [ -f "${SCRIPT_DIR}/wheelhouse/SHA256SUMS" ]; then
+  require_cmd sha256sum
+  echo "Verifying wheelhouse checksums"
+  (cd "${SCRIPT_DIR}/wheelhouse" && sha256sum -c SHA256SUMS)
+fi
+
+mkdir -p "${SCRIPT_DIR}/data"
+
+if [ "${PREFLIGHT_ONLY:-0}" = "1" ]; then
+  echo "Offline backend preflight passed. No container was changed."
+  exit 0
+fi
+
+if ! docker image inspect "${IMAGE_TAG}" >/dev/null 2>&1; then
+  require_file "${IMAGE_TAR}"
+  echo "Loading Docker image from ${IMAGE_TAR}"
+  docker load -i "${IMAGE_TAR}"
+fi
+
+if docker ps -a --format '{{.Names}}' | awk -v name="${CONTAINER_NAME}" '$0 == name { found=1 } END { exit(found ? 0 : 1) }'; then
+  echo "Removing existing container ${CONTAINER_NAME}"
+  docker rm -f "${CONTAINER_NAME}" >/dev/null
+fi
+
+mount_args=(
+  -v "${SCRIPT_DIR}/backend/app:/app/backend/app"
+  -v "${SCRIPT_DIR}/backend/packages:/app/backend/packages"
+  -v "${SCRIPT_DIR}/backend/scripts:/app/backend/scripts"
+  -v "${SCRIPT_DIR}/backend/pyproject.toml:/app/backend/pyproject.toml:ro"
+  -v "${SCRIPT_DIR}/backend/langgraph.json:/app/backend/langgraph.json:ro"
+  -v "${SCRIPT_DIR}/config.yaml:/app/config.yaml:ro"
+  -v "${EXTENSIONS_CONFIG_FILE}:/app/extensions_config.json:ro"
+  -v "${SCRIPT_DIR}/skill_evolution.json:/app/skill_evolution.json:ro"
+  -v "${SKILLS_DIR}:/app/skills:ro"
+  -v "${SCRIPT_DIR}/wheelhouse:/app/wheelhouse:ro"
+  -v "${SCRIPT_DIR}/backend/knowledge/knowledge_base:/mnt/knowledge_base:ro"
+  -v "${SCRIPT_DIR}/data:/app/backend/.deer-flow"
+)
+
+# uv.lock is not needed by uv run --no-sync, but mount it when the release
+# bundle contains one so the container exactly reflects the online build.
+if [ -f "${SCRIPT_DIR}/backend/uv.lock" ]; then
+  mount_args+=(
+    -v "${SCRIPT_DIR}/backend/uv.lock:/app/backend/uv.lock:ro"
+  )
+fi
+
+echo "Starting backend container ${CONTAINER_NAME}"
+docker run -d \
+  --name "${CONTAINER_NAME}" \
+  --restart unless-stopped \
+  -p "${HOST_PORT}:8001" \
+  --env-file "${SCRIPT_DIR}/.env" \
+  -e TZ="${CONTAINER_TZ}" \
+  -e DEER_FLOW_PROJECT_ROOT=/app \
+  -e DEER_FLOW_HOME=/app/backend/.deer-flow \
+  -e DEER_FLOW_CONFIG_PATH=/app/config.yaml \
+  -e DEER_FLOW_EXTENSIONS_CONFIG_PATH=/app/extensions_config.json \
+  -e DEER_FLOW_SKILL_EVOLUTION_CONFIG_PATH=/app/skill_evolution.json \
+  -e DEER_FLOW_SKILLS_PATH=/app/skills \
+  -e UV_LINK_MODE=copy \
+  "${mount_args[@]}" \
+  "${IMAGE_TAG}" \
+  sh -c 'set -eu
+cd backend
+if [ -n "${TZ:-}" ] && [ -f "/usr/share/zoneinfo/${TZ}" ]; then
+  ln -sf "/usr/share/zoneinfo/${TZ}" /etc/localtime 2>/dev/null || true
+  echo "${TZ}" > /etc/timezone 2>/dev/null || true
+fi
+
+if [ "${OFFLINE_RUNTIME_INSTALL_DEPS:-1}" != "0" ]; then
+  if ! uv run --no-sync python -c "import croniter, hindsight_client, redis" >/dev/null 2>&1; then
+    echo "Installing offline runtime packages from /app/wheelhouse"
+    PYTHON_BIN="$(uv run --no-sync python -c "import sys; print(sys.executable)")"
+    # Word splitting is intentional so .env may override this package list.
+    uv pip install \
+      --python "${PYTHON_BIN}" \
+      --no-index \
+      --find-links /app/wheelhouse \
+      ${OFFLINE_RUNTIME_PIP_PACKAGES:-croniter==6.2.2 hindsight-client==0.6.2 redis==5.2.1}
+  fi
+  uv run --no-sync python -c "import importlib.metadata as m; names=(\"croniter\", \"hindsight-client\", \"redis\"); print(\"Offline runtime packages:\", \", \".join(f\"{name}={m.version(name)}\" for name in names))"
+fi
+
+PYTHONPATH=. uv run --no-sync python -m uvicorn app.gateway.app:app --host 0.0.0.0 --port 8001 --workers "${GATEWAY_WORKERS:-2}"'
+
+echo
+echo "Backend started."
+echo "  URL:       http://:${HOST_PORT}"
+echo "  Logs:      docker logs -f ${CONTAINER_NAME}"
+echo "  Stop:      docker rm -f ${CONTAINER_NAME}"

From 5d2997b6a0c6eca772052e990c1acdaeb5543ceb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=90=89=E9=91=AB=E5=8D=87?= 
Date: Tue, 21 Jul 2026 14:39:57 +0800
Subject: [PATCH 14/16] fix: preserve offline deployment LF files

---
 offline-backend-20260512/.gitattributes           | 3 +++
 offline-backend-20260512/start_backend_offline.sh | 3 ++-
 2 files changed, 5 insertions(+), 1 deletion(-)
 create mode 100644 offline-backend-20260512/.gitattributes

diff --git a/offline-backend-20260512/.gitattributes b/offline-backend-20260512/.gitattributes
new file mode 100644
index 00000000..8fa659ea
--- /dev/null
+++ b/offline-backend-20260512/.gitattributes
@@ -0,0 +1,3 @@
+*.sh text eol=lf
+offline-runtime-requirements.txt text eol=lf
+wheelhouse/SHA256SUMS text eol=lf
diff --git a/offline-backend-20260512/start_backend_offline.sh b/offline-backend-20260512/start_backend_offline.sh
index 1952f766..9e612b41 100755
--- a/offline-backend-20260512/start_backend_offline.sh
+++ b/offline-backend-20260512/start_backend_offline.sh
@@ -87,8 +87,9 @@ require_wheel 'redis-5.2.1-*.whl'
 
 if [ -f "${SCRIPT_DIR}/wheelhouse/SHA256SUMS" ]; then
   require_cmd sha256sum
+  require_cmd sed
   echo "Verifying wheelhouse checksums"
-  (cd "${SCRIPT_DIR}/wheelhouse" && sha256sum -c SHA256SUMS)
+  (cd "${SCRIPT_DIR}/wheelhouse" && sed 's/\r$//' SHA256SUMS | sha256sum -c -)
 fi
 
 mkdir -p "${SCRIPT_DIR}/data"

From e571caaf2cfa305869018fa0623d336f960f042d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=90=89=E9=91=AB=E5=8D=87?= 
Date: Tue, 21 Jul 2026 14:46:58 +0800
Subject: [PATCH 15/16] fix: keep offline start script minimal

---
 offline-backend-20260512/README.md            |  42 +++---
 offline-backend-20260512/backend/CLAUDE.md    |   2 +-
 .../offline-runtime-requirements.txt          |   2 +-
 ...rt_backend_offline.sh => start_backend.sh} | 121 ++++++------------
 4 files changed, 65 insertions(+), 102 deletions(-)
 rename offline-backend-20260512/{start_backend_offline.sh => start_backend.sh} (54%)

diff --git a/offline-backend-20260512/README.md b/offline-backend-20260512/README.md
index e77ac35a..09610f41 100644
--- a/offline-backend-20260512/README.md
+++ b/offline-backend-20260512/README.md
@@ -1,13 +1,13 @@
 # DeerFlow 离线后端部署补充
 
-本目录提供一套与 `start_backend_offline.sh` 配套的 wheelhouse 流程，用于在联网机器准备 Python 依赖，再将整个发布目录复制到无网络服务器运行。
+本目录提供与原始 `start_backend.sh` 配套的 wheelhouse。启动脚本保持原有结构，只新增一行 Redis Python 客户端的离线安装命令。
 
 ## 文件
 
-- `offline-runtime-requirements.txt`：启动时补装的固定版本依赖。
+- `offline-runtime-requirements.txt`：固定版本的完整依赖闭包。
 - `prepare_wheelhouse.sh`：联网机器下载 Linux/Python 3.12 wheel，并验证传递依赖完整性。
-- `start_backend_offline.sh`：从本地 wheelhouse 补装依赖后启动 Docker 后端。
-- `wheelhouse/`：下载脚本生成；应与启动脚本一起复制到离线机。
+- `start_backend.sh`：原始 Docker 启动脚本，加一行 Redis 缺包安装命令。
+- `wheelhouse/`：22 个 wheel 和 `SHA256SUMS`，需要随发布目录一起复制。
 
 ## 联网机器准备
 
@@ -18,12 +18,13 @@ cd offline-backend-20260512
 bash prepare_wheelhouse.sh
 ```
 
-脚本固定下载：
+直接依赖为：
 
 - `croniter==6.2.2`
 - `hindsight-client==0.6.2`
 - `redis==5.2.1`
-- 上述包在 Python 3.12/Linux x86_64 下的 19 个传递依赖；版本也已全部锁定在 `offline-runtime-requirements.txt`
+
+上述依赖在 Python 3.12/Linux x86_64 下的 19 个传递依赖也已锁定在 `offline-runtime-requirements.txt`。
 
 国内镜像示例：
 
@@ -37,28 +38,29 @@ ARM64 镜像需要改目标平台：
 TARGET_PLATFORM=manylinux2014_aarch64 bash prepare_wheelhouse.sh
 ```
 
-下载完成后，`wheelhouse/SHA256SUMS` 可用于离线机校验文件完整性。
-
 ## 离线机器启动
 
-确保发布目录还包含 Docker 镜像 tar、`.env`、`config.yaml`、`skill_evolution.json` 和 `backend/`，然后运行。`extensions_config.json` 与 `skills/` 会优先从发布根目录读取，不存在时自动回退到 `backend/`：
+保持原脚本要求的发布目录结构，包括 Docker 镜像 tar、`.env`、`config.yaml`、`extensions_config.json`、`skill_evolution.json`、`skills/`、`wheelhouse/`、`backend/uv.lock` 和 `backend/`。
+
+可先校验 wheel：
 
 ```bash
-cd offline-backend-20260512
-bash start_backend_offline.sh
+cd offline-backend-20260512/wheelhouse
+sha256sum -c SHA256SUMS
+cd ..
+```
+
+然后启动：
+
+```bash
+bash start_backend.sh
 docker logs -f deerflow-backend
 ```
 
-只检查目录、配置项、wheel 和 checksum，不操作容器：
-
-```bash
-PREFLIGHT_ONLY=1 bash start_backend_offline.sh
-```
-
-启动脚本只使用 `/app/wheelhouse`，不会访问 PyPI。它会验证 checksum，并在镜像虚拟环境缺包时一次性安装三个固定版本的运行时包。
+原脚本已有 `croniter` 和 `hindsight-client` 的离线补装逻辑；新增的一行命令只负责在缺少 `redis` 模块时，从 `/app/wheelhouse` 安装 `redis==5.2.1`。
 
 ## Redis 说明
 
-`redis==5.2.1` 只是 Python 客户端，不包含 Redis 服务端。`config.yaml` 中 `concurrency.redis.enabled: false` 时无需部署 Redis；启用后必须另行准备可访问的 Redis 服务或离线 Redis Docker 镜像。
+`redis==5.2.1` 只是 Python 客户端，不包含 Redis 服务端。`config.yaml` 中 `concurrency.redis.enabled: false` 时无需部署 Redis；启用后必须另行准备 Redis 服务或离线 Redis Docker 镜像。
 
-如果 Redis 运行在另一个容器，`redis://127.0.0.1:6379/0` 指向的是后端容器自身，不能直接使用。应将两个容器加入同一个 Docker network，并把 URL 改成类似 `redis://redis:6379/0`，同时为不同环境设置不同的 `namespace`。
+如果 Redis 运行在另一个容器，`redis://127.0.0.1:6379/0` 指向后端容器自身，不能直接使用。应将两个容器加入同一个 Docker network，并将 URL 改为类似 `redis://redis:6379/0`，同时为不同环境配置不同的 `namespace`。
diff --git a/offline-backend-20260512/backend/CLAUDE.md b/offline-backend-20260512/backend/CLAUDE.md
index 839e9615..320b879a 100644
--- a/offline-backend-20260512/backend/CLAUDE.md
+++ b/offline-backend-20260512/backend/CLAUDE.md
@@ -998,7 +998,7 @@ Both the `view_image` tool and `ViewImageMiddleware` are registered whenever `vi
 ## Documentation
 
 See `docs/` directory for detailed documentation:
-- [../README.md](../README.md) - Offline Docker wheelhouse preparation and startup workflow. The pinned versions in `../offline-runtime-requirements.txt` and `../start_backend_offline.sh` must stay synchronized. The download target defaults to CPython 3.12 on Linux x86_64 and verifies dependency closure using wheelhouse only.
+- [../README.md](../README.md) - Offline Docker wheelhouse preparation and startup workflow. The pinned versions in `../offline-runtime-requirements.txt` and `../start_backend.sh` must stay synchronized. `start_backend.sh` stays aligned with the deployment script and adds only the Redis wheel fallback line. The download target defaults to CPython 3.12 on Linux x86_64 and verifies dependency closure using wheelhouse only.
 - [CONFIGURATION.md](docs/CONFIGURATION.md) - Configuration options
 - [ARCHITECTURE.md](docs/ARCHITECTURE.md) - Architecture details
 - [API.md](docs/API.md) - API reference
diff --git a/offline-backend-20260512/offline-runtime-requirements.txt b/offline-backend-20260512/offline-runtime-requirements.txt
index b493ed02..bcea964d 100644
--- a/offline-backend-20260512/offline-runtime-requirements.txt
+++ b/offline-backend-20260512/offline-runtime-requirements.txt
@@ -1,5 +1,5 @@
 # Direct runtime packages layered into the prebuilt offline backend image.
-# Keep these three pins synchronized with start_backend_offline.sh.
+# Keep these three pins synchronized with start_backend.sh.
 croniter==6.2.2
 hindsight-client==0.6.2
 redis==5.2.1
diff --git a/offline-backend-20260512/start_backend_offline.sh b/offline-backend-20260512/start_backend.sh
similarity index 54%
rename from offline-backend-20260512/start_backend_offline.sh
rename to offline-backend-20260512/start_backend.sh
index 9e612b41..f6cfa872 100755
--- a/offline-backend-20260512/start_backend_offline.sh
+++ b/offline-backend-20260512/start_backend.sh
@@ -7,20 +7,13 @@ IMAGE_TAG="${IMAGE_TAG:-deerflow-backend-offline:20260505}"
 CONTAINER_NAME="${CONTAINER_NAME:-deerflow-backend}"
 HOST_PORT="${HOST_PORT:-8001}"
 IMAGE_TAR="${IMAGE_TAR:-${SCRIPT_DIR}/deerflow-backend-offline-image-20260505.tar}"
+# Default timezone for the container. The offline image is built from a base
+# layer that ships with UTC, which made scheduled tasks fire at the wrong wall
+# clock and made log timestamps confusing for ops. ``Asia/Shanghai`` keeps the
+# scheduler aligned with the operator's expectation; override with TZ=... when
+# deploying to a different region.
 CONTAINER_TZ="${CONTAINER_TZ:-${TZ:-Asia/Shanghai}}"
 
-default_extensions_config="${SCRIPT_DIR}/extensions_config.json"
-if [ ! -f "${default_extensions_config}" ]; then
-  default_extensions_config="${SCRIPT_DIR}/backend/extensions_config.json"
-fi
-EXTENSIONS_CONFIG_FILE="${EXTENSIONS_CONFIG_FILE:-${default_extensions_config}}"
-
-default_skills_dir="${SCRIPT_DIR}/skills"
-if [ ! -d "${default_skills_dir}" ]; then
-  default_skills_dir="${SCRIPT_DIR}/backend/skills"
-fi
-SKILLS_DIR="${SKILLS_DIR:-${default_skills_dir}}"
-
 die() {
   echo "ERROR: $*" >&2
   exit 1
@@ -43,11 +36,6 @@ require_env_key() {
   grep -Eq "^[[:space:]]*${key}=" "${SCRIPT_DIR}/.env" || die "required .env key not found: ${key}"
 }
 
-require_wheel() {
-  local pattern="$1"
-  compgen -G "${SCRIPT_DIR}/wheelhouse/${pattern}" >/dev/null || die "required wheel not found: wheelhouse/${pattern}"
-}
-
 require_cmd docker
 
 if [ ! -f "${SCRIPT_DIR}/.env" ]; then
@@ -68,37 +56,21 @@ if [ ! -f "${SCRIPT_DIR}/config.yaml" ]; then
   die "missing config.yaml"
 fi
 
-require_env_key AUTH_JWT_SECRET
-require_env_key SCHEDULED_TASKS_ENABLED
-require_env_key ELEMENT_ENABLED
+require_env_key "AUTH_JWT_SECRET"
+require_env_key "SCHEDULED_TASKS_ENABLED"
+require_env_key "ELEMENT_ENABLED"
 require_file "${SCRIPT_DIR}/config.yaml"
-require_file "${EXTENSIONS_CONFIG_FILE}"
+require_file "${SCRIPT_DIR}/extensions_config.json"
 require_file "${SCRIPT_DIR}/skill_evolution.json"
 require_file "${SCRIPT_DIR}/backend/pyproject.toml"
 require_file "${SCRIPT_DIR}/backend/app/gateway/app.py"
 require_file "${SCRIPT_DIR}/backend/packages/harness/deerflow/runtime/scheduler/element.py"
-require_file "${SCRIPT_DIR}/backend/langgraph.json"
-require_dir "${SKILLS_DIR}"
+require_dir "${SCRIPT_DIR}/skills"
 require_dir "${SCRIPT_DIR}/wheelhouse"
+# Knowledge base directory, mounted into the container at /mnt/knowledge_base.
 require_dir "${SCRIPT_DIR}/backend/knowledge/knowledge_base"
-require_wheel 'croniter-6.2.2-*.whl'
-require_wheel 'hindsight_client-0.6.2-*.whl'
-require_wheel 'redis-5.2.1-*.whl'
-
-if [ -f "${SCRIPT_DIR}/wheelhouse/SHA256SUMS" ]; then
-  require_cmd sha256sum
-  require_cmd sed
-  echo "Verifying wheelhouse checksums"
-  (cd "${SCRIPT_DIR}/wheelhouse" && sed 's/\r$//' SHA256SUMS | sha256sum -c -)
-fi
-
 mkdir -p "${SCRIPT_DIR}/data"
 
-if [ "${PREFLIGHT_ONLY:-0}" = "1" ]; then
-  echo "Offline backend preflight passed. No container was changed."
-  exit 0
-fi
-
 if ! docker image inspect "${IMAGE_TAG}" >/dev/null 2>&1; then
   require_file "${IMAGE_TAR}"
   echo "Loading Docker image from ${IMAGE_TAR}"
@@ -110,29 +82,6 @@ if docker ps -a --format '{{.Names}}' | awk -v name="${CONTAINER_NAME}" '$0 == n
   docker rm -f "${CONTAINER_NAME}" >/dev/null
 fi
 
-mount_args=(
-  -v "${SCRIPT_DIR}/backend/app:/app/backend/app"
-  -v "${SCRIPT_DIR}/backend/packages:/app/backend/packages"
-  -v "${SCRIPT_DIR}/backend/scripts:/app/backend/scripts"
-  -v "${SCRIPT_DIR}/backend/pyproject.toml:/app/backend/pyproject.toml:ro"
-  -v "${SCRIPT_DIR}/backend/langgraph.json:/app/backend/langgraph.json:ro"
-  -v "${SCRIPT_DIR}/config.yaml:/app/config.yaml:ro"
-  -v "${EXTENSIONS_CONFIG_FILE}:/app/extensions_config.json:ro"
-  -v "${SCRIPT_DIR}/skill_evolution.json:/app/skill_evolution.json:ro"
-  -v "${SKILLS_DIR}:/app/skills:ro"
-  -v "${SCRIPT_DIR}/wheelhouse:/app/wheelhouse:ro"
-  -v "${SCRIPT_DIR}/backend/knowledge/knowledge_base:/mnt/knowledge_base:ro"
-  -v "${SCRIPT_DIR}/data:/app/backend/.deer-flow"
-)
-
-# uv.lock is not needed by uv run --no-sync, but mount it when the release
-# bundle contains one so the container exactly reflects the online build.
-if [ -f "${SCRIPT_DIR}/backend/uv.lock" ]; then
-  mount_args+=(
-    -v "${SCRIPT_DIR}/backend/uv.lock:/app/backend/uv.lock:ro"
-  )
-fi
-
 echo "Starting backend container ${CONTAINER_NAME}"
 docker run -d \
   --name "${CONTAINER_NAME}" \
@@ -147,33 +96,45 @@ docker run -d \
   -e DEER_FLOW_SKILL_EVOLUTION_CONFIG_PATH=/app/skill_evolution.json \
   -e DEER_FLOW_SKILLS_PATH=/app/skills \
   -e UV_LINK_MODE=copy \
-  "${mount_args[@]}" \
+  -v "${SCRIPT_DIR}/backend/app:/app/backend/app" \
+  -v "${SCRIPT_DIR}/backend/packages:/app/backend/packages" \
+  -v "${SCRIPT_DIR}/backend/scripts:/app/backend/scripts" \
+  -v "${SCRIPT_DIR}/backend/pyproject.toml:/app/backend/pyproject.toml:ro" \
+  -v "${SCRIPT_DIR}/backend/uv.lock:/app/backend/uv.lock:ro" \
+  -v "${SCRIPT_DIR}/backend/langgraph.json:/app/backend/langgraph.json:ro" \
+  -v "${SCRIPT_DIR}/config.yaml:/app/config.yaml:ro" \
+  -v "${SCRIPT_DIR}/extensions_config.json:/app/extensions_config.json:ro" \
+  -v "${SCRIPT_DIR}/skill_evolution.json:/app/skill_evolution.json:ro" \
+  -v "${SCRIPT_DIR}/skills:/app/skills:ro" \
+  -v "${SCRIPT_DIR}/wheelhouse:/app/wheelhouse:ro" \
+  -v "${SCRIPT_DIR}/backend/knowledge/knowledge_base:/mnt/knowledge_base:ro" \
+  -v "${SCRIPT_DIR}/data:/app/backend/.deer-flow" \
   "${IMAGE_TAG}" \
-  sh -c 'set -eu
-cd backend
+  sh -c 'cd backend
+# Apply timezone inside the container. ``TZ`` alone is enough for Python /
+# uvicorn log timestamps, but linking /etc/localtime keeps C-level tools
+# (e.g. anything that reads /etc/localtime directly) on the same clock.
 if [ -n "${TZ:-}" ] && [ -f "/usr/share/zoneinfo/${TZ}" ]; then
   ln -sf "/usr/share/zoneinfo/${TZ}" /etc/localtime 2>/dev/null || true
   echo "${TZ}" > /etc/timezone 2>/dev/null || true
 fi
-
-if [ "${OFFLINE_RUNTIME_INSTALL_DEPS:-1}" != "0" ]; then
-  if ! uv run --no-sync python -c "import croniter, hindsight_client, redis" >/dev/null 2>&1; then
-    echo "Installing offline runtime packages from /app/wheelhouse"
+if [ "${SCHEDULER_INSTALL_DEPS:-1}" != "0" ]; then
+  uv run --no-sync python -c "import croniter" >/dev/null 2>&1 || {
     PYTHON_BIN="$(uv run --no-sync python -c "import sys; print(sys.executable)")"
-    # Word splitting is intentional so .env may override this package list.
-    uv pip install \
-      --python "${PYTHON_BIN}" \
-      --no-index \
-      --find-links /app/wheelhouse \
-      ${OFFLINE_RUNTIME_PIP_PACKAGES:-croniter==6.2.2 hindsight-client==0.6.2 redis==5.2.1}
-  fi
-  uv run --no-sync python -c "import importlib.metadata as m; names=(\"croniter\", \"hindsight-client\", \"redis\"); print(\"Offline runtime packages:\", \", \".join(f\"{name}={m.version(name)}\" for name in names))"
+    uv pip install --python "${PYTHON_BIN}" --no-index --find-links /app/wheelhouse ${SCHEDULER_PIP_PACKAGES:-croniter==6.2.2}
+  }
 fi
-
-PYTHONPATH=. uv run --no-sync python -m uvicorn app.gateway.app:app --host 0.0.0.0 --port 8001 --workers "${GATEWAY_WORKERS:-2}"'
+if [ "${HINDSIGHT_INSTALL_DEPS:-1}" != "0" ]; then
+  uv run --no-sync python -c "import hindsight_client" >/dev/null 2>&1 || {
+    PYTHON_BIN="$(uv run --no-sync python -c "import sys; print(sys.executable)")"
+    uv pip install --python "${PYTHON_BIN}" --no-index --find-links /app/wheelhouse ${HINDSIGHT_PIP_PACKAGES:-hindsight-client==0.6.2}
+  }
+fi
+uv run --no-sync python -c "import redis" >/dev/null 2>&1 || uv pip install --python "$(uv run --no-sync python -c "import sys; print(sys.executable)")" --no-index --find-links /app/wheelhouse ${REDIS_PIP_PACKAGES:-redis==5.2.1}
+PYTHONPATH=. uv run --no-sync python -m uvicorn app.gateway.app:app --host 0.0.0.0 --port 8001 --workers ${GATEWAY_WORKERS:-2}'
 
 echo
 echo "Backend started."
 echo "  URL:       http://:${HOST_PORT}"
 echo "  Logs:      docker logs -f ${CONTAINER_NAME}"
-echo "  Stop:      docker rm -f ${CONTAINER_NAME}"
+echo "  Stop:      ./stop_backend.sh"

From 59f2f56124e69cc373ae0954024614f99a7b2bab Mon Sep 17 00:00:00 2001
From: Rocky Kwok 
Date: Sun, 26 Jul 2026 17:01:09 +0800
Subject: [PATCH 16/16] =?UTF-8?q?fix:iframe=E5=88=87=E6=8D=A2=E9=A2=9C?=
 =?UTF-8?q?=E8=89=B2?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 frontend-web/src/App.tsx                        |  5 ++++-
 .../page-layout/chats-iframe-route.test.mjs     | 17 +++++++++++++++++
 .../src/core/page-layout/chats-iframe-route.ts  | 13 +++++++++++--
 3 files changed, 32 insertions(+), 3 deletions(-)

diff --git a/frontend-web/src/App.tsx b/frontend-web/src/App.tsx
index a4a67c48..0235b013 100644
--- a/frontend-web/src/App.tsx
+++ b/frontend-web/src/App.tsx
@@ -61,7 +61,10 @@ function AppProviders({ children }: { children: React.ReactNode }) {
     // 走 next-themes forcedTheme（不写 localStorage）→ 同浏览器另开的正常系统页不受影响。
     const embedTheme = resolveForcedEmbedTheme(location.search);
     if (embedTheme) return embedTheme;
-    const routeTheme = resolveForcedThemeForPath(location.pathname);
+    const routeTheme = resolveForcedThemeForPath(
+      location.pathname,
+      location.search,
+    );
     if (routeTheme) return routeTheme;
     if (location.pathname === "/embed/roundtable") {
       const wp = readWujieProps();
diff --git a/frontend-web/src/core/page-layout/chats-iframe-route.test.mjs b/frontend-web/src/core/page-layout/chats-iframe-route.test.mjs
index eefe33b2..5cdb649b 100644
--- a/frontend-web/src/core/page-layout/chats-iframe-route.test.mjs
+++ b/frontend-web/src/core/page-layout/chats-iframe-route.test.mjs
@@ -31,3 +31,20 @@ test("forces light theme for landing and chats_iframe routes only", () => {
   );
   assert.equal(resolveForcedThemeForPath("/page/workspace/chats/new"), undefined);
 });
+
+test("honors the dark-blue theme query on chats_iframe routes", () => {
+  assert.equal(
+    resolveForcedThemeForPath(
+      "/page/workspace/chats_iframe/new",
+      "?from=iframe&theme=dark-blue",
+    ),
+    "dark",
+  );
+  assert.equal(
+    resolveForcedThemeForPath(
+      "/page/workspace/chats_iframe/new",
+      "?from=iframe&theme=red",
+    ),
+    "light",
+  );
+});
diff --git a/frontend-web/src/core/page-layout/chats-iframe-route.ts b/frontend-web/src/core/page-layout/chats-iframe-route.ts
index 30c6ba78..06b13117 100644
--- a/frontend-web/src/core/page-layout/chats-iframe-route.ts
+++ b/frontend-web/src/core/page-layout/chats-iframe-route.ts
@@ -1,3 +1,5 @@
+import { resolveEmbedThemeMode } from "@/core/embed/embed-session";
+
 export function isChatsIframeWorkspacePath(pathname: string): boolean {
   return (
     pathname === "/page/workspace/chats_iframe/new" ||
@@ -7,10 +9,17 @@ export function isChatsIframeWorkspacePath(pathname: string): boolean {
 
 export function resolveForcedThemeForPath(
   pathname: string,
-): "light" | undefined {
-  if (pathname === "/landing" || isChatsIframeWorkspacePath(pathname)) {
+  search = "",
+): "light" | "dark" | undefined {
+  if (pathname === "/landing") {
     return "light";
   }
 
+  if (isChatsIframeWorkspacePath(pathname)) {
+    const normalizedSearch = search.startsWith("?") ? search : `?${search}`;
+    const rawTheme = new URLSearchParams(normalizedSearch).get("theme")?.trim();
+    return rawTheme ? resolveEmbedThemeMode(rawTheme) : "light";
+  }
+
   return undefined;
 }