|
500 | 500 | <a href="#toolshell"><span class="nav-icon">🐚</span><span data-i18n="sidebar.toolshell">ToolShell</span></a> |
501 | 501 | <a href="#mcp"><span class="nav-icon">🔌</span><span data-i18n="sidebar.mcp">MCP 协议</span></a> |
502 | 502 | <a href="#memory"><span class="nav-icon">🧠</span><span data-i18n="sidebar.memory">记忆系统</span></a> |
| 503 | + <a href="#knowledge"><span class="nav-icon">📚</span><span data-i18n="sidebar.knowledge">知识库</span></a> |
| 504 | + <a href="#workflows"><span class="nav-icon">🌿</span><span data-i18n="sidebar.workflows">并行与版本工作流</span></a> |
503 | 505 | <a href="#skills"><span class="nav-icon">📦</span><span data-i18n="sidebar.skills">技能系统</span></a> |
504 | 506 | <a href="#search"><span class="nav-icon">🔍</span><span data-i18n="sidebar.search">Web 搜索</span></a> |
505 | 507 | <a href="#experts"><span class="nav-icon">👤</span><span data-i18n="sidebar.experts">领域专家</span></a> |
@@ -906,6 +908,39 @@ <h4>3. <span data-i18n="memory.useTitle">使用记忆</span></h4> |
906 | 908 | </div> |
907 | 909 | </section> |
908 | 910 |
|
| 911 | +<!-- ==================== 知识库 ==================== --> |
| 912 | +<section class="doc-section" id="knowledge"> |
| 913 | + <h2><span class="section-icon">📚</span> <span data-i18n="knowledge.title">知识库</span></h2> |
| 914 | + <p class="section-desc" data-i18n="knowledge.desc">知识库用于把本地文档建立为可检索的向量资料集。每个知识库拥有独立的嵌入配置与文档集合，可在对话中按会话接入一个或多个知识库。</p> |
| 915 | + <h3 data-i18n="knowledge.flowTitle">使用流程</h3> |
| 916 | + <ol> |
| 917 | + <li data-i18n="knowledge.flow1">进入「服务 → 知识库」，创建知识库并选择嵌入模型</li> |
| 918 | + <li data-i18n="knowledge.flow2">导入文档，RemindAI 会提取文本、分块并批量建立向量索引</li> |
| 919 | + <li data-i18n="knowledge.flow3">在对话输入区点击「知识库」，选择当前会话要接入的资料库</li> |
| 920 | + <li data-i18n="knowledge.flow4">发送问题后，相关片段会在模型调用前自动召回并注入上下文</li> |
| 921 | + </ol> |
| 922 | + <div class="callout callout-info"><span class="callout-icon">ℹ️</span><span data-i18n="knowledge.tip">知识库与长期记忆彼此独立：知识库面向成批资料，长期记忆面向跨会话事实、决策和偏好。</span></div> |
| 923 | +</section> |
| 924 | + |
| 925 | +<!-- ==================== 并行与版本工作流 ==================== --> |
| 926 | +<section class="doc-section" id="workflows"> |
| 927 | + <h2><span class="section-icon">🌿</span> <span data-i18n="workflows.title">并行理解与版本工作流</span></h2> |
| 928 | + <p class="section-desc" data-i18n="workflows.desc">面向大型项目，RemindAI 同时提供只读并行发现与 Git Worktree 隔离修改能力：先快速理解，再安全实验。</p> |
| 929 | + <h3 data-i18n="workflows.parallelTitle">并行只读理解</h3> |
| 930 | + <ul> |
| 931 | + <li data-i18n="workflows.parallel1"><code>toolshell_run_parallel</code> 可并发执行互不依赖的读取、搜索和记忆召回</li> |
| 932 | + <li data-i18n="workflows.parallel2"><code>/sub-readers</code> 会先规划 1–6 个不重叠子任务，再由独立只读子 Agent 并行执行，最后由主模型去重汇总</li> |
| 933 | + <li data-i18n="workflows.parallel3">子 Agent 不能写入、删除或执行命令，因此可以安全地指向同一个项目目录</li> |
| 934 | + </ul> |
| 935 | + <h3 data-i18n="workflows.worktreeTitle">Git Worktree 隔离修改</h3> |
| 936 | + <ul> |
| 937 | + <li data-i18n="workflows.worktree1">实验性重构或依赖升级可在 <code>.toolshell/worktrees/</code> 的独立分支中完成</li> |
| 938 | + <li data-i18n="workflows.worktree2">支持创建检查点、列出检查点、硬回退和查看相对主分支的差异</li> |
| 939 | + <li data-i18n="workflows.worktree3">验证通过后合并回主分支；不满意可整体丢弃，主工作目录不受影响</li> |
| 940 | + </ul> |
| 941 | + <div class="callout callout-warn"><span class="callout-icon">⚠️</span><span data-i18n="workflows.tip">Worktree 功能要求当前工作目录是 Git 仓库；是否启用由 AI 根据改动风险判断，不会自动强制开启。</span></div> |
| 942 | +</section> |
| 943 | + |
909 | 944 | <!-- ==================== 技能系统 ==================== --> |
910 | 945 | <section class="doc-section" id="skills"> |
911 | 946 | <h2><span class="section-icon">📦</span> <span data-i18n="skills.title">技能系统</span></h2> |
@@ -1291,6 +1326,8 @@ <h4 data-i18n="faq.q6">Q: Linux / macOS 如何使用？</h4> |
1291 | 1326 | 'sidebar.toolshell': 'ToolShell', |
1292 | 1327 | 'sidebar.mcp': 'MCP 协议', |
1293 | 1328 | 'sidebar.memory': '记忆系统', |
| 1329 | + 'sidebar.knowledge': '知识库', |
| 1330 | + 'sidebar.workflows': '并行与版本工作流', |
1294 | 1331 | 'sidebar.skills': '技能系统', |
1295 | 1332 | 'sidebar.search': 'Web 搜索', |
1296 | 1333 | 'sidebar.experts': '领域专家', |
@@ -1451,6 +1488,25 @@ <h4 data-i18n="faq.q6">Q: Linux / macOS 如何使用？</h4> |
1451 | 1488 | 'memory.useL3': '你可以在「🧠 记忆」页面查看、搜索和删除记忆条目', |
1452 | 1489 | 'memory.writeScreenshot': '记忆存储', |
1453 | 1490 | 'memory.writeCaption': 'AI 自动识别并存储重要信息到记忆', |
| 1491 | + 'knowledge.title': '知识库', |
| 1492 | + 'knowledge.desc': '知识库用于把本地文档建立为可检索的向量资料集。每个知识库拥有独立的嵌入配置与文档集合，可在对话中按会话接入一个或多个知识库。', |
| 1493 | + 'knowledge.flowTitle': '使用流程', |
| 1494 | + 'knowledge.flow1': '进入「服务 → 知识库」，创建知识库并选择嵌入模型', |
| 1495 | + 'knowledge.flow2': '导入文档，RemindAI 会提取文本、分块并批量建立向量索引', |
| 1496 | + 'knowledge.flow3': '在对话输入区点击「知识库」，选择当前会话要接入的资料库', |
| 1497 | + 'knowledge.flow4': '发送问题后，相关片段会在模型调用前自动召回并注入上下文', |
| 1498 | + 'knowledge.tip': '知识库与长期记忆彼此独立：知识库面向成批资料，长期记忆面向跨会话事实、决策和偏好。', |
| 1499 | + 'workflows.title': '并行理解与版本工作流', |
| 1500 | + 'workflows.desc': '面向大型项目，RemindAI 同时提供只读并行发现与 Git Worktree 隔离修改能力：先快速理解，再安全实验。', |
| 1501 | + 'workflows.parallelTitle': '并行只读理解', |
| 1502 | + 'workflows.parallel1': '<code>toolshell_run_parallel</code> 可并发执行互不依赖的读取、搜索和记忆召回', |
| 1503 | + 'workflows.parallel2': '<code>/sub-readers</code> 会先规划 1–6 个不重叠子任务，再由独立只读子 Agent 并行执行，最后由主模型去重汇总', |
| 1504 | + 'workflows.parallel3': '子 Agent 不能写入、删除或执行命令，因此可以安全地指向同一个项目目录', |
| 1505 | + 'workflows.worktreeTitle': 'Git Worktree 隔离修改', |
| 1506 | + 'workflows.worktree1': '实验性重构或依赖升级可在 <code>.toolshell/worktrees/</code> 的独立分支中完成', |
| 1507 | + 'workflows.worktree2': '支持创建检查点、列出检查点、硬回退和查看相对主分支的差异', |
| 1508 | + 'workflows.worktree3': '验证通过后合并回主分支；不满意可整体丢弃，主工作目录不受影响', |
| 1509 | + 'workflows.tip': 'Worktree 功能要求当前工作目录是 Git 仓库；是否启用由 AI 根据改动风险判断，不会自动强制开启。', |
1454 | 1510 | 'skills.title': '技能系统', |
1455 | 1511 | 'skills.desc': '技能系统允许你为 AI 添加额外的工具和能力。每个技能包含一个 SKILL.md（技能说明）和 tools.json（工具定义），打包为 ZIP 文件导入。', |
1456 | 1512 | 'skills.screenshot': '技能系统', |
@@ -1650,6 +1706,8 @@ <h4 data-i18n="faq.q6">Q: Linux / macOS 如何使用？</h4> |
1650 | 1706 | 'sidebar.toolshell': 'ToolShell', |
1651 | 1707 | 'sidebar.mcp': 'MCP Protocol', |
1652 | 1708 | 'sidebar.memory': 'Memory System', |
| 1709 | + 'sidebar.knowledge': 'Knowledge Bases', |
| 1710 | + 'sidebar.workflows': 'Parallel & Version Workflows', |
1653 | 1711 | 'sidebar.skills': 'Skill System', |
1654 | 1712 | 'sidebar.search': 'Web Search', |
1655 | 1713 | 'sidebar.experts': 'Domain Experts', |
@@ -1810,6 +1868,25 @@ <h4 data-i18n="faq.q6">Q: Linux / macOS 如何使用？</h4> |
1810 | 1868 | 'memory.useL3': 'You can view, search, and delete memory entries on the "🧠 Memory" page', |
1811 | 1869 | 'memory.writeScreenshot': 'Memory Storage', |
1812 | 1870 | 'memory.writeCaption': 'AI automatically identifies and stores important information to memory', |
| 1871 | + 'knowledge.title': 'Knowledge Bases', |
| 1872 | + 'knowledge.desc': 'Knowledge bases turn local documents into searchable vector collections. Each base has its own embedding configuration and document set, and one or more bases can be attached per chat session.', |
| 1873 | + 'knowledge.flowTitle': 'Workflow', |
| 1874 | + 'knowledge.flow1': 'Open Services → Knowledge Base, create a base, and select an embedding model', |
| 1875 | + 'knowledge.flow2': 'Import documents; RemindAI extracts text, chunks it, and builds vector indexes in batches', |
| 1876 | + 'knowledge.flow3': 'Click Knowledge Base in the chat composer and choose the collections for the current session', |
| 1877 | + 'knowledge.flow4': 'Before the model call, relevant passages are automatically recalled and injected into context', |
| 1878 | + 'knowledge.tip': 'Knowledge bases and long-term memory are separate: bases are for document collections, while memory stores cross-session facts, decisions, and preferences.', |
| 1879 | + 'workflows.title': 'Parallel Comprehension & Version Workflows', |
| 1880 | + 'workflows.desc': 'For large projects, RemindAI combines read-only parallel discovery with Git Worktree isolation: understand quickly, then experiment safely.', |
| 1881 | + 'workflows.parallelTitle': 'Parallel Read-Only Comprehension', |
| 1882 | + 'workflows.parallel1': '<code>toolshell_run_parallel</code> concurrently runs independent reads, searches, and memory recalls', |
| 1883 | + 'workflows.parallel2': '<code>/sub-readers</code> plans 1–6 non-overlapping tasks, runs independent read-only sub-agents in parallel, then asks the main model to deduplicate and synthesize the findings', |
| 1884 | + 'workflows.parallel3': 'Sub-agents cannot write, delete, or execute commands, so they can safely share the same project root', |
| 1885 | + 'workflows.worktreeTitle': 'Git Worktree Isolation', |
| 1886 | + 'workflows.worktree1': 'Experimental refactors and dependency upgrades can run on an isolated branch under <code>.toolshell/worktrees/</code>', |
| 1887 | + 'workflows.worktree2': 'Create and list checkpoints, hard-revert to a checkpoint, and inspect diffs against the main branch', |
| 1888 | + 'workflows.worktree3': 'Merge verified changes back to main, or discard the entire experiment without touching the main working directory', |
| 1889 | + 'workflows.tip': 'Worktree support requires the current workspace to be a Git repository. The AI decides when it is useful based on risk; it is never forced automatically.', |
1813 | 1890 | 'skills.title': 'Skill System', |
1814 | 1891 | 'skills.desc': 'The skill system lets you add additional tools and capabilities to AI. Each skill includes a SKILL.md (skill description) and tools.json (tool definitions), packaged as a ZIP file for import.', |
1815 | 1892 | 'skills.screenshot': 'Skill System', |
|
0 commit comments