[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-616":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":23,"hasPages":23,"topics":25,"createdAt":10,"pushedAt":10,"updatedAt":26,"readmeContent":27,"aiSummary":28,"trendingCount":16,"starSnapshotCount":16,"syncStatus":29,"lastSyncTime":30,"discoverSource":31},616,"llm_wiki","nashsu\u002Fllm_wiki","nashsu","LLM Wiki is a cross-platform desktop application that turns your documents into an organized, interlinked knowledge base — automatically. Instead of traditional RAG (retrieve-and-answer from scratch every time), the LLM incrementally builds and maintains a persistent wiki from your sources。","",null,"TypeScript",11069,1363,46,104,0,467,1102,2100,1401,44.4,"Other",false,"main",[],"2026-06-12 02:00:16","# LLM Wiki\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"logo.jpg\" width=\"128\" height=\"128\" style=\"border-radius: 22%;\" alt=\"LLM Wiki Logo\">\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cstrong>A personal knowledge base that builds itself.\u003C\u002Fstrong>\u003Cbr>\n  LLM reads your documents, builds a structured wiki, and keeps it current.\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"#what-is-this\">What is this?\u003C\u002Fa> •\n  \u003Ca href=\"#what-we-changed--added\">Features\u003C\u002Fa> •\n  \u003Ca href=\"#tech-stack\">Tech Stack\u003C\u002Fa> •\n  \u003Ca href=\"#installation\">Installation\u003C\u002Fa> •\n  \u003Ca href=\"#credits\">Credits\u003C\u002Fa> •\n  \u003Ca href=\"#license\">License\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  English | \u003Ca href=\"README_CN.md\">中文\u003C\u002Fa>\n\u003C\u002Fp>\n\n---\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002Foverview.jpg\" width=\"100%\" alt=\"Overview\">\n\u003C\u002Fp>\n\n## Features\n\n- **Two-Step Chain-of-Thought Ingest** — LLM analyzes first, then generates wiki pages with source traceability and incremental cache\n- **Multimodal Image Ingestion** — extract embedded images from PDFs, generate factual captions with a vision LLM, surface them in image-aware search results with lightbox preview and jump-to-source\n- **4-Signal Knowledge Graph** — relevance model with direct links, source overlap, Adamic-Adar, and type affinity\n- **Louvain Community Detection** — automatic knowledge cluster discovery with cohesion scoring\n- **Graph Insights** — surprising connections and knowledge gaps with one-click Deep Research\n- **Vector Semantic Search** — optional embedding-based retrieval via LanceDB, supports any OpenAI-compatible endpoint\n- **Persistent Ingest Queue** — serial processing with crash recovery, cancel, retry, and progress visualization\n- **Folder Import** — recursive folder import preserving directory structure, folder context as LLM classification hint\n- **Deep Research** — LLM-optimized search topics, multi-query web search, auto-ingest results into wiki\n- **Async Review System** — LLM flags items for human judgment, predefined actions, pre-generated search queries\n- **Chrome Web Clipper** — one-click web page capture with auto-ingest into knowledge base\n\n## What is this?\n\nLLM Wiki is a cross-platform desktop application that turns your documents into an organized, interlinked knowledge base — automatically. Instead of traditional RAG (retrieve-and-answer from scratch every time), the LLM **incrementally builds and maintains a persistent wiki** from your sources. Knowledge is compiled once and kept current, not re-derived on every query.\n\nThis project is based on [Karpathy's LLM Wiki pattern](https:\u002F\u002Fgist.github.com\u002Fkarpathy\u002F442a6bf555914893e9891c11519de94f) — a methodology for building personal knowledge bases using LLMs. We implemented the core ideas as a full desktop application with significant enhancements.\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002Fllm_wiki_arch.jpg\" width=\"100%\" alt=\"LLM Wiki Architecture\">\n\u003C\u002Fp>\n\n## Credits\n\nThe foundational methodology comes from **Andrej Karpathy**'s [llm-wiki.md](https:\u002F\u002Fgist.github.com\u002Fkarpathy\u002F442a6bf555914893e9891c11519de94f), which describes the pattern of using LLMs to incrementally build and maintain a personal wiki. The original document is an abstract design pattern; this project is a concrete implementation with substantial extensions.\n\n## What We Kept from the Original\n\nThe core architecture follows Karpathy's design faithfully:\n\n- **Three-layer architecture**: Raw Sources (immutable) → Wiki (LLM-generated) → Schema (rules & config)\n- **Three core operations**: Ingest, Query, Lint\n- **index.md** as the content catalog and LLM navigation entry point\n- **log.md** as the chronological operation record with parseable format\n- **[[wikilink]]** syntax for cross-references\n- **YAML frontmatter** on every wiki page\n- **Obsidian compatibility** — the wiki directory works as an Obsidian vault\n- **Human curates, LLM maintains** — the fundamental role division\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002F5-obsidian_compatibility.jpg\" width=\"100%\" alt=\"Obsidian Compatibility\">\n\u003C\u002Fp>\n\n## What We Changed & Added\n\n### 1. From CLI to Desktop Application\n\nThe original is an abstract pattern document designed to be copy-pasted to an LLM agent. We built it into a **full cross-platform desktop application** with:\n- **Three-column layout**: Knowledge Tree \u002F File Tree (left) + Chat (center) + Preview (right)\n- **Icon sidebar** for switching between Wiki, Sources, Search, Graph, Lint, Review, Deep Research, Settings\n- **Custom resizable panels** — drag-to-resize left and right panels with min\u002Fmax constraints\n- **Activity panel** — real-time processing status showing file-by-file ingest progress\n- **All state persisted** — conversations, settings, review items, project config survive restarts\n- **Scenario templates** — Research, Reading, Personal Growth, Business, General — each pre-configures purpose.md and schema.md\n\n### 2. Purpose.md — The Wiki's Soul\n\nThe original has Schema (how the wiki works) but no formal place for **why** the wiki exists. We added `purpose.md`:\n- Defines goals, key questions, research scope, evolving thesis\n- LLM reads it during every ingest and query for context\n- LLM can suggest updates based on usage patterns\n- Different from schema — schema is structural rules, purpose is directional intent\n\n### 3. Two-Step Chain-of-Thought Ingest\n\nThe original describes a single-step ingest where the LLM reads and writes simultaneously. We split it into **two sequential LLM calls** for significantly better quality:\n\n```\nStep 1 (Analysis): LLM reads source → structured analysis\n  - Key entities, concepts, arguments\n  - Connections to existing wiki content\n  - Contradictions & tensions with existing knowledge\n  - Recommendations for wiki structure\n\nStep 2 (Generation): LLM takes analysis → generates wiki files\n  - Source summary with frontmatter (type, title, sources[])\n  - Entity pages, concept pages with cross-references\n  - Updated index.md, log.md, overview.md\n  - Review items for human judgment\n  - Search queries for Deep Research\n```\n\nAdditional ingest enhancements beyond the original:\n- **SHA256 incremental cache** — source file content is hashed before ingest; unchanged files are skipped automatically, saving LLM tokens and time\n- **Persistent ingest queue** — serial processing prevents concurrent LLM calls; queue persisted to disk, survives app restart; failed tasks auto-retry up to 3 times\n- **Folder import** — recursive folder import preserving directory structure; folder path passed to LLM as classification context (e.g., \"papers > energy\" helps categorize content)\n- **Queue visualization** — Activity Panel shows progress bar, pending\u002Fprocessing\u002Ffailed tasks with cancel and retry buttons\n- **Auto-embedding** — when vector search is enabled, new pages are automatically embedded after ingest\n- **Source traceability** — every generated wiki page includes a `sources: []` field in YAML frontmatter, linking back to the raw source files that contributed to it\n- **overview.md auto-update** — global summary page regenerated on every ingest to reflect the latest state of the wiki\n- **Guaranteed source summary** — fallback ensures a source summary page is always created, even if the LLM omits it\n- **Language-aware generation** — LLM responds in the user's configured language (English or Chinese)\n\n### 4. Knowledge Graph with Relevance Model\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002F3-knowledge_graph.jpg\" width=\"100%\" alt=\"Knowledge Graph\">\n\u003C\u002Fp>\n\nThe original mentions `[[wikilinks]]` for cross-references but has no graph analysis. We built a **full knowledge graph visualization and relevance engine**:\n\n**4-Signal Relevance Model:**\n| Signal | Weight | Description |\n|--------|--------|-------------|\n| Direct link | ×3.0 | Pages linked via `[[wikilinks]]` |\n| Source overlap | ×4.0 | Pages sharing the same raw source (via frontmatter `sources[]`) |\n| Adamic-Adar | ×1.5 | Pages sharing common neighbors (weighted by neighbor degree) |\n| Type affinity | ×1.0 | Bonus for same page type (entity↔entity, concept↔concept) |\n\n**Graph Visualization (sigma.js + graphology + ForceAtlas2):**\n- Node colors by page type or community, sizes scaled by link count (√ scaling)\n- Edge thickness and color by relevance weight (green=strong, gray=weak)\n- Hover interaction: neighbors stay visible, non-neighbors dim, edges highlight with relevance score label\n- Zoom controls (ZoomIn, ZoomOut, Fit-to-screen)\n- Position caching prevents layout jumps when data updates\n- Legend switches between type counts and community info based on coloring mode\n\n### 5. Louvain Community Detection\n\nNot in the original. Automatic discovery of knowledge clusters using the **Louvain algorithm** (graphology-communities-louvain):\n\n- **Auto-clustering** — discovers which pages naturally group together based on link topology, independent of predefined page types\n- **Type \u002F Community toggle** — switch between coloring nodes by page type (entity, concept, source...) or by discovered knowledge cluster\n- **Cohesion scoring** — each community scored by intra-edge density (actual edges \u002F possible edges); low-cohesion clusters (\u003C 0.15) flagged with warning\n- **12-color palette** — distinct visual separation between clusters\n- **Community legend** — shows top node label, member count, and cohesion per cluster\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002Fkg_community.jpg\" width=\"100%\" alt=\"Louvain Community Detection\">\n\u003C\u002Fp>\n\n### 6. Graph Insights — Surprising Connections & Knowledge Gaps\n\nNot in the original. The system **automatically analyzes graph structure** to surface actionable insights:\n\n**Surprising Connections:**\n- Detects unexpected relationships: cross-community edges, cross-type links, peripheral↔hub couplings\n- Composite surprise score ranks the most noteworthy connections\n- Dismissable — mark connections as reviewed so they don't reappear\n\n**Knowledge Gaps:**\n- **Isolated pages** (degree ≤ 1) — pages with few or no connections to the rest of the wiki\n- **Sparse communities** (cohesion \u003C 0.15, ≥ 3 pages) — knowledge areas with weak internal cross-references\n- **Bridge nodes** (connecting 3+ clusters) — critical junction pages that hold multiple knowledge areas together\n\n**Interactive:**\n- Click any insight card to **highlight** corresponding nodes and edges in the graph; click again to deselect\n- Knowledge gaps and bridge nodes have a **Deep Research button** — triggers LLM-optimized research with domain-aware topics (reads overview.md + purpose.md for context)\n- Research topic shown in **editable confirmation dialog** before starting — user can refine topic and search queries\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002Fkg_insights.jpg\" width=\"100%\" alt=\"Graph Insights\">\n\u003C\u002Fp>\n\n### 7. Optimized Query Retrieval Pipeline\n\nThe original describes a simple query where the LLM reads relevant pages. We built a **multi-phase retrieval pipeline** with optional vector search and budget control:\n\n```\nPhase 1: Tokenized Search\n  - English: word splitting + stop word removal\n  - Chinese: CJK bigram tokenization (每个 → [每个, 个…])\n  - Title match bonus (+10 score)\n  - Searches both wiki\u002F and raw\u002Fsources\u002F\n\nPhase 1.5: Vector Semantic Search (optional)\n  - Embedding via any OpenAI-compatible \u002Fv1\u002Fembeddings endpoint\n  - Stored in LanceDB (Rust backend) for fast ANN retrieval\n  - Cosine similarity finds semantically related pages even without keyword overlap\n  - Results merged into search: boosts existing matches + adds new discoveries\n\nPhase 2: Graph Expansion\n  - Top search results used as seed nodes\n  - 4-signal relevance model finds related pages\n  - 2-hop traversal with decay for deeper connections\n\nPhase 3: Budget Control\n  - Configurable context window: 4K → 1M tokens\n  - Proportional allocation: 60% wiki pages, 20% chat history, 5% index, 15% system\n  - Pages prioritized by combined search + graph relevance score\n\nPhase 4: Context Assembly\n  - Numbered pages with full content (not just summaries)\n  - System prompt includes: purpose.md, language rules, citation format, index.md\n  - LLM instructed to cite pages by number: [1], [2], etc.\n```\n\n**Vector Search** is fully optional — disabled by default, enabled in Settings with independent endpoint, API key, and model configuration. When disabled, the pipeline falls back to tokenized search + graph expansion. Benchmark: overall recall improved from 58.2% to 71.4% with vector search enabled.\n\n### 8. Multi-Conversation Chat with Persistence\n\nThe original has a single query interface. We built **full multi-conversation support**:\n\n- **Independent chat sessions** — create, rename, delete conversations\n- **Conversation sidebar** — quick switching between topics\n- **Per-conversation persistence** — each conversation saved to `.llm-wiki\u002Fchats\u002F{id}.json`\n- **Configurable history depth** — limit how many messages are sent as context (default: 10)\n- **Cited references panel** — collapsible section on each response showing which wiki pages were used, grouped by type with icons\n- **Reference persistence** — cited pages stored directly in message data, stable across restarts\n- **Regenerate** — re-generate the last response with one click (removes last assistant + user message pair, re-sends)\n- **Save to Wiki** — archive valuable answers to `wiki\u002Fqueries\u002F`, then auto-ingest to extract entities\u002Fconcepts into the knowledge network\n\n### 9. Thinking \u002F Reasoning Display\n\nNot in the original. For LLMs that emit `\u003Cthink>` blocks (DeepSeek, QwQ, etc.):\n\n- **Streaming thinking** — rolling 5-line display with opacity fade during generation\n- **Collapsed by default** — thinking blocks hidden after completion, click to expand\n- **Visual separation** — thinking content shown in distinct style, separate from the main response\n\n### 10. KaTeX Math Rendering\n\nNot in the original. Full LaTeX math support across all views:\n\n- **KaTeX rendering** — inline `$...$` and block `$$...$$` formulas rendered via remark-math + rehype-katex\n- **Milkdown math plugin** — preview editor renders math natively via @milkdown\u002Fplugin-math\n- **Auto-detection** — bare `\\begin{aligned}` and other LaTeX environments automatically wrapped with `$$` delimiters\n- **Unicode fallback** — 100+ symbol mappings (α, ∑, →, ≤, etc.) for simple inline notation outside math blocks\n\n### 11. Review System (Async Human-in-the-Loop)\n\nThe original suggests staying involved during ingest. We added an **asynchronous review queue**:\n\n- LLM flags items needing human judgment during ingest\n- **Predefined action types**: Create Page, Deep Research, Skip — constrained to prevent LLM hallucination of arbitrary actions\n- **Search queries generated at ingest time** — LLM pre-generates optimized web search queries for each review item\n- User handles reviews at their convenience — doesn't block ingest\n\n### 12. Deep Research\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002F1-deepresearch.jpg\" width=\"100%\" alt=\"Deep Research\">\n\u003C\u002Fp>\n\nNot in the original. When the LLM identifies knowledge gaps:\n\n- **Web search** (Tavily API) finds relevant sources with full content extraction (no truncation)\n- **Multiple search queries** per topic — LLM-generated at ingest time, optimized for search engines\n- **LLM-optimized research topics** — when triggered from Graph Insights, LLM reads overview.md + purpose.md to generate domain-specific topics and queries (not generic keywords)\n- **User confirmation dialog** — editable topic and search queries shown for review before research starts\n- **LLM synthesizes** findings into a wiki research page with cross-references to existing wiki\n- **Thinking display** — `\u003Cthink>` blocks shown as collapsible sections during synthesis, auto-scroll to latest content\n- **Auto-ingest** — research results automatically processed to extract entities\u002Fconcepts into the wiki\n- **Task queue** with 3 concurrent tasks\n- **Research Panel** — dedicated sidebar panel with dynamic height, real-time streaming progress\n\n### 13. Browser Extension (Web Clipper)\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002F4-chrome_extension_webclipper.jpg\" width=\"100%\" alt=\"Chrome Extension Web Clipper\">\n\u003C\u002Fp>\n\nThe original mentions Obsidian Web Clipper. We built a **dedicated Chrome Extension** (Manifest V3):\n\n- **Mozilla Readability.js** for accurate article extraction (strips ads, nav, sidebars)\n- **Turndown.js** for HTML → Markdown conversion with table support\n- **Project picker** — choose which wiki to clip into (supports multi-project)\n- **Local HTTP API** (port 19827, tiny_http) — Extension ↔ App communication\n- **Auto-ingest** — clipped content automatically triggers the two-step ingest pipeline\n- **Clip watcher** — polls every 3 seconds for new clips, processes automatically\n- **Offline preview** — shows extracted content even when app is not running\n\n### 14. Multi-format Document Support\n\nThe original focuses on text\u002Fmarkdown. We support structured extraction preserving document semantics:\n\n| Format | Method |\n|--------|--------|\n| PDF | pdf-extract (Rust) with file caching |\n| DOCX | docx-rs — headings, bold\u002Fitalic, lists, tables → structured Markdown |\n| PPTX | ZIP + XML — slide-by-slide extraction with heading\u002Flist structure |\n| XLSX\u002FXLS\u002FODS | calamine — proper cell types, multi-sheet support, Markdown tables |\n| Images | Native preview (png, jpg, gif, webp, svg, etc.) |\n| Video\u002FAudio | Built-in player |\n| Web clips | Readability.js + Turndown.js → clean Markdown |\n\n### 15. File Deletion with Cascade Cleanup\n\nThe original has no deletion mechanism. We added **intelligent cascade deletion**:\n\n- Deleting a source file removes its wiki summary page\n- **3-method matching** finds related wiki pages: frontmatter `sources[]` field, source summary page name, frontmatter section references\n- **Shared entity preservation** — entity\u002Fconcept pages linked to multiple sources only have the deleted source removed from their `sources[]` array, not deleted entirely\n- **Index cleanup** — removed pages are purged from index.md\n- **Wikilink cleanup** — dead `[[wikilinks]]` to deleted pages are removed from remaining wiki pages\n\n### 16. Configurable Context Window\n\nNot in the original. Users can configure how much context the LLM receives:\n\n- **Slider from 4K to 1M tokens** — adapts to different LLM capabilities\n- **Proportional budget allocation** — larger windows get proportionally more wiki content\n- **60\u002F20\u002F5\u002F15 split** — wiki pages \u002F chat history \u002F index \u002F system prompt\n\n### 17. Cross-Platform Compatibility\n\nThe original is platform-agnostic (abstract pattern). We handle concrete cross-platform concerns:\n\n- **Path normalization** — unified `normalizePath()` used across 22+ files, backslash → forward slash\n- **Unicode-safe string handling** — char-based slicing instead of byte-based (prevents crashes on CJK filenames)\n- **macOS close-to-hide** — close button hides window (app stays running in background), click dock icon to restore, Cmd+Q to quit\n- **Windows\u002FLinux close confirmation** — confirmation dialog before quitting to prevent accidental data loss\n- **Tauri v2** — native desktop on macOS, Windows, Linux\n- **GitHub Actions CI\u002FCD** — automated builds for macOS (ARM + Intel), Windows (.msi), Linux (.deb \u002F .AppImage)\n\n### 18. Other Additions\n\n- **i18n** — English + Chinese interface (react-i18next)\n- **Settings persistence** — LLM provider, API key, model, context size, language saved via Tauri Store\n- **Obsidian config** — auto-generated `.obsidian\u002F` directory with recommended settings\n- **Markdown rendering** — GFM tables with borders, proper code blocks, wikilink processing in chat and preview\n- **Multi-provider LLM support** — OpenAI, Anthropic, Google, Ollama, Custom — each with provider-specific streaming and headers\n- **15-minute timeout** — long ingest operations won't fail prematurely\n- **dataVersion signaling** — graph and UI automatically refresh when wiki content changes\n\n## Tech Stack\n\n| Layer | Technology |\n|-------|-----------|\n| Desktop | Tauri v2 (Rust backend) |\n| Frontend | React 19 + TypeScript + Vite |\n| UI | shadcn\u002Fui + Tailwind CSS v4 |\n| Editor | Milkdown (ProseMirror-based WYSIWYG) |\n| Graph | sigma.js + graphology + ForceAtlas2 |\n| Search | Tokenized search + graph relevance + optional vector (LanceDB) |\n| Vector DB | LanceDB (Rust, embedded, optional) |\n| PDF | pdf-extract |\n| Office | docx-rs + calamine |\n| i18n | react-i18next |\n| State | Zustand |\n| LLM | Streaming fetch (OpenAI, Anthropic, Google, Ollama, Custom) |\n| Web Search | Tavily API |\n\n## Installation\n\n### Pre-built Binaries\n\nDownload from [Releases](https:\u002F\u002Fgithub.com\u002Fnashsu\u002Fllm_wiki\u002Freleases):\n- **macOS**: `.dmg` (Apple Silicon + Intel)\n- **Windows**: `.msi`\n- **Linux**: `.deb` \u002F `.AppImage`\n\n### Build from Source\n\n```bash\n# Prerequisites: Node.js 20+, Rust 1.70+\ngit clone https:\u002F\u002Fgithub.com\u002Fnashsu\u002Fllm_wiki.git\ncd llm_wiki\nnpm install\nnpm run tauri dev      # Development\nnpm run tauri build    # Production build\n```\n\n### Chrome Extension\n\n1. Open `chrome:\u002F\u002Fextensions`\n2. Enable \"Developer mode\"\n3. Click \"Load unpacked\"\n4. Select the `extension\u002F` directory\n\n## Quick Start\n\n1. Launch the app → Create a new project (choose a template)\n2. Go to **Settings** → Configure your LLM provider (API key + model)\n3. Go to **Sources** → Import documents (PDF, DOCX, MD, etc.)\n4. Watch the **Activity Panel** — LLM automatically builds wiki pages\n5. Use **Chat** to query your knowledge base\n6. Browse the **Knowledge Graph** to see connections\n7. Check **Review** for items needing your attention\n8. Run **Lint** periodically to maintain wiki health\n\n## Project Structure\n\n```\nmy-wiki\u002F\n├── purpose.md              # Goals, key questions, research scope\n├── schema.md               # Wiki structure rules, page types\n├── raw\u002F\n│   ├── sources\u002F            # Uploaded documents (immutable)\n│   └── assets\u002F             # Local images\n├── wiki\u002F\n│   ├── index.md            # Content catalog\n│   ├── log.md              # Operation history\n│   ├── overview.md         # Global summary (auto-updated)\n│   ├── entities\u002F           # People, organizations, products\n│   ├── concepts\u002F           # Theories, methods, techniques\n│   ├── sources\u002F            # Source summaries\n│   ├── queries\u002F            # Saved chat answers + research\n│   ├── synthesis\u002F          # Cross-source analysis\n│   └── comparisons\u002F        # Side-by-side comparisons\n├── .obsidian\u002F              # Obsidian vault config (auto-generated)\n└── .llm-wiki\u002F              # App config, chat history, review items\n```\n\n## Star History\n\n\u003Ca href=\"https:\u002F\u002Fwww.star-history.com\u002F?repos=nashsu%2Fllm_wiki&type=date&legend=top-left\">\n \u003Cpicture>\n   \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Fapi.star-history.com\u002Fchart?repos=nashsu\u002Fllm_wiki&type=date&theme=dark&legend=top-left\" \u002F>\n   \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Fapi.star-history.com\u002Fchart?repos=nashsu\u002Fllm_wiki&type=date&legend=top-left\" \u002F>\n   \u003Cimg alt=\"Star History Chart\" src=\"https:\u002F\u002Fapi.star-history.com\u002Fchart?repos=nashsu\u002Fllm_wiki&type=date&legend=top-left\" \u002F>\n \u003C\u002Fpicture>\n\u003C\u002Fa>\n\n## License\n\nThis project is licensed under the **GNU General Public License v3.0** — see [LICENSE](LICENSE) for details.\n","LLM Wiki 是一款跨平台的桌面应用程序，能够自动将您的文档转换为有组织且相互链接的知识库。其核心功能包括通过两步链式思维分析和生成带有源可追溯性的wiki页面、多模态图像提取与处理、基于四信号的知识图谱构建以及矢量语义搜索等。该应用使用TypeScript编写，支持文件夹导入、深度研究及异步审核系统等功能，旨在提供一个持久化且不断更新的知识管理系统。适用于需要管理和利用大量文档资源的研究人员、学生或任何希望高效组织个人知识的人士。",2,"2026-06-11 02:38:08","CREATED_QUERY"]