[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-1480":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":12,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":14,"stars7d":14,"stars30d":14,"stars90d":14,"forks30d":14,"starsTrendScore":14,"compositeScore":15,"rankGlobal":10,"rankLanguage":10,"license":16,"archived":17,"fork":17,"defaultBranch":18,"hasWiki":19,"hasPages":17,"topics":20,"createdAt":10,"pushedAt":10,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":14,"starSnapshotCount":14,"syncStatus":28,"lastSyncTime":29,"discoverSource":30},1480,"llm-wiki","mduongvandinh\u002Fllm-wiki","mduongvandinh","Hệ thống knowledge base cá nhân hoàn toàn tự động, vận hành bởi LLM. Dựa trên pattern LLM Wiki của Andrej Karpathy.","",null,"HTML",211,94,0,45.93,"MIT License",false,"main",true,[21,22,23,24],"andrej-karpathy","claude-code","llm","obsidian","2026-06-12 04:00:09","# LLM Wiki\n\n**A personal knowledge base that builds and maintains itself.**\n\nDrop articles, notes, PDFs, or Reddit threads into a folder. Your AI reads everything, builds a structured wiki with cross-references, detects contradictions between sources, and keeps it all up to date — automatically.\n\nYou just read and ask questions.\n\n> Inspired by [Andrej Karpathy's LLM Wiki pattern](https:\u002F\u002Fgist.github.com\u002Fkarpathy\u002F442a6bf555914893e9891c11519de94f). Extended with auto-discovery, Reddit scanning, GitHub tracking, and ready-made variants for common use cases.\n\n---\n\n## Why not just use RAG?\n\n| | Traditional RAG | LLM Wiki |\n|---|---|---|\n| Knowledge | Re-retrieved fresh every query | Compiled once, maintained continuously |\n| Cross-references | None | Built and maintained automatically |\n| Contradictions | Not detected | Flagged immediately on ingest |\n| Compounding | Flat | Grows richer over time |\n| Infrastructure | Vector DB + embeddings | Just folders + markdown |\n\n---\n\n## Quick Start\n\n### Option A — Start with a ready-made variant (recommended)\n\nDon't configure from scratch. Pick a variant that matches your use case — config, schema, and demo data are pre-tuned:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fmduongvandinh\u002Fllm-wiki.git my-wiki\ncd my-wiki\n\n# Install skill (Claude Code only — other tools skip this)\ncp -r skills\u002Fllm-wiki ~\u002F.claude\u002Fskills\u002Fllm-wiki\n\n# Pick your variant:\n\u002Fllm-wiki setup book-companion      # 📚 Reading companion\n\u002Fllm-wiki setup competitive-intel   # 🎯 Track competitors\n\u002Fllm-wiki setup job-search          # 💼 Job search research\n```\n\n### Option B — Start from scratch\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fmduongvandinh\u002Fllm-wiki.git my-wiki\ncd my-wiki\n\ncp config.example.yaml config.yaml\n# Edit config.yaml: set your topics, keywords, feeds\n```\n\nThen drop content into `raw\u002Farticles\u002F` and run your first cycle.\n\n---\n\n## Variants\n\nPre-configured templates for specific use cases. One command setup, includes demo data.\n\n| | Variant | Use Case | Key Command |\n|---|---------|----------|-------------|\n| 📚 | [book-companion](variants\u002Fbook-companion\u002FREADME.md) | Build a companion wiki as you read — characters, timeline, factions, lore auto-extracted | `book-summary` · [demo](demo\u002Fbook-companion\u002F) |\n| 🎯 | [competitive-intel](variants\u002Fcompetitive-intel\u002FREADME.md) | Track competitors, detect pricing\u002Ffeature changes, generate battlecards | `competitive-brief \"Linear\"` |\n| 💼 | [job-search](variants\u002Fjob-search\u002FREADME.md) | Research target companies, prep interview 1-pagers | `interview-prep \"Stripe\"` |\n\n→ [See all variants →](variants\u002FREADME.md)\n\n---\n\n## Using with Your AI Tool\n\nLLM Wiki works with any AI coding assistant. The wiki is just folders and markdown — the AI reads `CLAUDE.md` (or `AGENTS.md`) to understand the schema and rules, then operates on files.\n\n### Claude Code\n\nThe primary supported tool. Has a slash command skill for the best experience.\n\n**Install the skill:**\n```bash\ncp -r skills\u002Fllm-wiki ~\u002F.claude\u002Fskills\u002Fllm-wiki\n```\n\n**Commands:**\n```bash\n\u002Fllm-wiki setup book-companion     # Set up a variant\n\u002Fllm-wiki run                       # Full cycle: discover → ingest → lint\n\u002Fllm-wiki ingest                    # Process new files in raw\u002F\n\u002Fllm-wiki query \"What is X?\"        # Ask questions against the wiki\n\u002Fllm-wiki discover                  # Auto-find new sources from the web\n\u002Fllm-wiki lint                      # Health check: contradictions, orphans, gaps\n\u002Fllm-wiki status                    # Overview of wiki state\n\u002Fllm-wiki digest                    # Daily brief: new sources, insights, changes\n\u002Fllm-wiki book-summary              # Structured book summary (book-companion variant)\n\u002Fllm-wiki competitive-brief \"Name\"  # Battlecard for a competitor\n\u002Fllm-wiki interview-prep \"Company\"  # Interview prep 1-pager\n```\n\n**Auto-run:**\n```bash\n\u002Floop 1h \u002Fllm-wiki run    # Runs every hour in current session\n\u002Floop 2h \u002Fllm-wiki run    # Every 2 hours\n```\n\n---\n\n### Cursor\n\nCursor reads `AGENTS.md` from the project root automatically. No installation needed.\n\n**Open the project in Cursor, then chat:**\n```\nFollow AGENTS.md. Run full cycle: discover new sources, ingest into wiki, then lint.\n```\n```\nFollow AGENTS.md. Ingest everything new in raw\u002F and update the wiki.\n```\n```\nFollow AGENTS.md. Query the wiki: what do we know about [topic]?\n```\n```\nFollow AGENTS.md. Run competitive-brief for \"Linear\" and save to outputs\u002F.\n```\n\n---\n\n### GitHub Copilot (VS Code)\n\nCopilot reads `.github\u002Fcopilot-instructions.md` automatically when you open the project. No installation needed.\n\n**Open project in VS Code, then use Copilot Chat:**\n```\nRead CLAUDE.md then run discover, ingest, and lint for this wiki.\n```\n```\nBased on the wiki schema in CLAUDE.md, ingest the new files in raw\u002Farticles\u002F.\n```\n```\nQuery the wiki: summarize everything we know about [topic].\n```\n\n> **Note:** Copilot Chat doesn't support background loops. Run manually or use system scheduler.\n\n---\n\n### Windsurf (Codeium)\n\nWindsurf reads `AGENTS.md` or `.windsurfrules` from the project root.\n\n**Chat in Windsurf:**\n```\nRead AGENTS.md. Run the full llm-wiki cycle: discover → ingest → lint.\n```\n```\nRead AGENTS.md. I've added new files to raw\u002Farticles\u002F. Please ingest them into the wiki.\n```\n\n---\n\n### Gemini CLI\n\nGemini CLI reads `GEMINI.md` from the project root (or falls back to `AGENTS.md`).\n\n**Create `GEMINI.md`** (copy from `AGENTS.md`):\n```bash\ncp AGENTS.md GEMINI.md\n```\n\n**Then run:**\n```bash\ngemini \"Read GEMINI.md and run full cycle: discover, ingest, lint\"\ngemini \"Read GEMINI.md and query the wiki: what do we know about [topic]?\"\n```\n\n---\n\n### Codex CLI (OpenAI)\n\nCodex reads `AGENTS.md` from the project root.\n\n```bash\ncodex \"Read AGENTS.md and run full cycle: discover → ingest → lint\"\ncodex \"Read AGENTS.md and ingest new files in raw\u002F\"\ncodex \"Read AGENTS.md and query: summarize the wiki on [topic]\"\n```\n\n---\n\n### Aider\n\nAider works best with explicit file references.\n\n```bash\naider --message \"Read CLAUDE.md to understand the schema. Run ingest on all new files in raw\u002Farticles\u002F. Update wiki\u002F accordingly.\" CLAUDE.md raw\u002Farticles\u002F*.md\n```\n\n---\n\n### Continue.dev\n\nAdd `CLAUDE.md` to Continue's context, then chat:\n```\n@CLAUDE.md Ingest the new articles in raw\u002F and update the wiki with entities, concepts, and cross-references.\n```\n\n---\n\n### Tool Comparison\n\n| Tool | Config File | Skill\u002FSlash Command | Auto-loop | Best For |\n|------|------------|-------------------|-----------|----------|\n| Claude Code | `CLAUDE.md` | ✅ `\u002Fllm-wiki` | ✅ `\u002Floop` | Full experience, all features |\n| Cursor | `AGENTS.md` | ❌ chat only | ❌ | IDE-integrated workflow |\n| GitHub Copilot | `.github\u002Fcopilot-instructions.md` | ❌ chat only | ❌ | VS Code users |\n| Windsurf | `AGENTS.md` | ❌ chat only | ❌ | VS Code alternative |\n| Gemini CLI | `GEMINI.md` | ❌ chat only | via cron | Google ecosystem |\n| Codex CLI | `AGENTS.md` | ❌ chat only | via cron | OpenAI preference |\n| Aider | `CLAUDE.md` | ❌ explicit | via cron | Terminal-only |\n\n> All tools use the same `raw\u002F`, `wiki\u002F`, `outputs\u002F` folder structure. The AI config file tells each tool how to operate — same content, different filenames.\n\n---\n\n## Auto-Run Without Claude Code\n\nFor tools without `\u002Floop`, use your system scheduler.\n\n**Windows — Task Scheduler** (run once as Admin):\n```powershell\npowershell -ExecutionPolicy Bypass -File \"scripts\u002Fsetup-scheduler.ps1\"\n```\n\n**macOS \u002F Linux — crontab:**\n```bash\n# Claude Code (headless)\n0 *\u002F2 * * * cd \u002Fpath\u002Fto\u002Fmy-wiki && claude --print \"\u002Fllm-wiki run\" >> outputs\u002Fauto-run.log 2>&1\n\n# Codex\n0 *\u002F2 * * * cd \u002Fpath\u002Fto\u002Fmy-wiki && codex \"Read AGENTS.md and run full cycle\" >> outputs\u002Fauto-run.log 2>&1\n\n# Gemini CLI\n0 *\u002F2 * * * cd \u002Fpath\u002Fto\u002Fmy-wiki && gemini \"Read GEMINI.md and run full cycle\" >> outputs\u002Fauto-run.log 2>&1\n```\n\n---\n\n## How It Works\n\n```\n┌─────────────────────────────────────────────────┐\n│              YOU  (read & ask)                   │\n├──────────────┬──────────────┬────────────────────┤\n│   wiki\u002F      │   outputs\u002F   │   wiki-viewer.html │\n│  (Obsidian)  │  (reports)   │   (browser view)   │\n├──────────────┴──────────────┴────────────────────┤\n│              LLM  (writes & maintains)            │\n├──────────┬──────────┬──────────┬─────────────────┤\n│ discover │  ingest  │  query   │      lint        │\n│ (find)   │ (process)│ (answer) │  (health check)  │\n├──────────┴──────────┴──────────┴─────────────────┤\n│   raw\u002F        │  config.yaml   │   CLAUDE.md      │\n│  (sources)    │  (settings)    │   (schema\u002Frules) │\n└───────────────┴────────────────┴──────────────────┘\n```\n\n**Main flow:** `discover` → `raw\u002F` → `ingest` → `wiki\u002F` → `query` → `outputs\u002F`\n\n**Self-improvement loop:** `lint` finds gaps → `discover` finds sources → `ingest` fills them → wiki grows richer\n\n### Core Workflows\n\n**`discover`** — Searches the web for new sources based on your topics in `config.yaml`. Can scan Reddit for pain points, track GitHub trending repos, follow RSS feeds, and snowball from references found in existing sources.\n\n**`ingest`** — Reads each new file in `raw\u002F`. Extracts entities, concepts, and key claims. Creates wiki pages. Adds cross-references. Flags contradictions with existing content. One source typically creates 5–15 wiki pages.\n\n**`query`** — Reads `wiki\u002FINDEX.md`, finds relevant pages, synthesizes an answer with citations. Valuable answers are automatically saved as synthesis pages.\n\n**`lint`** — Scans the entire wiki for: contradictions, orphan pages, broken links, stale claims, and knowledge gaps. Updates `.discoveries\u002Fgaps.json` so `discover` knows what to find next.\n\n---\n\n## Adding Sources\n\n### Drop a file manually\n\nCreate a markdown file in `raw\u002Farticles\u002F` with frontmatter, then run `ingest`:\n\n```markdown\n---\ntitle: \"Article Title\"\nurl: \"https:\u002F\u002Fexample.com\u002Farticle\"\nauthor: \"Author Name\"\ndiscovered: 2026-04-14\ntopic: \"your topic\"\n---\n\nArticle content here...\n```\n\nFor **book notes** (book-companion variant), add `chapter: N`:\n```markdown\n---\ntitle: \"Dune Chapter 5 Notes\"\nchapter: 5\nbook: \"Dune\"\n---\n```\n\nFor **competitor data** (competitive-intel variant), add `competitor: \"Name\"`:\n```markdown\n---\ntitle: \"Linear — Pricing Page\"\ncompetitor: \"Linear\"\n---\n```\n\n### Obsidian Web Clipper\n\n1. Install [Obsidian Web Clipper](https:\u002F\u002Fobsidian.md\u002Fclipper) browser extension\n2. See an interesting article → click clip → save to `raw\u002Farticles\u002F`\n3. Run ingest\n\n### Auto-discovery\n\nConfigure `config.yaml` with topics, Reddit subreddits, GitHub orgs, and RSS feeds — the `discover` command handles the rest.\n\n---\n\n## Viewing Your Wiki\n\n**Obsidian** (recommended): `File → Open vault → select your wiki folder` → Graph View shows connections\n\n**Browser**: Open `wiki-viewer.html` — mobile-first dark theme dashboard with graph view\n\n**Any markdown editor**: All wiki pages are plain `.md` files\n\n---\n\n## Folder Structure\n\n```\nmy-wiki\u002F\n├── CLAUDE.md                    ← Schema + rules for Claude Code\n├── AGENTS.md                    ← Schema + rules for Cursor, Codex, Windsurf, Aider\n├── GEMINI.md                    ← Schema + rules for Gemini CLI (copy of AGENTS.md)\n├── .github\u002F\n│   └── copilot-instructions.md  ← Schema for GitHub Copilot\n├── config.yaml                  ← Your topics, feeds, schedule\n├── config.example.yaml          ← Template to start from\n├── wiki-viewer.html             ← Browser-based wiki viewer\n│\n├── variants\u002F                    ← Ready-made use case templates\n│   ├── book-companion\u002F          ← 📚 Reading companion\n│   ├── competitive-intel\u002F       ← 🎯 Competitor tracking\n│   └── job-search\u002F              ← 💼 Job search research\n│\n├── raw\u002F                         ← YOUR SOURCES (LLM never modifies these)\n│   ├── articles\u002F                ← Web articles, blog posts, papers\n│   ├── reddit\u002F                  ← Reddit threads, pain points\n│   ├── notes\u002F                   ← Personal notes\n│   └── media\u002F                   ← Screenshots, diagrams\n│\n├── wiki\u002F                        ← LLM-maintained wiki (don't edit manually)\n│   ├── INDEX.md                 ← Master catalog of all wiki pages\n│   ← LOG.md                     ← Activity log\n│   ├── entities\u002F                ← People, orgs, tools, projects\n│   ├── concepts\u002F                ← Ideas, patterns, methodologies\n│   ├── sources\u002F                 ← Summary of each raw source\n│   └── syntheses\u002F               ← Analysis, comparisons, rankings\n│\n├── outputs\u002F                     ← Query results, reports, battlecards, briefs\n├── .discoveries\u002F                ← Auto-discovery metadata\n│   ├── feeds.json               ← Sources being tracked\n│   ├── gaps.json                ← Knowledge gaps to fill\n│   └── history.json             ← Processed sources (dedup)\n│\n└── skills\u002Fllm-wiki\u002FSKILL.md     ← Claude Code skill definition\n```\n\n---\n\n## Configuration\n\nAll behavior is controlled by `config.yaml`. Copy `config.example.yaml` to get started.\n\nKey sections:\n\n```yaml\nwiki:\n  name: \"My Wiki\"\n  language: \"en\"   # Any language — wiki content will be in this language\n\ntopics:\n  - name: \"Topic Name\"\n    keywords: [\"keyword1\", \"keyword2\"]\n    priority: high\n\nfeeds:\n  reddit:\n    enabled: true\n    subreddits_pain_points:\n      - sub: \"r\u002FSaaS\"\n  hackernews:\n    enabled: true\n    min_score: 50\n  github_trending:\n    enabled: true\n    languages: [\"python\", \"typescript\"]\n\nschedule:\n  run:\n    loop_interval: \"1h\"   # For \u002Floop in Claude Code\n```\n\n→ See `config.example.yaml` for the full reference with all options.\n\n---\n\n## Commands Reference\n\n| Command | Description |\n|---------|-------------|\n| `setup \u003Cvariant>` | Initialize wiki from a variant template (book-companion, competitive-intel, job-search) |\n| `init \"Topic\"` | Add a new topic to an existing wiki |\n| `ingest` | Process all new files in `raw\u002F` |\n| `query \"...\"` | Answer a question using the wiki; saves valuable answers |\n| `discover` | Auto-find new sources from the web based on topics |\n| `run` | Full cycle: discover → ingest → lint |\n| `lint` | Health check: contradictions, orphans, gaps, stale content |\n| `status` | Show wiki stats and health overview |\n| `digest` | Daily brief: new sources, top insights, detected changes |\n| `pain-rank` | Rank pain points by business opportunity (Reddit data) |\n| `book-summary` | Structured book summary: cast, timeline, themes, mysteries |\n| `competitive-brief \"Name\"` | Battlecard 1-pager for a specific competitor |\n| `interview-prep \"Company\"` | Interview prep 1-pager for a target company |\n\nFor tools without slash commands, use the equivalent natural language prompt:\n```\nRead CLAUDE.md (or AGENTS.md). Then [describe what you want].\n```\n\n---\n\n## FAQ\n\n**What AI model do I need?**\nAny capable model with tool use (web search, file read\u002Fwrite). Claude Sonnet\u002FOpus recommended. Works with GPT-4o, Gemini 1.5 Pro, and others via their respective CLI tools.\n\n**How much does it cost to run?**\nA typical `\u002Fllm-wiki run` cycle (discover + ingest 5 sources + lint) costs $0.10–0.40 depending on wiki size and source length. Running every 2 hours = ~$50\u002Fmonth.\n\n**My wiki is getting large. How do I query it efficiently?**\nThe skill reads `wiki\u002FINDEX.md` first to find relevant pages, then reads only those. Even large wikis (500+ pages) query efficiently because you never load everything at once.\n\n**Can I use this for a team?**\nYes — put the wiki folder in a shared Git repo. Each team member runs ingest on their own sources; the wiki is a shared output. Conflicts are resolved by `lint`.\n\n**Does it hallucinate?**\nThe query and ingest commands only use information from your `raw\u002F` sources. They cannot add claims not supported by your sources. If something is uncertain, the wiki page will say so.\n\n**Can I write in a language other than English?**\nYes. Set `wiki.language` in `config.yaml`. The LLM will write all wiki content in that language while keeping file names and frontmatter in English.\n\n---\n\n## Credits\n\n- Original pattern: [Andrej Karpathy — LLM Wiki](https:\u002F\u002Fgist.github.com\u002Fkarpathy\u002F442a6bf555914893e9891c11519de94f)\n- Inspiration: [Nick Spisak — How to Build Your Second Brain](https:\u002F\u002Fx.com\u002FNickSpisak_\u002Fstatus\u002F2040448463540830705)\n- Built for: [Claude Code](https:\u002F\u002Fclaude.ai\u002Fcode) (Anthropic) + compatible with all major AI coding assistants\n- Wiki viewer: [Obsidian](https:\u002F\u002Fobsidian.md\u002F)\n\n## License\n\nMIT — use it, fork it, build variants, share back.\n","LLM Wiki 是一个基于大型语言模型自动构建和维护的个人知识库系统。它能够自动读取用户放入文件夹中的文章、笔记、PDF 或 Reddit 帖子，生成结构化的维基文档，并自动创建交叉引用，检测来源间的矛盾，持续更新内容。该系统支持自动发现新信息源、扫描 Reddit 和跟踪 GitHub 项目，并为常见应用场景提供了预配置模板。适用于需要整理大量阅读材料、跟踪竞争对手动态或进行求职研究等场景，用户只需简单配置即可开始使用。",2,"2026-06-11 02:44:01","CREATED_QUERY"]