[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-1862":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":8,"htmlUrl":8,"language":8,"languages":8,"totalLinesOfCode":8,"stars":9,"forks":10,"watchers":11,"openIssues":12,"contributorsCount":13,"subscribersCount":13,"size":13,"stars1d":13,"stars7d":11,"stars30d":14,"stars90d":13,"forks30d":13,"starsTrendScore":13,"compositeScore":15,"rankGlobal":8,"rankLanguage":8,"license":8,"archived":16,"fork":16,"defaultBranch":17,"hasWiki":18,"hasPages":16,"topics":19,"createdAt":8,"pushedAt":8,"updatedAt":20,"readmeContent":21,"aiSummary":22,"trendingCount":13,"starSnapshotCount":13,"syncStatus":23,"lastSyncTime":24,"discoverSource":25},1862,"llm-wiki-karpathy","balukosuri\u002Fllm-wiki-karpathy","balukosuri",null,178,76,3,1,0,25,5.66,false,"main",true,[],"2026-06-12 02:00:33","# LLM Wiki (Karpathy Pattern)\n\nA self-maintaining personal knowledge base powered by LLMs, based on [Andrej Karpathy's LLM Wiki pattern](https:\u002F\u002Fgist.github.com\u002Fkarpathy\u002F1dd0294ef9567971c1e4348a90d69285).\n\nInstead of re-searching raw documents on every question (like RAG), the LLM **reads your sources once and builds a persistent, interlinked wiki** that compounds over time. The more sources you feed it, the richer and more connected it gets.\n\n---\n\n## Prerequisites\n\n- [Cursor](https:\u002F\u002Fcursor.sh\u002F) (or any LLM-powered editor that reads a schema file)\n- [Obsidian](https:\u002F\u002Fobsidian.md\u002F) (free) for browsing the wiki in real time\n\n---\n\n## Getting Started\n\n### 1. Clone the repo\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fbalukosuri\u002Fllm-wiki-karpathy.git\ncd llm-wiki-karpathy\n```\n\n### 2. Open the project in Cursor\n\nCursor reads `CLAUDE.md` automatically and understands the wiki's structure, page formats, and workflows.\n\nIf you use a different AI agent (Claude Code, Codex, etc.), paste the contents of `CLAUDE.md` into your agent's context.\n\n### 3. Open the same folder in Obsidian\n\nOpen the project directory as an Obsidian vault. You'll have two windows side by side — Cursor on the left where you talk to the AI, Obsidian on the right where you browse the wiki as pages appear.\n\n### 4. Drop a source into `raw\u002F`\n\nAny document works:\n\n- Product specs, design docs, or PRDs\n- Meeting transcripts\n- Web articles (use [Obsidian Web Clipper](https:\u002F\u002Fobsidian.md\u002Fclipper) to save pages as markdown)\n- Style guides\n- PDFs, reports, email threads saved as text\n- Competitor documentation\n\n### 5. Say \"ingest\"\n\nType this in Cursor:\n\n> ingest raw\u002Fmy-document.pdf\n\nThe AI will:\n\n1. Read the document\n2. Discuss key takeaways with you\n3. Create a source summary page in `wiki\u002Fsources\u002F`\n4. Create new pages for any products, features, personas, or concepts it finds\n5. Update the glossary with new terms\n6. Update the index with all new pages\n7. Update the overview if the big picture shifted\n8. Log everything in `wiki\u002Flog.md`\n\nA single source can touch 5-15 wiki pages. Watch them appear in Obsidian in real time.\n\n### 6. Ask questions\n\n> What are the main risks identified across all my sources?\n\nThe AI reads the wiki, synthesizes an answer with citations, and asks: *\"Should I save this as a wiki page?\"* If you say yes, the answer becomes a permanent analysis page. Your questions make the knowledge base richer over time.\n\n### 7. Lint the wiki\n\nEvery 10 ingests or so, run a health check:\n\n> lint the wiki\n\nThe AI checks for contradictions between pages, stale claims, orphan pages with no links, missing cross-references, and inconsistent terminology. It reports what it found and asks which fixes to apply.\n\n---\n\n## Repo Structure\n\n```\nllm-wiki-karpathy\u002F\n├── CLAUDE.md          # Schema — the AI's operating manual\n├── llm-wiki.md        # Karpathy's original idea document\n├── article.md         # Walkthrough article explaining this project\n│\n├── raw\u002F               # Your source documents (AI reads, never writes)\n│   └── .gitkeep\n│\n├── wiki\u002F              # AI-generated knowledge base (AI owns this layer)\n│   ├── index.md       # Master catalog — the AI reads this first on every query\n│   ├── overview.md    # Big-picture synthesis (evolves with each ingest)\n│   ├── glossary.md    # Terms, definitions, and style conventions\n│   └── log.md         # Chronological record of all activity\n│\n└── .obsidian\u002F         # Pre-configured Obsidian vault settings\n```\n\n### How the layers work\n\n| Layer | Folder | Who owns it | Purpose |\n|-------|--------|-------------|---------|\n| Raw sources | `raw\u002F` | You | Immutable source documents. The AI reads from here but never modifies anything. |\n| The wiki | `wiki\u002F` | The AI | Structured, interlinked markdown pages. The AI creates, updates, and maintains everything here. |\n| The schema | `CLAUDE.md` | You + AI | Defines page types, workflows, and conventions. Edit this to customize the AI's behavior for your domain. |\n\n### Wiki page types\n\nThe AI creates different page types depending on what it finds in your sources:\n\n| Type | Location | What it captures |\n|------|----------|-----------------|\n| Source | `wiki\u002Fsources\u002F` | Summary of a raw document — key facts, quotes, metadata |\n| Feature | `wiki\u002Ffeatures\u002F` | A product feature — what it does, how it works |\n| Product | `wiki\u002Fproducts\u002F` | A product or tool — overview, versions, related features |\n| Persona | `wiki\u002Fpersonas\u002F` | A user type — goals, pain points, expertise level |\n| Concept | `wiki\u002Fconcepts\u002F` | A domain idea — definition, related terms, common misconceptions |\n| Style Rule | `wiki\u002Fstyle\u002F` | A writing convention — when to apply, examples, exceptions |\n| Analysis | `wiki\u002Fanalyses\u002F` | A synthesized output — comparison table, gap analysis, outline |\n\n---\n\n## Customizing for Your Domain\n\nThe schema file `CLAUDE.md` is not set in stone. Edit it to fit your needs:\n\n- **Add new page types.** If your domain needs \"API endpoints\" or \"customer segments\" or \"recipe variations\", add them to the entity types table and tell the AI.\n- **Change the ingest workflow.** If you want the AI to skip the discussion step and just process silently, update the workflow section.\n- **Adjust output formats.** The AI can produce markdown pages, comparison tables, doc outlines, release notes drafts, or persona briefs. Add formats that make sense for your work.\n\n---\n\n## Tips\n\n**Ingest one source at a time.** You can batch-ingest, but you lose the chance to guide the AI. Stay involved — read the summaries, tell it what to emphasize, ask follow-ups during ingestion.\n\n**Save your best questions.** When you ask something and get a useful answer, tell the AI to save it as an analysis page. Your explorations compound in the wiki instead of disappearing into chat history.\n\n**Use graph view often.** Press `Cmd+G` in Obsidian. The visual map shows which pages are hubs, which are orphans, and how everything connects.\n\n**Check the glossary before writing.** Open `wiki\u002Fglossary.md` before you write anything. It has the right terms, the wrong terms, and the reasons behind each choice.\n\n**Don't write wiki pages yourself.** Your job is to find good sources and ask good questions. The AI handles the summarizing, cross-referencing, filing, and bookkeeping.\n\n---\n\n## Use Cases\n\n- **Technical writers** — Ingest specs, transcripts, and competitor docs. Get a living glossary, persona pages, and structured outlines without writing them yourself.\n- **Researchers** — Feed it papers, articles, and reports over weeks. End up with a wiki that has an evolving thesis and all the connections already made.\n- **Product managers** — Ingest PRDs, customer interviews, competitive analyses, and retros. The wiki maintains the big picture.\n- **Students** — Ingest textbook chapters one at a time. The AI builds concept pages, links them together, and flags connections between chapters.\n- **Anyone accumulating knowledge** — Trip planning, hobby research, health tracking, course notes, book clubs. Anything where information comes from multiple sources and you want it organized.\n\n---\n\n## Credits\n\n- Pattern by [Andrej Karpathy](https:\u002F\u002Fgist.github.com\u002Fkarpathy\u002F1dd0294ef9567971c1e4348a90d69285)\n- Implementation and article by [Balu Kosuri](https:\u002F\u002Fgithub.com\u002Fbalukosuri)\n","该项目是一个基于大语言模型（LLM）的自维护个人知识库，采用Andrej Karpathy提出的LLM Wiki模式。其核心功能是通过一次读取源文档并构建一个持久且相互链接的wiki，随着时间推移不断丰富和扩展。技术特点包括使用Cursor等LLM驱动的编辑器理解和管理wiki结构，以及借助Obsidian实现实时浏览与探索。适合需要从大量文档中提炼、组织信息，并持续更新知识体系的场景，如产品开发团队的知识管理、研究项目的信息整理等。",2,"2026-06-11 02:46:28","CREATED_QUERY"]