{"id":50693915,"url":"https://github.com/james-julius/latent-space-explorer","last_synced_at":"2026-06-09T05:05:23.406Z","repository":{"id":361772139,"uuid":"1255760610","full_name":"james-julius/latent-space-explorer","owner":"james-julius","description":"A flythrough 3D map of meaning — type any concept and watch it land near related ideas. In-browser embeddings (no install, no key), pre-seeded scenes, and a multi-model knowledge explorer.","archived":false,"fork":false,"pushed_at":"2026-06-01T07:43:23.000Z","size":353,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-01T08:23:37.388Z","etag":null,"topics":["3d-visualization","ai","claude","data-visualization","embeddings","gemini","latent-space","llm","machine-learning","nextjs","openai","react-three-fiber","semantic-search","text-embeddings","threejs","transformers-js","typescript","umap","vector-search","webgpu"],"latest_commit_sha":null,"homepage":null,"language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/james-julius.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2026-06-01T06:36:21.000Z","updated_at":"2026-06-01T07:43:27.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/james-julius/latent-space-explorer","commit_stats":null,"previous_names":["james-julius/latent-space-explorer"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/james-julius/latent-space-explorer","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/james-julius%2Flatent-space-explorer","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/james-julius%2Flatent-space-explorer/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/james-julius%2Flatent-space-explorer/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/james-julius%2Flatent-space-explorer/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/james-julius","download_url":"https://codeload.github.com/james-julius/latent-space-explorer/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/james-julius%2Flatent-space-explorer/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34092316,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-09T02:00:06.510Z","response_time":63,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["3d-visualization","ai","claude","data-visualization","embeddings","gemini","latent-space","llm","machine-learning","nextjs","openai","react-three-fiber","semantic-search","text-embeddings","threejs","transformers-js","typescript","umap","vector-search","webgpu"],"created_at":"2026-06-09T05:05:22.718Z","updated_at":"2026-06-09T05:05:23.401Z","avatar_url":"https://github.com/james-julius.png","language":"TypeScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n\n# 🌌 Latent Space Explorer\n\n### Fly through a 3D map of meaning.\n\nType any concept and watch it drop into space — right next to the ideas an AI model thinks are similar. The closer two points sit, the closer their meaning. No install, no API key, no sign-up. It runs entirely in your browser.\n\n![Latent Space Explorer — the concept “ocean” selected, with its nearest neighbours coast, continent, ecosystem, sand and island](docs/cover.png)\n\n\u003cp\u003e\n  \u003cimg alt=\"Next.js\" src=\"https://img.shields.io/badge/Next.js-16-000000?logo=nextdotjs\u0026logoColor=white\"\u003e\n  \u003cimg alt=\"React\" src=\"https://img.shields.io/badge/React-19-149eca?logo=react\u0026logoColor=white\"\u003e\n  \u003cimg alt=\"TypeScript\" src=\"https://img.shields.io/badge/TypeScript-5-3178c6?logo=typescript\u0026logoColor=white\"\u003e\n  \u003cimg alt=\"Three.js\" src=\"https://img.shields.io/badge/Three.js-r184-000000?logo=threedotjs\u0026logoColor=white\"\u003e\n  \u003cimg alt=\"transformers.js\" src=\"https://img.shields.io/badge/🤗%20Transformers.js-in--browser-yellow\"\u003e\n\u003c/p\u003e\n\n\u003c/div\u003e\n\n---\n\n## ✨ What is this?\n\nLanguage models turn words into long lists of numbers — **embeddings** — where similar meanings end up close together in a high-dimensional space. That space is impossible to picture… so this project flattens it down to 3D and lets you **fly around inside it**.\n\n- **Type a concept** → it gets embedded and lands near things that mean something similar.\n- **Get close to a point** → related ideas bloom around it, so the map keeps unfolding as you wander.\n- **Watch ideas cluster** → emotions drift one way, animals another, abstract concepts somewhere else entirely.\n\nIt loads pre-filled with a rich cloud of ~600 everyday concepts, so there's something to explore the instant you open it.\n\n## 🚀 Quick start\n\nNo keys, no setup — just clone and run:\n\n```bash\ngit clone https://github.com/james-julius/latent-space-explorer.git\ncd latent-space-explorer\nnpm install\nnpm run dev\n```\n\nOpen **http://localhost:3000** and start flying. The embedding model (~30 MB) downloads once on first load and is cached in your browser forever after.\n\n## 🧠 How it works\n\n```\nyour text ─▶ embedding (bge-small, in-browser) ─▶ 3D projection (UMAP) ─▶ a glowing point\n```\n\n- **Embeddings run on your device.** [`bge-small-en-v1.5`](https://huggingface.co/Xenova/bge-small-en-v1.5) runs locally via [🤗 Transformers.js](https://github.com/huggingface/transformers.js) (WASM/WebGPU). Nothing you type is sent to a server.\n- **Pre-seeded scenes load instantly.** A build-time script embeds a curated concept atlas and precomputes its 3D layout, shipped as a small static file — so the cloud renders before the model even finishes loading.\n- **Your additions stay in the same space.** New points are placed next to their nearest semantic neighbours, growing the map without rearranging it. Everything you build is saved locally in your browser and can be exported/imported as JSON.\n\n## 🤖 Compare how different models think (optional)\n\nBring your own API key for **Claude**, **ChatGPT**, or **Gemini** and let each model suggest related concepts as you explore. Switch on **“by source”** colouring to see, in one shared space, *where the models agree and where they diverge* on what relates to what.\n\nKeys are stored only in your browser and sent only to that provider — never anywhere else.\n\n## ⌨️ Controls\n\n| | |\n|---|---|\n| `W` `A` `S` `D` | move · `E` / `Q` up / down · hold `Shift` to warp |\n| drag | look around |\n| **click a point** | inspect it + reveal its nearest neighbours |\n| `Tab` | jump to the nearest neighbour |\n| `F` / `X` | expand the selected point into related concepts |\n| `B` | bridge — find the conceptual path between two points |\n| `G` | GPS — take a guided tour between two points |\n| `[` / `]` | step back / forward through your trail |\n| `H` | fly home · `D` | dream mode (auto-wander) |\n| `/` | search · `I` | import text or a saved graph · `⌘/Ctrl + Z` undo |\n| `1`–`5` | load a themed preset |\n\n## 🛠️ Tech stack\n\n- **[Next.js 16](https://nextjs.org)** (App Router, Turbopack) + **React 19** + **TypeScript**\n- **[Three.js](https://threejs.org)** via **[@react-three/fiber](https://github.com/pmndrs/react-three-fiber)** \u0026 **drei** for the 3D scene\n- **[Transformers.js](https://github.com/huggingface/transformers.js)** for in-browser embeddings\n- **[UMAP](https://github.com/PAIR-code/umap-js)** for dimensionality reduction\n- **Zustand**-light state, Tailwind CSS\n\n## 🌱 Regenerate the seed scene\n\nThe pre-seeded concept atlas is generated at build time. To edit the concepts or build your own scene, tweak [`scripts/seed-topics.mjs`](scripts/seed-topics.mjs) and run:\n\n```bash\nnpm run seed\n```\n\nThis embeds every term, runs a deterministic UMAP projection, quantizes the vectors, and writes the static scene to `public/scenes/`.\n\n---\n\n\u003cdiv align=\"center\"\u003e\n\u003csub\u003eBuilt to make the invisible geometry of meaning something you can actually fly through. ✦\u003c/sub\u003e\n\u003c/div\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjames-julius%2Flatent-space-explorer","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjames-julius%2Flatent-space-explorer","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjames-julius%2Flatent-space-explorer/lists"}