20:"$Sreact.view_transition" 21:I[9754,["/_next/static/chunks/2t36f5p4pjzlb.js","/_next/static/chunks/1gyqeidjm5eoa.js","/_next/static/chunks/0_fjnxc4tt_l_.js","/_next/static/chunks/1ymt1shyhmu-4.js","/_next/static/chunks/3kv2zf0i7s2wc.js","/_next/static/chunks/1vs_3g_g75zmk.js","/_next/static/chunks/2g4dh802s8oz3.js","/_next/static/chunks/1bc3f0shukf1j.js","/_next/static/chunks/192kgs9v_t14j.js","/_next/static/chunks/0z6pw0v5c_lx5.js","/_next/static/chunks/08r60e6icmtkc.js","/_next/static/chunks/3trx6c0i7_x-3.js","/_next/static/chunks/40sro5bzp9bi_.js","/_next/static/chunks/1wgf3tk38y6cf.js"],"ProjectDetail"] 22:T78a5, Dataset Viewer · 流式查看超大文件数据集

Dataset Viewer stardustai / dataset-viewer

秒开超大文件，毫秒级搜索。

完全由 AI Agent 构建的桌面数据集查看器。它把 WebDAV、S3、SSH、本地或 Hugging Face 的大文件当作流来读取——100GB 以上的数据集不用等它加载完，一条可搜索的窄带就已经在眼前。

100GB+ 流式 WebDAV · S3 · SSH Hugging Face Tauri / React / TypeScript

查看仓库 → 官网 ↗

FLOW CONDUIT · SOURCE → VIEWPORT THROUGHPUT ▲

主数据流远端源虚拟化缓冲

02 / SOURCES

多协议接入，一条流

不止是打开文件，而是从任意远程或本地数据源把大文件作为流吸入。每个源是管线上的一个节点，进查看视窗前都能被即时预览。

WebDAV

挂载到 web 的远程目录，像读本地盘一样浏览并流式预览其中的大文件。

WEBDAV

S3

对象存储桶里的数据集，无需整包下载，按需拉取目标文件进入查看视窗。

OBJECT STORE

SSH

通过安全的 shell 通道读取远端服务器上的日志与数据文件，边读边看。

SECURE SHELL

Local

直接打开本机文件系统中的文件，作为即时查看的基线源。

FILESYSTEM

Hugging Face

直接从 Hugging Face 上的数据集仓库读取资源并流式预览，而不用先下载全部内容。

HF HUB

03 / WINDOW

虚拟化视窗：只渲染你看到的

文件的尺度可以非常大，但视窗总是刚刚好。越界行不载入内存，滚动即时，搜索毫秒级命中—— 因为无限大的文件被压缩成了一条可搜索的窄带。

train_000 000 012.jsonl VIRTUAL WINDOW

0001"text""samples / 4096"

0002"text""labels updated"

0003····· 跳过的越界行 ·····

0901"text""tokens=2048"

0902"text""done=true"

表与文本，一屏之内

虚拟滚动保证一个巨大的表格或日志文件也能平滑翻页，配合毫秒级搜索，你能跨越数百 GB 直达目标行，而不必在整个数据集里逐段翻找。

文件规模 100GB+

搜索延迟 ms

内存占用仅当前视窗

04 / FORMATS

多种格式，就地解析

常见数据格式不用额外导出或转换，直接拖进视窗即可解析。从表格到归档，从结构化文本到代码，一条管线通吃。

.parquet 列式数据集 · 分块读取

COLUMNAR → STREAM

.xlsx 电子表格 · 工作表

SHEET → CELLS

.csv 行式文本 · 分隔符

ROWS → COLUMNS

.json 嵌套结构 · jsonl

TREE → KEYS

.zip / .tar 归档 · 内直接预览

ARCHIVE → ENTRY

code 源代码 · 高亮阅读

SRC → TOKENS

官方文档以 ZIP/TAR 归档的直接预览、以及代码文件的高亮阅读作为能力示例，覆盖从数据到代码的常见工作流。

05 / EXTEND

用插件扩展新的格式

不是所有格式都值得内置。通过插件开发 SDK 和示例插件，你可以为自定义格式接入自己的查看器分支。

把自己的格式接进来

SDK 暴露查看与解析的钩子，示例插件演示从零接入一个新格式的最小路径。社区与团队可以针对私有格式或特殊工作流构建专属查看器分支。

适用于尚未内置的领域格式
示例插件作为开发的起点
查看器的能力随插件一起生长

06 / LOCAL-FIRST

本地优先，随手可用

基于 Tauri 的桌面应用，安装即是本地工具。多平台可装，重数据留在你的机器上，打开与搜索都在本地完成。

发行

为 Windows、macOS 与 Linux 提供安装包，跨平台安装即用。

v1.6.4WINDOWS

v1.6.4macOS

v1.6.4LINUX

stackTauri · React · TypeScript

如实说明

该项目 100% 由 AI Agent 生成，稳定性请在关键场景自行评估。

插件生态仍处于早期，可用插件与文档相对有限。

项目未提供官方支持渠道，问题排查依赖社区与源码。

Dataset Viewer · stardustai/dataset-viewer · TypeScript

GitHub 官网

23:T73f0, Dataset Viewer · Stream & inspect massive datasets

Dataset Viewer stardustai / dataset-viewer

Open massive files in seconds, search in milliseconds.

A dataset viewer built entirely by an AI agent. It reads large files from WebDAV, S3, SSH, Local, or Hugging Face as streams — a 100GB+ dataset doesn't have to finish loading before a searchable band of it is already on screen.

100GB+ streaming WebDAV · S3 · SSH Hugging Face Tauri / React / TypeScript

View repository → Website ↗

FLOW CONDUIT · SOURCE → VIEWPORT THROUGHPUT ▲

Primary data flow Remote source Virtualized buffer

02 / SOURCES

One stream from many protocols

Rather than just opening a file, it draws large files in as streams from any remote or local source. Each source is a node on the conduit, and everything is previewable before it reaches the viewport.

WebDAV

Remote directories exposed over the web, browsed like a local drive while large files stream into view.

WEBDAV

S3

Object storage buckets, ready to pull the exact file you need without downloading the whole dataset.

OBJECT STORE

SSH

Read logs and data files from a remote server over a secure shell channel, and watch them as they come.

SECURE SHELL

Local

Open files straight from the local filesystem — the baseline source for instant inspection.

FILESYSTEM

Hugging Face

Reach into datasets hosted on Hugging Face and preview resources without staging everything locally first.

HF HUB

03 / WINDOW

A virtual window that renders only what you see

Files can be enormous, but the window stays just right. Out-of-view rows never load into memory, scrolling is instant, and search hits in milliseconds — because an effectively infinite file is compressed into one searchable band.

train_000 000 012.jsonl VIRTUAL WINDOW

0001"text""samples / 4096"

0002"text""labels updated"

0003····· skipped rows ·····

0901"text""tokens=2048"

0902"text""done=true"

Tables and text on a single page

Virtual scrolling keeps a massive table or log smooth to page through, and millisecond search lets you jump straight past hundreds of gigabytes to the exact row — no need to hunt through the whole dataset.

File size 100GB+

Search latency ms

Memory footprint viewport only

04 / FORMATS

Many formats, parsed in place

Common data formats need no extra export step. Drop them into the window and they parse on the spot — from tables to archives, from structured text to source code, all down one pipeline.

.parquet Columnar datasets · chunked reads

COLUMNAR → STREAM

.xlsx Spreadsheets · worksheets

SHEET → CELLS

.csv Delimited row text

ROWS → COLUMNS

.json Nested structures · JSONL

TREE → KEYS

.zip / .tar Archives · preview from within

ARCHIVE → ENTRY

code Source files · syntax-highlighted

SRC → TOKENS

The project highlights direct preview of ZIP/TAR archives and syntax-highlighted code files as capabilities, spanning the common workflow from data to code.

05 / EXTEND

Extend the viewer with plugins

Not every format deserves to be built in. With the plugin SDK and example plugins, you can wire your own parser branch onto the core for a custom format.

Bring your own format

The SDK exposes the hooks for viewing and parsing, and the example plugins show the minimal path to wiring up a new format from scratch. Teams can build dedicated viewer branches for private formats or specialized workflows.

Fits domain formats not yet built in
Example plugins as a starting point
Capabilities grow with the plugin

06 / LOCAL-FIRST

Local-first, ready when you are

A Tauri-based desktop app, installing to a tool that lives on your machine. Multi-platform, your heavy data stays local, and opening and searching all happen right there.

Releases

Installers for Windows, macOS, and Linux — cross-platform, install and go.

v1.6.4WINDOWS

v1.6.4macOS

v1.6.4LINUX

stackTauri · React · TypeScript

Honest notes

This project is 100% AI-generated; evaluate its stability for anything critical.

The plugin ecosystem is still early — available plugins and docs are limited.

There is no official support channel; troubleshooting relies on the community and the source.

Dataset Viewer · stardustai/dataset-viewer · TypeScript

GitHub Website

a:[["$","$20",null,{"enter":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-enter","default":"none"},"exit":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-exit","default":"none"},"update":{"secondary-forward":"secondary-forward","secondary-back":"secondary-back-update","default":"none"},"default":"none","children":["$","$L21","cmsqzq1eb00y9zn19d4793n3d:zh-CN:beautified",{"project":{"id":"cmsqzq1eb00y9zn19d4793n3d","slug":"stardustai-dataset-viewer","name":"dataset-viewer","description":"AI 构建的高性能数据集查看器","authorName":"stardustai","authorVerification":{"verified":false,"verifiedAt":null},"category":"development_tools","subcategory":"data_ai_tools","experienceUrl":"https://stardustai.github.io/dataset-viewer","qualityScore":85,"featured":false,"officialSite":null,"views":{"original":{"available":true},"official":{"available":false,"reason":"disabled_by_policy"},"beautified":{"available":true,"reason":null},"defaultView":"beautified"},"beautifiedView":{"revisionId":"cmt2zh31w01mu6271tzfqcy3d","kind":"artifact","artifact":{"schemaVersion":"localized-static-html-v4","kind":"static_html","locales":["zh-CN","en"],"defaultLocale":"zh-CN","documents":{"zh-CN":"$22","en":"$23"},"designDirections":{"schemaVersion":"dodot-design-directions-v1","directions":[{"id":"conduit-pipeline","name":"The Streaming Conduit","projectMetaphor":"远程数据源是一条条加压管道，把 100GB+ 的文件流送进一个即时打开的查看视窗；查看器不是文件浏览器而是流的中枢分岔点。","signatureVisual":"一条从左向右的动态流管网：若干来自不同源的语义节点（WebDAV/S3/SSH/Local/HF）汇入一条主管线，管线中段切开露出虚拟化的行缓冲窗格，数据在其间流动。","heroGeometry":"横向张力构图：左侧五个源节点以流速各异的管线汇入右侧主查看窗格，一条贯穿的水平流轴承载延迟/吞吐刻度，无限大文件被压缩成一条可搜索的窄带。","sectionSequence":["conduit-hero","sources-routing","virtual-windowing","formats-parsing","plugins-sdk","local-first"],"surfaceModel":"工业级流控表面：亚光深色基材上叠加发光管线层，节点为高对比圆角瓣体，窗格为内凹视窗带刻度，全部层级带可感知的进深次序。","palette":"base #0A0F0E · surface #10181A · ink #E7EFEA · positive #3DDC97 管线 · amber #FFB454 流速负荷 · info #56CCF2 远端源 · muted #7E8C88","typography":"标题用硬朗的等宽强调（JetBrains Mono 式），正文用中性无衬线；协议名与流速标签用等宽小字。","desktopComposition":"1280px 下采用 12 栏：左 6 栏为源-管线带（含实时流速刻度），右 6 栏为查看视窗叠层，管线穿缝连接两栏；下方章节按证据顺序纵向展开。","mobileTransformation":"390px 下管线由横向改为垂直栈：五个源节点自上而下叠加为路由清单，主查看窗格收为单屏视窗，每章横向滚动改为纵向卡片。","mediaStrategy":"无本地合格素材，删除媒体区；以证据驱动的管线-流速度量图解与格式解析排版承担标志视觉，不伪造产品截图。","forbiddenPatterns":["三卡指标横幅","功能卡片墙（可套用任何项目的文案）","浏览器/终端/编辑器外壳冒充产品截图","装饰性发光网格或粒子替代项目视觉","伪造的开关/搜索/图表控件"],"compositionFingerprint":"多源节点管线汇入一个虚拟化查看视窗的横向流控布局，流速刻度贯穿窗格，格式/插件章节以管线分支延伸。"},{"id":"infinite-ruler","name":"The Infinite Ruler","projectMetaphor":"这个查看器把整份巨型文件当成一根无限长的标尺，用户只移动一个可搜索的取景窗口，从不加载整个文件；查看的本质是移动视界而非打开文件。","signatureVisual":"一根横贯页面的长标尺，刻度从文件起始排到不可见尽头，上方一个高亮取景窗沿尺滑动，窗内渲染当前行/字段，窗两侧标尺按距离压缩。","heroGeometry":"极横向构图：一条占据全宽的连续标尺作为主地平线，取景窗悬浮其上，游标与搜索刻度交互，整个首屏是一条被压缩成带状的超长文件。","sectionSequence":["ruler-hero","moving-window","format-lenses","remote-reach","extensibility","honesty-notes"],"surfaceModel":"测绘工具面：标尺与刻度用细线刻蚀，取景窗为悬浮半透明量窗，导航游标为唯一强色点，整体走冷灰金属与量测精度。","palette":"base #0C0E11 · surface #14171C · ink #E8ECEF · primary #7C9FF0 量窗 · accent #E8A87C 游标/搜索 · remote #3F4A55 · muted #89939E","typography":"标题用宽体无衬线压缩字距，正文用细无衬线；刻度、偏移量与格式标签全用等宽数字字体强化量测感。","desktopComposition":"1280px 下为整根横向标尺布局：标尺贯穿全宽作视觉主轴，取景窗与游标在尺上横置，章节随尺再下方纵向排列，形成 T 形总构图。","mobileTransformation":"390px 下全宽标尺转为垂直标尺靠在页面左缘，取景窗收为靠右单列内容，章节纵向堆叠，游标交互改为触控滑距。","mediaStrategy":"无本地合格素材，删除媒体区；以刻度-取景窗的矢量量测图解与行/字段排版承担标志视觉，不伪造产品截图。","forbiddenPatterns":["首屏大标语配三卡指标","浏览器/桌面外壳框住内容","同一圆角/阴影/间距尺度复用每章","发光网格或粒子背景作主视觉","伪造选项卡/图标墙"],"compositionFingerprint":"贯穿全宽的巨型标尺加取景窗的横向量测布局，游标为唯一强色，刻度按距离压缩，章节在尺下纵向排开。"},{"id":"core-orbit","name":"The Orbiting Sources","projectMetaphor":"查看器是数据核心，远程源——WebDAV、S3、SSH、本地、Hugging Face——是围绕它公转的端点；拖动画布即可把任意远端拉近到核心即时预览，而不是逐个打开。","signatureVisual":"一个径向天球图：核心查看器居中，五个源端点在同心轨道上散布，连接线按活跃度加粗，核心外圈露出当前预览视窗的一个切面。","heroGeometry":"中心放射构图：圆形/椭圆视窗居首屏中心，五个源节点沿离心率不同的轨道环绕，指向核心的矢径承担数据到达的隐喻。","sectionSequence":["orbit-hero","endpoint-ring","preview-core","plugin-expansion","integrity-baseline"],"surfaceModel":"天球层叠面：扁平深空基材，多个同心轨道环作为分层容器，核心为实心高亮盘，源节点为漂浮瓣体，切面视窗由内向外翻出。","palette":"base #0B0F1C · surface #131A2B · ink #ECEDF7 · primary #8B7CFF 核心 · accent #FF7EAA 活跃连接 · source #9FB4FF · muted #6E7690","typography":"标题用大号几何无衬线加圆角收笔，正文用中性无衬线；协议名用等宽小标签配轨道半径刻度。","desktopComposition":"1280px 下以超大同心轨道为背景层承担首屏对称，核心盘居左一档、源节点环列右右三档，下方章节从轨道区域垂直泄出。","mobileTransformation":"390px 下同心轨道摊平为递进圆环列表，核心收至顶部，源端点变为可纵向滚动的接入清单，轨道刻度改为进度条样式。","mediaStrategy":"无本地合格素材，删除媒体区；以同心轨道连接图与核心预览切面的矢量表达承担标志视觉，不伪造产品截图。","forbiddenPatterns":["三列功能卡片","浏览器/编辑器壳作为主画面","发光粒子背景顶替项目图形","无证据的仪表盘/图表墙","首屏只放标语无标志视觉"],"compositionFingerprint":"中心放射的同心轨道天球布局，核心查看器居中，五个源端点在独立轨道上环布，矢径连接承担数据到达隐喻。"}],"selectedDirectionId":"conduit-pipeline","selectionRationale":"三个方向都能承载事实（多源、流式、100GB+、虚拟滚动），但 conduit-pipeline 最能直接表达项目核心机制：数据从多个远程/文件源被流式送入一个即时打开的查看视窗——这正是 README 与 verifiedClaims 反复出现的主承诺。infinite-ruler 聚焦单文件的虚拟化视界，方向正确但把多源接入弱化为从属；core-orbit 偏产品化的对称天球，更强的装饰性而机制驱动力较弱。conduit 的横向流控布局让多协议路由、流式速度与虚拟化窗格在同一视觉里并列成立，且其表面、章节与手机重构与前两项及历史 seed 的地层纵剖面均有实质差异，是最具证据契合度、也最不贴近营销模板的组合。"},"designBrief":{"schemaVersion":"dodot-design-brief-v1","selectedDirectionId":"conduit-pipeline","projectMetaphor":"远程数据源是一条条加压管道，把 100GB+ 的文件流送进一个即时打开的查看视窗；查看器不是文件浏览器而是流的中枢分岔点。","signatureVisual":"一条从左向右的动态流管网：若干来自不同源的语义节点（WebDAV/S3/SSH/Local/HF）汇入一条主管线，管线中段切开露出虚拟化的行缓冲窗格，数据在其间流动。","compositionFingerprint":"多源节点管线汇入一个虚拟化查看视窗的横向流控布局，流速刻度贯穿窗格，格式/插件章节以管线分支延伸。","heroGeometry":"横向张力构图：左侧五个源节点以流速各异的管线汇入右侧主查看窗格，一条贯穿的水平流轴承载延迟/吞吐刻度，无限大文件被压缩成一条可搜索的窄带。","sectionSequence":["conduit-hero","sources-routing","virtual-windowing","formats-parsing","plugins-sdk","local-first"],"surfaceModel":"工业级流控表面：亚光深色基材上叠加发光管线层，节点为高对比圆角瓣体，窗格为内凹视窗带刻度，全部层级带可感知的进深次序。","palette":"base #0A0F0E · surface #10181A · ink #E7EFEA · positive #3DDC97 管线 · amber #FFB454 流速负荷 · info #56CCF2 远端源 · muted #7E8C88","typography":"标题用硬朗的等宽强调（JetBrains Mono 式），正文用中性无衬线；协议名与流速标签用等宽小字。","desktopComposition":"1280px 下采用 12 栏：左 6 栏为源-管线带（含实时流速刻度），右 6 栏为查看视窗叠层，管线穿缝连接两栏；下方章节按证据顺序纵向展开。","mobileTransformation":"390px 下管线由横向改为垂直栈：五个源节点自上而下叠加为路由清单，主查看窗格收为单屏视窗，每章横向滚动改为纵向卡片。","mediaStrategy":"无本地合格素材，删除媒体区；以证据驱动的管线-流速度量图解与格式解析排版承担标志视觉，不伪造产品截图。","evidencePlan":["A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face.（repository.description / verifiedClaims）","它基于 Tauri、React 和 TypeScript 构建，可处理 100GB 以上的文件，并支持毫秒级搜索。（summary.claims[1]）","支持多种文件格式，包括 Parquet、Excel、CSV、JSON、代码文件，以及 ZIP/TAR 归档的直接预览。（summary.claims[2]）","提供插件开发 SDK 和示例插件，可扩展自定义文件格式查看器。（summary.claims[3]）","仓库 topics 含 ai、file-streaming、huggingface、s3、smb、ssh、tauri、text、vibe-coding、viewer、webdav；主语言为 TypeScript。（project.topics / primary_language）","本地化文案标注 100% AI 生成、插件生态尚早、无官方支持等 caveats。（summary.localizations.zh-CN.caveats）"],"localeContract":{"zh-CN":"默认中文，秒开的标题用「秒开超大文件，毫秒级搜索」；专名 WebDAV/S3/SSH/Local/Hugging Face/Parquet/Excel/CSV/JSON/ZIP/TAR/Tauri/TypeScript 保留原文；AI 生成与稳定性提示如实呈现。","en":"English title uses «Open massive files in seconds, search in milliseconds»; proper nouns WebDAV/S3/SSH/Local/Hugging Face/Parquet/Excel/CSV/JSON/ZIP/TAR/Tauri/TypeScript are kept; AI-generated and stability caveats shown truthfully."},"forbiddenPatterns":["三卡指标横幅","功能卡片墙（可套用任何项目的文案）","浏览器/终端/编辑器外壳冒充产品截图","装饰性发光网格或粒子替代项目视觉","伪造的开关/搜索/图表控件"],"historicalReuse":{"sourceAssetId":"HEAD:deepseek-v2/stardustai-dataset-viewer","decision":"rebuild","rationale":"历史 seed 的地层剖面隐喻（WebDAV/S3/SSH/Local/HF 五层纵堆叠，数据向下渗透）是上一代的孤立视觉资产，若沿用只会变成对旧 HTML 的改名而非 V4 产物；其侧重点在地质剖面而非「流式汇入 + 虚拟化查看」这一当前事实更突出的机制。决定重建为横向流控管线方向。仅提炼对该事实仍成立、且不与 antiTemplateContext 冲突的通用构图/排版色彩偏好。","reusedElements":["深色高对比基材下的等高显色语言（沿用暗底 + 亮色强调的同类对比手法）","将以协议为单位的源表示法保留为节点维度（各源作为一等对象呈现）","等宽字体用于协议/技术标签的排版语感","内凹窗格承载视图叠层的表面思路"],"discardedElements":["地质地层剖面/纵层堆叠隐喻","数据流自上而下的竖向渗透方向","源层标注的固定五层视觉与透明度阶梯透明度","竖向滚动的固定 main 布局与深绿 teal/sand 配色系统"]}},"visualReview":{"schemaVersion":"dodot-automated-design-review-v1","status":"approved","summary":"页面忠实实现项目专属 design-brief 选定的 conduit-pipeline 横向流控管线方向：五个源节点（WebDAV/S3/SSH/Local/Hugging Face）以不同流速汇入主管线，中段切开露出虚拟化行缓冲窗格并流向右侧视窗，signatureVisual 真实以内联 SVG 存在；章节顺序与 brief 的 sectionSequence 完全一致。所有文案事实（AI代理构建、100GB+流式、毫秒级搜索、Tauri/React/TypeScript、parquet/xlsx/csv/json/zip-tar/code 格式、插件 SDK、v1.6.4 多平台发行、AI 生成/插件早期/无官方支持等 caveats）均可溯源至 material.json 的 verifiedClaims/summary/localizations，无编造数据。手机的 720px 断点将多源网格、路由节点、视窗、格式板、插件行与本地区块全部重构为单列纵向栈，且 QA 确认移动端无横向溢出。zh-CN 与 en 语义等价、专名保留原文、英文自然流畅。未回退到通用 Hero、三卡指标横幅或装饰性发光网格（发光仅作管线符号的点缀而非主视觉），未出现浏览器/终端/编辑器外壳冒充截图。SVG 中的 LAT/THR 等仅为图解刻度标注，虚拟化行缓冲窗格是对产品真实功能的示意（role=img），无伪造开关/搜索/图表控件。","checks":{"projectFit":{"passed":true,"evidence":"页面采用 brief 选定方向 conduit-pipeline，标志视觉为五源节点经管线汇入虚拟化查看视窗的横向流控布局；sectionSequence 逐个实现为 conduit-hero → sources-routing → virtual-windowing → formats-parsing → plugins-sdk → local-first，与 design-brief.json 的 sectionSequence 完全一致；hero 几何（左侧源-管线带、右侧查看视窗叠层、流速刻度贯穿）符合 heroGeometry 与 desktopComposition。"},"evidenceBound":{"passed":true,"evidence":"核心主张均源自 material.json：'built entirely by AI Agent'、支持 WebDAV/S3/SSH/Local/Hugging Face 流式加载（description/topics）、100GB+ 与毫秒级搜索（summary claims）、Tauri+React+TypeScript（readme）、parquet/xlsx/csv/json/代码及 ZIP/TAR 直接预览（summary claims[1]）、插件 SDK 与示例插件（claims[3]）、v1.6.4 多平台发行（release）、AI 生成/插件早期/无官方支持等 caveats（localizations.caveats）。SVG 中 LAT 24ms/THR 812 MB/s 等为 diagrammatic 流速刻度注记，非陈述性事实。发行版本号 v1.6.4 与仓库链接均有材料依据。"},"signatureVisual":{"passed":true,"evidence":"conduit 标志视觉为真实内联 SVG（viewBox 1080x300），包含五个源节点、汇入分支管、主管线、中段切开的虚拟化行缓冲窗格（row 000 012/virtualized buffer）与右侧 VIEWPORT 目标窗格、流速刻度，role=img 且带描述性 aria-label；不依赖截图、外部图片或占位假图，真实存在于源码。"},"mobileTransformation":{"passed":true,"evidence":"@media (max-width:720px) 将 routing-grid 由双列改为单列、route 节点垂直栈排布、route-flow 由竖条改为横向条、window-band/format-board/plugin-row/local-band 全部改为单列纵向堆叠，且保留标题尺度的移动端收紧；qa-report 确认 zh-CN 与 en 移动端 scrollWidth==clientWidth==390（无横向溢出），形成真实重构而非单纯缩放。"},"localeParity":{"passed":true,"evidence":"zh-CN 与 en 语义等价：H1『秒开超大文件，毫秒级搜索』与『Open massive files in seconds, search in milliseconds』对应，六个章节逐节一致，WebDAV/S3/SSH/Local/Hugging Face/Parquet/Excel/CSV/JSON/ZIP/TAR/Tauri/TypeScript 等专名在两版均保留原文，英文为自然地道表达，标题与 meta description 均按语言适配，无缺失或错位内容。"},"antiTemplate":{"passed":true,"evidence":"未出现通用 Hero+三卡指标或功能卡片墙：采用项目专属的管线-视窗流控构图、证据驱动的格式解析面板与插件管线分支示意；标志视觉由 conduit 图解承担而非无依据的发光网格/粒子背景（glow 滤镜仅作管线光效点缀）；未用浏览器/终端/编辑器外壳冒充产品截图（虚拟化窗格与 plugin 示意图皆为 role=img 的示意性图解）。"},"noFakeUi":{"passed":true,"evidence":"页面无伪造的开关/搜索/图表控件：SVG 中 LAT 24ms/THR 812 MB/s 及 VIRTUAL WINDOW 等为图解标注或对产品真实能力（虚拟化行缓冲、秒开搜索）的示意，均标记 role=img & aria-hidden 或 figcaption；全部链接为真实仓库/官网链接，qa-report 记 fakeControlCount=0、emptyLinkCount=0，未引入不可交互的仿冒 UI。"}},"issues":[]},"repairEvidence":[],"provenance":{"executor":"opencode","model":"deepseek/deepseek-v4-flash","opencodeVersion":"1.1.58","promptVersion":"dodot-showcase-deepseek-v4","inputHash":"66136590d995d9cfac5a179613c8b54687e9143b11774e1760792fd5982e72ff","artifactHash":"c6f96358e18b673e57d7a48a8fd7cf83aef1b166b2f0f045716c55775ff8aab6","designSkill":"dodot-showcase-design","historicalSeed":{"source":"git_head","sourceCanonicalKey":"github:stardustai/dataset-viewer","assetId":"HEAD:deepseek-v2/stardustai-dataset-viewer","sourceSchemaVersion":"v1","sourceGeneratorVersion":"anthropic:claude-fable-5","sourceArtifactHash":"f8874051e668b4cbc58d0c227ed93bb0740d4c93b46bf8e091756c2fed1c6c50","disposition":"repair_or_regenerate","seedPayloadHash":"141f78d5a6ab792fcfab0869b9a1083523d831a161ecde482693542e8f312c33"}},"qaEvidence":{"zh-CN":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":5,"sectionCount":6,"textLength":2119,"cjkCount":867,"latinLetterCount":719,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":4471,"h1Top":147.546875,"bottomInset":330.84375,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"5ddc3dcf8c01b7a0a268a72e914e56eaeb6baa262e634f0abceca633637536be","middle":"dddb62fdc0e7d53e67b3f78ee388ff1a4e78a5e13ba29267cf8e4dd6102e7475","bottom":"11a0de783232783c77e9303c3483d6c9e0c335ce2c438874c31f0c1993e11abf"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":5,"sectionCount":6,"textLength":2119,"cjkCount":867,"latinLetterCount":719,"scrollWidth":390,"clientWidth":390,"scrollHeight":6361,"h1Top":163.09375,"bottomInset":350.484375,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"27b6c290aeaf2f1cd012c2237f69ce8fbf67f7b9c532a3b035ff6db7fb257584","middle":"ed7ab0810fe3182b1885d08812a0ea398a421b45c391eb2ed373625899b12c8a","bottom":"9ef5ef5becd9e528ccf32f5ecec7da3c24f3d9045e8b623031e43ca9b41e665a"},"consoleErrors":[],"networkRequests":[]}],"en":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":5,"sectionCount":6,"textLength":4310,"cjkCount":0,"latinLetterCount":3328,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":4702,"h1Top":147.546875,"bottomInset":331.328125,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"e863ed80f78a86b24b27b16bea2ab986fe07a97e993eafad939e07d139ad2042","middle":"41c0030be30b9285b1e3a7d8b87fcbd3a09eee9fdf82ed91cab9ac9766a2cb00","bottom":"c967b873abab988ca1588884a27dc6b172420214e18323cc444ef03710e7a2bb"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":5,"sectionCount":6,"textLength":4310,"cjkCount":0,"latinLetterCount":3328,"scrollWidth":390,"clientWidth":390,"scrollHeight":6987,"h1Top":163.09375,"bottomInset":351.03125,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"3dddd1432984fbc780a8019a85e4224fb78a178fb175cb435e5b868178da991e","middle":"e6c97f74188b495a450c097c795aed877f004f2ea31c38919a3a335c69258bce","bottom":"9560ac2936f54c0b7b6e87c3f53577a23d36e3a6bb4d77c03278fd9dfae5b26e"},"consoleErrors":[],"networkRequests":[]}]}},"generatorVersion":"opencode-1.1.58:deepseek/deepseek-v4-flash:dodot-showcase-deepseek-v4","publishedAt":"2026-08-20T14:58:07.302Z"},"rawExtraction":{"projectRevisionId":"cmt2zh31k01mo6271edwh4njg","projectVersion":1,"sources":[{"canonicalKey":"github:stardustai/dataset-viewer","role":"primary","revisionVersion":1,"url":"https://github.com/stardustai/dataset-viewer"}],"facts":[{"key":"repository_owner","value":"stardustai","evidence":[{"path":"repository.owner","excerpt":"stardustai"}]},{"key":"repository_name","value":"dataset-viewer","evidence":[{"path":"repository.name","excerpt":"dataset-viewer"}]},{"key":"description","value":"A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face.","evidence":[{"path":"repository.description","excerpt":"A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face."}]},{"key":"primary_language","value":"TypeScript","evidence":[{"path":"repository.language","excerpt":"TypeScript"}]},{"key":"topics","value":"ai, file-streaming, huggingface, s3, smb, ssh, tauri, text, vibe-coding, viewer, webdav","evidence":[{"path":"repository.topics","excerpt":"ai, file-streaming, huggingface, s3, smb, ssh, tauri, text, vibe-coding, viewer, webdav"}]},{"key":"source_count","value":"1","evidence":[{"path":"projectRevision.sources","excerpt":"github:stardustai/dataset-viewer"}]}],"taxonomy":{"category":"development_tools","subcategory":"data_ai_tools","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face."}]},"lifecycle":{"stage":"active","confidence":0.75,"evidence":[{"path":"repository.pushedAt","excerpt":"2026-03-28T10:03:05.000Z"}]},"userIntents":[{"slug":"build_software","label":"构建或改进软件","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face."}]},{"slug":"automate_work","label":"自动化工作流程","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face."}]},{"slug":"create_content","label":"创作内容或设计","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face."}]},{"slug":"analyze_data","label":"分析与理解数据","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face."}]},{"slug":"learn_explore","label":"学习与探索","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"A sleek dataset viewer built entirely by AI Agent. Supports streaming large files from WebDAV, S3, SSH, Local or Hugging Face."}]}],"readme":{"headings":["Dataset Viewer","🚀 Key Features","📚 Supported File Types","📸 Screenshots","✨ Technical Highlights","🎯 Perfect For","🔌 Plugin Development","🤝 Contributing","🙏 Acknowledgments","📄 License"],"commands":[]},"changeSet":[{"field":"authorName","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"fdd777f5f13bee03c4fbb48934bf595da699f2820b6c5e59d13f61ec275eac7d"},{"field":"description","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"06406eb8b8dedb6e982ad8aae814c49a71810f0bcc7782e8874e325e155fd5fe"},{"field":"experienceUrl","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"540c28fe70575213b2d1ed4b79ba7ac92387ac14b6f8a5ce598c6e53ba02fd35"},{"field":"name","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"1b9e3550e57db4717c186a27897192244c92cbc5f52eca19ded00849f2f552f9"},{"field":"repository","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"dd393e1d2f3be7f775ec3ff12b3676ed6c9d5e8eb118d3d65c6b1a48ed824ef6"},{"field":"sourceCount","beforeHash":"74234e98afe7498fb5daf1f36ac2d78acc339464f950703b8c019892f982b90b","afterHash":"6b86b273ff34fce19d6b804eff5a3f5747ada4eaa22f1d49c01e52ddb7875b4b"}]},"publishedAt":"2026-08-13T04:00:37.540Z","updatedAt":"2026-08-21T13:26:53.904Z","summary":{"id":"cmt2zh31n01mq6271tjdqecuv","projectType":"product","oneLine":"AI 构建的高性能数据集查看器","overview":"Dataset Viewer 是一款由 AI 代理完全构建的桌面应用，基于 Tauri 和 React，专为处理超大文件设计。它支持从 WebDAV、S3、SSH、本地或 Hugging Face 流式加载 100GB 以上的文件，提供毫秒级搜索和虚拟滚动。支持 Parquet、Excel、CSV、JSON、代码文件及 ZIP/TAR 归档的直接预览，并支持插件扩展。适合数据科学家、日志分析和远程数据访问场景。","highlights":["秒开 100GB+ 文件","毫秒级搜索","多协议远程访问","插件扩展"],"bestFor":["数据科学家","日志分析","远程数据访问"],"caveats":["100% AI 生成，需自行评估稳定性","插件生态尚在早期","未提供官方支持渠道"],"categorySuggestion":"development_tools","subcategorySuggestion":"data_ai_tools","lifecycleStages":["testing_quality","deployment_growth"],"audienceIntents":["discover","evaluate","operate"],"confidence":0.85,"evidence":[{"claim":"Dataset Viewer 是一个由 AI 代理完全构建的桌面应用，支持从 WebDAV、S3、SSH、本地或 Hugging Face 流式加载大型文件。","sourceType":"description","locator":"https://github.com/stardustai/dataset-viewer"},{"claim":"它基于 Tauri、React 和 TypeScript 构建，可处理 100GB 以上的文件，并支持毫秒级搜索。","sourceType":"readme","locator":"https://github.com/stardustai/dataset-viewer"},{"claim":"支持多种文件格式，包括 Parquet、Excel、CSV、JSON、代码文件，以及 ZIP/TAR 归档的直接预览。","sourceType":"readme","locator":"https://github.com/stardustai/dataset-viewer"},{"claim":"提供插件开发 SDK 和示例插件，可扩展自定义文件格式查看器。","sourceType":"readme","locator":"https://github.com/stardustai/dataset-viewer"},{"claim":"最新版本 v1.6.4 提供 Windows、macOS 和 Linux 的安装包。","sourceType":"release","locator":"https://github.com/stardustai/dataset-viewer/releases/tag/v1.6.4"}],"publishedAt":"2026-08-17T14:13:43.246Z"},"github":{"url":"https://github.com/stardustai/dataset-viewer","ownerAvatarUrl":"https://avatars.githubusercontent.com/u/36030373?v=4","starsCount":1000,"forksCount":69,"language":"TypeScript","syncedAt":"2026-08-17T14:11:28.500Z","license":null,"defaultBranch":"main","archived":false,"disabled":false,"lastChangedAt":"2026-03-28T10:03:05.000Z","missing":false},"tags":[{"name":"Vibe coding","slug":"vibe-coding","verification":"verified","evidenceUrl":"https://github.com/stardustai/dataset-viewer","evidenceExcerpt":"GitHub topic: vibe-coding"}],"updates":[]},"entryPoint":"discovery","claim":{"state":"unclaimed","verified":false},"initialView":"beautified","community":{"viewer":null,"state":{"likesCount":0,"creativeCount":0,"followersCount":0,"commentsCount":0,"liked":false,"creative":false,"followed":false},"comments":{"comments":[],"nextCursor":null}}}]}],"$L24"]