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:T3751, Coffee CLI — 一边 Vibe Coding，一边保持时尚

Vibe Coding on one side, looking stylish on the other.

专业速记 · expert

coffee-cli — Rust / Tauri
并行 AI CLI 代理工作台 ×11

朴素阅读 · human

一边 Vibe Coding，
一边保持时尚。

Coffee CLI

一座供 AI 编程代理穿梭的原生桌面工作台：让只懂机器速记的人，和只读朴素文字的人，都站在同一条工作流的两侧，没有人被困在单边海岸。

共享工作 · shared

multi-tab · parallel · history · install · remote

tauri(rust) — 多标签会话 / 历史 / 文件浏览 / 任务板 / 代理安装器 / 远程终端 / 外部启动

朴素阅读 · human

同一个意图，同时跑在两条岸上

同一段工作流，既可以用密集的专业缩写来驱动，也可以用一行朴素的话来理解。多标签页并行会话让你同时对着多个 AI 代理而不必在终端里迷路。

语言矩阵 · locales

11 locales / 同一意图

代码语言 11界面语言，处处同义 i18n同一措辞，逐语同步

朴素阅读 · human

十一种语言，一个意思

界面支持 11 种语言，英语、简体中文、繁体中文、日语、韩语都在其中。这条镜像华美之处不在于有多少套译文，而在于每一套译文都忠实于同一个意图——专业的人写专业，不懂英语的人读自己的母语，互不缺席。

代理名录 · agents

claude-code · codex · qwen · cli

Rust — 一行安装 / 多代理并行 / 会话历史持久

朴素阅读 · human

让 AI 代理成为你的同事

Claude Code、Codex、Qwen……这些原本藏在命令行深处的代理，现在可以一键安装、并行调度、随时回看历史。你不需要记住任何指令，只要说出你想做的事。

多标签页并行会话，工作互不阻塞
内置任务板，把想法落成清单
会话历史与文件浏览器，随时捡起进度
一键安装 AI 代理，免去环境配置

等价契约 · pact

AGPL-3.0 · edison7009 · rust · 169*

coffee-cli — 开源镜像两岸，一等公民双语

朴素阅读 · human

开源、双语、人人能读

Coffee CLI 以 Rust 构建、遵循 AGPL-3.0 开源，由 edison7009 维护。它不是把界面翻译一遍就算了——而是让「专业的人不降级，朴素的人不敬畏」成为同一件事的两面。

项目主页 coffeecli.com GitHub 源码

注意 · caution

deps > features > evidence

cli install ≠ 零配置 / 能力依赖代理 / 证据=README

朴素阅读 · human

动身前，先看清这几处浅滩

仍需自行安装相应的 AI CLI 工具
部分高级功能可能取决于你选择的特定代理
以上描述基于项目 README，实际体验可能略有出入

Coffee CLI · edison7009

23:T360b, Coffee CLI — Vibe Coding, stylishly.

一边 Vibe Coding，一边保持时尚。

expert shorthand

coffee-cli — Rust / Tauri
parallel AI CLI agent bench ×11

plain reading

Vibe Coding,
stylishly.

Coffee CLI

A native desktop bench where AI coding agents shuttle between two shores — for the machine that speaks shorthand and the human who reads plain words. Nobody is locked on one side of the workflow.

shared work

multi-tab · parallel · history · install · remote

tauri(rust) — multi-tab / history / file explorer / task board / agent installer / remote terminal / external launch

plain reading

One workflow, mirrored on two shores

The same intent runs simultaneously as tight shorthand and as a sentence in plain words. Multi-tab parallel sessions let you sit with several AI agents at once without ever getting lost in a terminal.

locale matrix

11 locales / 1 intent

11interface languages, one meaning i18none wording, in sync across them all

plain reading

Eleven languages, one meaning

The interface ships in 11 languages — English, Simplified Chinese, Traditional Chinese, Japanese, Korean, and more. What makes the mirror matter is not how many translations exist, but that every one stays faithful to the same intent: the expert writes dense, the newcomer reads in their own tongue, and neither side goes missing.

agent registry

claude-code · codex · qwen · cli

Rust — one-command install / parallel agents / persistent session history

plain reading

Turn AI agents into coworkers

Claude Code, Codex, Qwen — agents that once lived deep inside a command prompt can now be installed in one step, scheduled in parallel, and revisited at any time. You never need to recall a single command; just say what you want done.

Multi-tab parallel sessions that never block each other
A built-in task board to turn thoughts into a list
Session history and a file explorer to pick up where you left off
One-click agent installation, no environment setup

equivalence pact

AGPL-3.0 · edison7009 · rust · 169*

coffee-cli — open source, mirrored across both shores

plain reading

Open, bilingual, readable by everyone

Coffee CLI is built with Rust, released under AGPL-3.0, and maintained by edison7009. Its promise is not a translated skin, but a pact: the expert is never dumbed down and the newcomer is never made to feel intimidated — two faces of the same act.

Project site — coffeecli.com Source on GitHub

caution

deps > features > evidence

cli install ≠ zero-config / capability depends on agent / evidence = README

plain reading

Before you set out, mind these shallows

You will still need to install the corresponding AI CLI tools yourself
Some advanced features may depend on the specific agent you choose
This page draws on the project README; real-world experience may differ

Coffee CLI · edison7009

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","cmt2zgzpi00fi6271hztymv5a:zh-CN:beautified",{"project":{"id":"cmt2zgzpi00fi6271hztymv5a","slug":"edison7009-coffee-cli","name":"coffee-cli","description":"AI CLI 代理的原生桌面工作台","authorName":"edison7009","authorVerification":{"verified":false,"verifiedAt":null},"category":"development_tools","subcategory":"data_ai_tools","experienceUrl":"https://coffeecli.com/","qualityScore":82,"featured":false,"officialSite":null,"views":{"original":{"available":true},"official":{"available":false,"reason":"disabled_by_policy"},"beautified":{"available":true,"reason":null},"defaultView":"beautified"},"beautifiedView":{"revisionId":"cmt2zgzq100fu6271j4kl6go4","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":"relay-relay","name":"The Relay Counter","projectMetaphor":"A phrase is passed across a counter between two very different worlds: the machine side that does the work (raw, exact, dense) and the human side that feels good looking at it (calm, stylish, open). The project's whole mechanism is relaying workflow across that same gap for people who never learned a terminal.","signatureVisual":"A horizontal relay track: one small message token (a rounded bar/capsule carrying a phrase) sliding from a left 'machine' pool of glyphs toward a right 'human' pool of words, with a line of tiny language emblems (their alphabets) standing between them like counter staff. The same phrase appears once as dense glyphs and once as plain words, linked by the passing token.","heroGeometry":"A wide horizontal relay bar that runs across the full viewport. On the left, a tight stack of code glyphs; on the right, an oversized clean word; between them a single capsule token with the slogan, resting on a baseline that visually 'passes' the message left to right.","sectionSequence":["relayDemonstration","translationBridge","parallelSessions","taskBoardFlow","languageMatrix","cautionNotes"],"surfaceModel":"A continuous countertop surface: the page reads as one horizontal plane across which content moves and settles. Cards are flattened into low 'slip' layers, no thick walls; depth comes from left-to-right travel rather than vertical stacking.","palette":"Warm coffee-cream base with two contrast poles — a machine ink (deep graphite) and a human accent (amber/espresso highlight). Small language emblems tinted in restrained jewel tones used only on the relay line.","typography":"A paired system: a crisp humanist sans (upright, wide, warm) for the human side, and a condensed monoline sans (tall, tight, mechanical) reserved for machine glyphs and layer labels. Pull quote uses the upright face at a large optical size.","desktopComposition":"The hero is a full-width horizontal relay not limited to a centered column. Sections alternate between left-anchored meter-like rows and a central rail; the relay line recurs at low opacity behind section tops to bind the page to the same passage.","mobileTransformation":"The horizontal relay collapses into a vertical 'passthrough': the token moves top-to-bottom between stacked pools, heavy elements sequence down the scroll instead of side by side; the machine glyphs remain thin and left-aligned so the concept survives without horizontal space.","mediaStrategy":"No local assets exist. The relay line becomes the media region: purely typographic and diagrammatic movement, no imported screenshots or frames. The passing-token motion is the sole 'image'.","forbiddenPatterns":["generic slogan over three metric cards","feature-card wall that can host any project copy","full terminal or browser shell frame","decorative gradient glow behind a centered headline"],"compositionFingerprint":"horizontal relay counter / machine-to-human passthrough token / recurring baseline rail"},{"id":"roast-stream","name":"The Roastery Line","projectMetaphor":"Independent beans (each an AI agent with its own flavor) are dropped into a roast line and emerge as one brewed cup of finished work. The project runs several agents in parallel and unifies them into a deliverable — exactly a roast, not a single appliance.","signatureVisual":"A processing stream: discrete rounded units (beans/releases) enter from the top-left along a downward incline, merge into one channel, and exit as a single clean disc. Between entry and exit, small step markers show parallel lanes folding into one — a literal diagram of 'many agents, one result'.","heroGeometry":"A descending diagonal stream across the viewport: multiple small inputs at the top merge into a single bold output at the bottom-right. The eye walks inputs → compression → output, mirroring the orchestration the app actually performs.","sectionSequence":["roastDemonstration","ingredientGlossary","unifiedBrew","sessionLedger","installLine","cautionNotes"],"surfaceModel":"Interlocking 'tray' layers that read as bodies in a roast hopper — each section is a shallow tilted tray whose lower edge overlaps the next, so the page physically conveys downward flow. Elements sit inside the trays with visible chutes rather than floating on a flat canvas.","palette":"Warm roast spectrum moving dark→light down the page: deep charcoal-roast at the hero, caramel mid-tones for working sections, cream toward the outcome. Parallel inputs share a cool slate counterpoint so multi-agent voices stay distinguishable.","typography":"A sturdy slab-formed serif for outcomes (the 'cup' feels finished and grounded) used sparingly, paired with a neutral grotesque sans for inputs and labels. The finished result always typeset in the serif to mark 'delivered'.","desktopComposition":"The hero stream occupies the upper two-thirds as a descending system; below, the page is organized as stacked tilted trays, each with a long intake label and a short settled outcome. Every section keeps the input-pool → unified-result grammar.","mobileTransformation":"The diagonal stream becomes a straight vertical fall — small inputs stack at the top and the single disc lands on screen as you scroll. Trays stack vertically with the overlap reduced to a dark seam so downward progress still reads on a single column.","mediaStrategy":"No local media. The stream and its merging lanes are the visual: a derived diagram expressing the verified 'multiple agents → one result' relationship, built with pure shape and motion, never impersonating a product screenshot.","forbiddenPatterns":["hero screenshot or editor frame","flat grid of identical cards","fake install progress bars or spinners","glowing particle field as background texture"],"compositionFingerprint":"descending roast stream / many-to-one merge lanes / overlapping tray sections"},{"id":"mirror-echo","name":"Bilingual Mirror","projectMetaphor":"Two speakers who share no vocabulary still carry on because each sees the same intent echoed back in their own language. The app's real promise is that mirroring: the same workflow runs on a dense, expert side and a plain, native-language side simultaneously — nobody is locked on one shore.","signatureVisual":"A two-line mirror stave: the same phrase traced once as sparse, expert shorthand (top) and once as a fuller, plain-language sentence (bottom), joined at the spine by a series of short matching tics. Like a bilingual score, both lines stay in sync across the whole piece.","heroGeometry":"A split field divided by a central vertical spine. The left half carries compact expert marks, the right half the expansive human reading; at the spine, tics connect matching positions so the whole reads as one translated score rather than two separate halves.","sectionSequence":["mirrorDemonstration","sharedWork","languageMatrix","agencyList","equivalencePact","cautionNotes"],"surfaceModel":"Mirrored 'pages' that face each other across the page — left and right bodies hold echo pairs of what is otherwise the same content, and the central spine is the only join. Sections alternate which half is visually dominant but never break the mirror pairing.","palette":"Two-tone symmetry: one cool 'expert shore' ink and one warm 'human shore' ink held apart by a neutral paper ground. Only the connecting tics are allowed a third, bright accent so the translation seam is the sole color event.","typography":"A dual typographic voice: a technical mono for the expert side and an open, easily-read text serif for the human side. Every echoed phrase is set in matching weights so both faces of the same sentence feel like translations, not two different designs.","desktopComposition":"A true two-column split sustained past the hero: content is echoed left/right across sections with the spine running the full page height. Eye movement is deliberately east-west along translation pairs, not top-down through marketing blocks.","mobileTransformation":"On a single column the two shores serialize: the expert line and its plain echo sit stacked one above the other, the shared tics now reading as short connecting rules between them. The mirror survives because each phrase is still shown twice, once per voice.","mediaStrategy":"No local assets. The mirror score and its matching tics form the hero visual — an evidence-driven relationship ('11 languages, same intent') drawn from the project, in pure type and line, never a faked localization dropdown or screenshot.","forbiddenPatterns":["fake translation dropdown or language picker UI","two-column marketing layout with no echoed content relation","terminal window that pretends to be the app","decorative symmetric gradient divider"],"compositionFingerprint":"bilateral mirror stave / expert-and-human echo pairs / continuous central spine translation seam"}],"selectedDirectionId":"mirror-echo","selectionRationale":"mirror-echo best expresses the project's mechanism rather than its packaging: the only verified relational claims are 'multiple AI agents' and '11 interface languages for non-developers', and this direction renders those as one continuous act — the same intent shown simultaneously in expert shorthand and plain native language. The relay direction courts the generic 'productive pipeline' reading and the roastery direction leans on an invented food metaphor; the mirror stays closest to the evidence (parallel agents + 11 languages) while its east-west translation seam is the least reusable grammar across the antiTemplateContext and the least confused with a typical hero/metric-card page. It also sacrifices no media honesty: with no local assets the mirror score itself is the composition, so the brief does not depend on doctored screenshots."},"designBrief":{"schemaVersion":"dodot-design-brief-v1","selectedDirectionId":"mirror-echo","projectMetaphor":"Two speakers who share no vocabulary still carry on because each sees the same intent echoed back in their own language. The app's real promise is that mirroring: the same workflow runs on a dense, expert side and a plain, native-language side simultaneously — nobody is locked on one shore.","signatureVisual":"A two-line mirror stave: the same phrase traced once as sparse, expert shorthand (top) and once as a fuller, plain-language sentence (bottom), joined at the spine by a series of short matching tics. Like a bilingual score, both lines stay in sync across the whole piece.","compositionFingerprint":"bilateral mirror stave / expert-and-human echo pairs / continuous central spine translation seam","heroGeometry":"A split field divided by a central vertical spine. The left half carries compact expert marks, the right half the expansive human reading; at the spine, tics connect matching positions so the whole reads as one translated score rather than two separate halves.","sectionSequence":["mirrorDemonstration","sharedWork","languageMatrix","agencyList","equivalencePact","cautionNotes"],"surfaceModel":"Mirrored 'pages' that face each other across the page — left and right bodies hold echo pairs of what is otherwise the same content, and the central spine is the only join. Sections alternate which half is visually dominant but never break the mirror pairing.","palette":"Two-tone symmetry: one cool 'expert shore' ink and one warm 'human shore' ink held apart by a neutral paper ground. Only the connecting tics are allowed a third, bright accent so the translation seam is the sole color event.","typography":"A dual typographic voice: a technical mono for the expert side and an open, easily-read text serif for the human side. Every echoed phrase is set in matching weights so both faces of the same sentence feel like translations, not two different designs.","desktopComposition":"A true two-column split sustained past the hero: content is echoed left/right across sections with the spine running the full page height. Eye movement is deliberately east-west along translation pairs, not top-down through marketing blocks.","mobileTransformation":"On a single column the two shores serialize: the expert line and its plain echo sit stacked one above the other, the shared tics now reading as short connecting rules between them. The mirror survives because each phrase is still shown twice, once per voice.","mediaStrategy":"No local assets. The mirror score and its matching tics form the hero visual — an evidence-driven relationship ('11 languages, same intent') drawn from the project, in pure type and line, never a faked localization dropdown or screenshot.","evidencePlan":["hero mirror demonstrates '11 interface languages' and 'for non-developers' from the summary highlights, so the split must present the same phrase as expert shorthand and plain native language without inventing a number or screenshot","sharedWork section must route off the verified multi-tab sessions + parallel agents claims from the overview, echoed as equivalent pairs across both shores","languageMatrix and equivalencePact must enumerate only the 11-language and AGPL-3.0/owner facts that appear in material.json, never fabricated locales","cautionNotes must carry the three verified caveats (requires installing the AI CLI tools, features depend on the agent, evidence is README-based)"],"localeContract":{"zh-CN":"同一条内容的专业速记与朴素中文阅读并存，左右两栏互为译文；直接使用简体中文 slogan「一边 Vibe Coding，一边保持时尚。」，长度、排版密度保持中文自然节奏。","en":"Expert shorthand and plain English reading echo each other across the split; use the English slogan 'Vibe Coding, stylishly.' with natural English typographic rhythm."},"forbiddenPatterns":["fake translation dropdown or language picker UI","two-column marketing layout with no echoed content relation","terminal window that pretends to be the app","decorative symmetric gradient divider","slogan over exactly three metric cards","feature-card wall that could accept any project's copy"]},"visualReview":{"schemaVersion":"dodot-automated-design-review-v1","status":"approved","summary":"Both zh-CN and en artifacts faithfully implement the selected mirror-echo direction: a sustained two-column mirror split (mono expert shore / serif human shore) joined by a continuous central spine and seam tics, with the same phrase echoed once as dense shorthand and once as plain language. All content is traceable to material.json facts, the signature mirror stave is genuinely realized in type and line (no screenshots), mobile collapses the two shores into a stacked vertical passthrough with tics re-reading as short rules, and the two locales are structurally and semantically equivalent. No generic Hero / three-metric-card / card-wall / device-shell / decorative gradient patterns and no fake controls were found.","checks":{"projectFit":{"passed":true,"evidence":"Both artifacts implement the mirror-echo composition: .stave and .mirror-pair use a 1fr/1fr split with .spine-rail/.spine-core running the full page height, mono .expert voices vs serif .human voices, and tics at the seam. Section sequence in HTML matches the brief exactly (mirrorDemonstration, sharedWork, languageMatrix, agencyList, equivalencePact, cautionNotes) and the desktop split is sustained past the hero."},"evidenceBound":{"passed":true,"evidence":"Every claim traces to material.json: Rust/Tauri (primary_language, overview), AGPL-3.0 license, edison7009 owner, 11 interface languages with en/zh-CN/ja/ko listed, multi-tab/parallel sessions/history/file explorer/task board/agent installer/remote terminal/external launch (overview claim), agents Claude Code/Codex/Qwen, star count 169 (repository.stars), and the three caveats (install CLI tools, feature depends on agent, evidence is README-based). Slogans match the stored localizations."},"signatureVisual":{"passed":true,"evidence":"The two-line mirror stave is real: the hero presents the same phrase as a tight expert shorthand (coffee-cli — Rust / Tauri / parallel AI CLI agent bench x11) above an oversized plain human reading (一边 Vibe Coding，一边保持时尚 / Vibe Coding, stylishly.), joined by seam-colored tics. It is built purely from type and line, not a doctored screenshot or faked asset."},"mobileTransformation":{"passed":true,"evidence":"At max-width:760px the media query collapses .mirror-pair and .stave to a single column, stacking expert above human (text-align flips to left, borders switch to horizontal), moves the spine to left:8px, and re-forms the seam tics as short vertical connecting rules. This is a genuine structural reflow (scrollHeight grows 3005->4685 zh, 3245->5442 en), not mere scaling."},"localeParity":{"passed":true,"evidence":"Both files share identical structure, section sequence (header/sections), and mirror grammar. zh-CN carries natural Simplified Chinese copy and en natural English copy, each echoing the equivalent meaning across the two shores. Slogans match the localization data, and the cross-language quirk (zh hero shows the English repo description line, en hero shows the Chinese slogan) is equivalent and intentional."},"antiTemplate":{"passed":true,"evidence":"No slogan-over-three-metric-cards block and no interchangeable feature-card wall appear; the hero is the mirror stave rather than a centered headline. No terminal/browser device shell is used, and the background is a flat paper tone without a decorative gradient glow. The recurring seam/rail grammar is specific to the mirror-echo concept."},"noFakeUi":{"passed":true,"evidence":"No fabricated translation dropdown, language picker, install progress bar, or spinner exists. All real links (coffeecli.com, github.com/edison7009/coffee-cli) point to the verified targets. QA report records fakeControlCount:0, brokenImageCount:0, emptyLinkCount:0 for both locales and viewports."}},"issues":[]},"repairEvidence":[],"provenance":{"executor":"opencode","model":"deepseek/deepseek-v4-flash","opencodeVersion":"1.1.58","promptVersion":"dodot-showcase-deepseek-v4","inputHash":"0ff210117fd5d68d31448926b517387228eb7e531699062f7448e88922d147d1","artifactHash":"75768c239b583b909c49044b447bc7ec90af40f13027f422edb6149f5adbdce9","designSkill":"dodot-showcase-design"},"qaEvidence":{"zh-CN":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":10,"sectionCount":6,"textLength":1333,"cjkCount":607,"latinLetterCount":384,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":3005,"h1Top":467.4375,"bottomInset":292.421875,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"058cdad6dc80afd5e30e4be6df295ca36ff31cf9082a11dcf0d35423cd02ec6d","middle":"4c2de4534bf18b80dedcf8358f1268bc03e9ec7f78ef1bd24bde357351833b46","bottom":"ce6ae4e7126b5aea564864d46030cc141614bfe5e043a9a76640876f90a831cb"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":10,"sectionCount":6,"textLength":1333,"cjkCount":607,"latinLetterCount":384,"scrollWidth":390,"clientWidth":390,"scrollHeight":4685,"h1Top":648.265625,"bottomInset":283.71875,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"7e0014ac17f169251db048b3e66834a0b8fe578a67859e57ac0edade164d4369","middle":"a73b7fa39e97e36d27b39fcaea4030d88664386f4f300e866093ed40a0ca2051","bottom":"7ca4a87d2499c34e17164f99442df999ceb8cc1d76ef68463d1194e44e04253b"},"consoleErrors":[],"networkRequests":[]}],"en":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":10,"sectionCount":6,"textLength":2759,"cjkCount":8,"latinLetterCount":2144,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":3245,"h1Top":448.03125,"bottomInset":291.953125,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"f903994e4dd732bc317d570cbd75b2597b1fbbd8ce9fadc08940940ef9283fb8","middle":"43c879e344bfc8ebdf1e52a8d2383dacbf2f8e43065d20ff846da1b3d434f9d9","bottom":"a8fac7eab4da7a074e3e1b6f40b70dbf972459229534e695097e84053286ae2e"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":10,"sectionCount":6,"textLength":2759,"cjkCount":8,"latinLetterCount":2144,"scrollWidth":390,"clientWidth":390,"scrollHeight":5442,"h1Top":633.984375,"bottomInset":283.640625,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"5af2870b6ef164f35a92a6eb86dd7b7822a3107d9fda0d2a4ada4980219177c1","middle":"06f045dbb64b6333b848bbd39130707f527fbcfbad0d5bafdb2adbbb2eb8aefa","bottom":"16508a6ec6840470a197fe7108827e0b2970fbed6e04f271621a94519fecaaaf"},"consoleErrors":[],"networkRequests":[]}]}},"generatorVersion":"opencode-1.1.58:deepseek/deepseek-v4-flash:dodot-showcase-deepseek-v4","publishedAt":"2026-08-17T15:21:48.068Z"},"rawExtraction":{"projectRevisionId":"cmt2zgzpt00fo62714er9p1pp","projectVersion":1,"sources":[{"canonicalKey":"github:edison7009/coffee-cli","role":"primary","revisionVersion":1,"url":"https://github.com/edison7009/Coffee-CLI"}],"facts":[{"key":"repository_owner","value":"edison7009","evidence":[{"path":"repository.owner","excerpt":"edison7009"}]},{"key":"repository_name","value":"coffee-cli","evidence":[{"path":"repository.name","excerpt":"coffee-cli"}]},{"key":"description","value":"Vibe Coding on one side, looking stylish on the other.","evidence":[{"path":"repository.description","excerpt":"Vibe Coding on one side, looking stylish on the other."}]},{"key":"primary_language","value":"Rust","evidence":[{"path":"repository.language","excerpt":"Rust"}]},{"key":"license","value":"AGPL-3.0","evidence":[{"path":"repository.license","excerpt":"AGPL-3.0"}]},{"key":"topics","value":"claude-code, cli, multi-agent, multi-agent-systems","evidence":[{"path":"repository.topics","excerpt":"claude-code, cli, multi-agent, multi-agent-systems"}]},{"key":"source_count","value":"1","evidence":[{"path":"projectRevision.sources","excerpt":"github:edison7009/coffee-cli"}]}],"taxonomy":{"category":"development_tools","subcategory":"data_ai_tools","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Vibe Coding on one side, looking stylish on the other."}]},"lifecycle":{"stage":"active","confidence":0.75,"evidence":[{"path":"repository.pushedAt","excerpt":"2026-08-17T12:30:40.000Z"}]},"userIntents":[{"slug":"build_software","label":"构建或改进软件","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Vibe Coding on one side, looking stylish on the other."}]},{"slug":"automate_work","label":"自动化工作流程","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Vibe Coding on one side, looking stylish on the other."}]},{"slug":"create_content","label":"创作内容或设计","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Vibe Coding on one side, looking stylish on the other."}]},{"slug":"learn_explore","label":"学习与探索","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"Vibe Coding on one side, looking stylish on the other."}]}],"readme":{"headings":["English","What is Coffee CLI?","Supported AI CLIs","Who is it for?","See It In Action","Theme Switching","Task Board","Session History","Multi-Tab: Vibe Coding + DOS Game","11-Language Interface","Key Features","External Launch (CLI)","macOS — call the binary directly (LaunchServices drops --args when the app is already running)","Windows — pair with an Explorer right-click entry that passes the folder as %V","Linux","Install","Build from Source","Prerequisites: Rust, Node.js","简体中文","Coffee CLI 是什么？"],"commands":["cargo tauri build"]},"changeSet":[{"field":"repository","beforeHash":"ad05240117d7302829cc94c872d5685b780a4b6003c7ca528ed540d1f59a21f4","afterHash":"d4a1b8dbf6ddb77673173d7ec41052816910b794f926b3031158c4decd4c7a0d"}]},"publishedAt":"2026-08-16T12:25:56.094Z","updatedAt":"2026-08-21T13:26:49.568Z","summary":{"id":"cmt2zgzpx00fq627114soyrwd","projectType":"product","oneLine":"AI CLI 代理的原生桌面工作台","overview":"Coffee CLI 是一个基于 Tauri 和 Rust 构建的原生桌面应用，用于运行和管理多个 AI CLI 代理，如 Claude Code、Codex 和 Qwen。它提供多标签页会话、会话历史、内置文件浏览器、任务板、代理安装器和远程终端等功能，并支持 11 种界面语言。旨在让非开发者也能轻松使用 AI 编程代理，无需终端经验或英语能力。","highlights":["多标签页并行会话","11 种界面语言","内置任务板","一键安装代理"],"bestFor":["希望使用 AI 编程代理的非开发者","需要管理多个 AI 代理的专业人士","偏好原生桌面应用的用户"],"caveats":["需要安装相应的 AI CLI 工具","部分高级功能可能依赖特定代理","证据基于 README，实际体验可能有所不同"],"categorySuggestion":"development_tools","subcategorySuggestion":"coding_agents","lifecycleStages":["idea_requirements","product_design","frontend","backend_data"],"audienceIntents":["discover","evaluate","build","operate"],"confidence":0.85,"evidence":[{"claim":"Coffee CLI 是一个基于 Tauri 和 Rust 构建的原生桌面应用，用于运行和管理多个 AI CLI 代理，如 Claude Code、Codex 和 Qwen。","sourceType":"readme","locator":"https://github.com/edison7009/Coffee-CLI"},{"claim":"它支持多标签页会话、会话历史、内置文件浏览器、任务板、代理安装器、远程终端和外部启动。","sourceType":"readme","locator":"https://github.com/edison7009/Coffee-CLI"},{"claim":"界面提供 11 种语言，包括英语、简体中文、繁体中文、日语、韩语等。","sourceType":"readme","locator":"https://github.com/edison7009/Coffee-CLI"},{"claim":"Coffee CLI 旨在消除终端专业知识和英语语言障碍，面向非开发者的专业人士。","sourceType":"readme","locator":"https://github.com/edison7009/Coffee-CLI"},{"claim":"最新版本 v3.4.6 改进了工具执行反馈，并保持所有 11 种语言的翻译同步。","sourceType":"release","locator":"https://github.com/edison7009/Coffee-CLI/releases/tag/v3.4.6"}],"publishedAt":"2026-08-17T13:22:04.426Z"},"github":{"url":"https://github.com/edison7009/Coffee-CLI","ownerAvatarUrl":"https://avatars.githubusercontent.com/u/244278055?v=4","starsCount":169,"forksCount":18,"language":"Rust","syncedAt":"2026-08-17T13:15:56.779Z","license":"AGPL-3.0","defaultBranch":"main","archived":false,"disabled":false,"lastChangedAt":"2026-08-17T12:30:40.000Z","missing":false},"tags":[],"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"]