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:T678a, EvoScientist · 自进化 AI 科学家 · 驾驭 Vibe Research

EvoScientist · 自进化 AI 科学家项目

自进化 AI 科学家，驾驭 Vibe Research

EvoScientist 是一个自进化的 AI 科学家系统：由规划、研究、编码、调试、分析、写作六个智能体组成的团队，依托自进化记忆，把科研工作流跑成一座越用越准的台架，而不是一次性批处理。

项目 · evoscientist/evoscientist 语言 · Python 授权 · Apache-2.0 分类 · AI 智能体 / AI4Science

打开项目站 GitHub 仓库

01 → 02 → 03 → 04 → 05 → 06 LOOP NODE · 上升不改闭合 · 每圈提炼更锐利的方法

VERIFY GATE · 交给人的审视

01接收intake

02规划plan

03执行execute

04评估evaluate

05写作write

06验证verify

active loop · 活动回路 近处带 · 正在运行的回路 六个站点沿带依次经过，持续上升。

memory · 底层记忆带 远处带 · 沉淀的记忆 回路之下的积累，随项目推进不断写入。

LOOP NODE · 升降梯 · 每圈提炼更锐利的方法

SEC 01 · WORKFLOW

六个站点的科研工作流

接收 → 规划 → 执行 → 评估 → 写作 → 验证。螺旋每一次经过同一组站点，却不会回到起点——它持续抬升，把每一次往返所得到的经验，提炼成更清晰的科研方法。

active loop · 活动回路

一条具体的流程带

六个命名的工作流站被一条连续的带子串起来：接收研究需求、规划路径、执行步骤、评估结果、写作产出，最后在验证闸口把成果交给人的审视。整条带子不断抬升，读作一次从未闭合的上升。

intakeplanexecuteevaluatewriteverify

memory · 底层记忆带

回路背后是积累

在活动回路的下方，是沉淀下来的记忆带。它不是一次会话的临时记录，而是随项目推进不断积累的上下文——回路的每一次往复，都从这里取用、也向这里写入。

SEC 02 · TEAM

多智能体的团队构成

协作体由六个专职角色组成，每个角色在流程带上各司其职，连接成一支共同工作的研究小组，而不是一个单一的黑箱模型。

01 · PLANNER规划拆解科研目标，编排任务的先后与依赖。

02 · RESEARCHER研究检索与收集资料，为方法提供依据。

03 · CODER编码把方案写成可运行的代码与实验脚本。

04 · DEBUGGER调试定位错误、修复执行中的故障。

05 · ANALYST分析解读实验结果，抽出趋势与结论。

06 · WRITER写作把结论整理成清晰的研究产出。

collaborating body

共享目标的工作台

六个角色沿流程带分工，却在同一套上下文里协作。它面向科研人员、数据科学家与 AI 开发者，适合科学研究、数据分析与代码生成等需要多人式分工的场景。

科研人员数据科学家AI 开发者

SEC 03 · MEMORY

从记忆到可复用技能

AutoSkills 从系统自身的记忆里提炼出可复用的技能，通过 /autoskills 交由你审查——沉淀的不只是聊天记录，而是提炼后的方法。

accumulated memory

原始记忆

回路运行中积累的完整上下文，是技能提炼的原料。它随时间增长，构成系统理解问题、改进方法的内在基础。

distilled skill

提炼为技能

AutoSkills 把高频、稳定、可复用的经验进一步压缩成技能，并主动提出供你审查。自进化不是即时的魔法，而是需要时间积累的沉淀过程。

SEC 04 · VERIFY

人的验证闸口

在回路顶端，系统把本轮产物交给人类审视——human-on-the-loop，AI 作为研究伙伴与你共同进化，而不是替你自作主张。

本轮产物（写作与验证结果）

闸口亮起时，意味着一个循环把手头的东西交到你手上：由你确认、补正或放行，再进入下一轮上升。

GATE OPEN · 等待人的审视

研究伙伴而非替代者。系统的价值不在于替你决定什么是对的，而在于把可检验的产出递到你面前，让你留在闭环里把控方向。这是它区别于普通批处理任务的关键。

SEC 05 · SURFACES

多界面与多渠道接入

同一套控制逻辑，贯通桌面 WebUI、CLI/TUI 与移动端，并可通过多家 LLM 提供商与常用消息渠道接入。

SURFACE桌面 WebUI浏览器内的图形工作台，适合完整的研究编排。

SURFACECLI / TUI终端内的命令行与界面，归入脚本化流程。

SURFACE移动端轻量接入，随时查看与调度。

channels

同一套逻辑，多渠道到达

通过消息渠道触达同一协作体，选择你所在的入口开始协作。

AnthropicOpenAIGoogleMiniMaxNVIDIA

TelegramSlack飞书微信

SEC 06 · INTEGRATION

调度与集成

支持 cron 风格的定时任务与 MCP 服务器集成，让研究工作可以定时触发、也可以接入更广的工具生态。

定时任务 /schedule 以 cron 表达式编排周期性的研究调度，交给回路在合适的时间自动运转。

MCP 服务器集成 通过 MCP 接入外部工具与数据源，扩展多智能体团队的能力边界。

honest notes

使用前请知悉

需要配置 LLM API 密钥之后才能运转。

自进化能力依赖时间积累，不会立即可见。

部分能力依赖第三方服务与提供商。

以上说明来自项目仓库的描述、功能列表与被验证的工作流设置；能力均以其实际支持为准。

23:T6649, EvoScientist · Self-evolving AI Scientist for Vibe Research

EvoScientist · self-evolving AI scientist project

Self-evolving AI Scientist for Vibe Research

EvoScientist is a self-evolving AI scientist: a team of six agents—plan, research, code, debug, analyze, write—backed by a self-evolving memory. It runs the scientific workflow as a bench that gets sharper every round, not a one-shot batch job.

Project · evoscientist/evoscientist Language · Python License · Apache-2.0 Category · AI agent / AI4Science

Visit project GitHub repository

01 → 02 → 03 → 04 → 05 → 06 LOOP NODE · reopening upward · each pass honing sharper method

VERIFY GATE · handed to human review

01Intakeintake

02Planplan

03Executeexecute

04Evaluateevaluate

05Writewrite

06Verifyverify

active loop · near band The running loop Six stations pass along the band in sequence, ascending.

memory · far band The accumulated context Context layered beneath the loop, written back as the project advances.

LOOP NODE · descent-to-ascent ladder · each pass honing sharper method

SEC 01 · WORKFLOW

A six-station scientific workflow

Intake → plan → execute → evaluate → write → verify. Every lap passes through the same stations yet returns to no starting point—it keeps ascending, distilling each round of experience into a sharper method.

active loop

A concrete flow band

Six named workflow stations are strung together by a continuous band: intake a research request, plan the path, execute steps, evaluate results, write output, and hand the artifact to human review at the verification gate. The band keeps lifting—read as an ascent that never closes.

intakeplanexecuteevaluatewriteverify

underlying memory band

A growing context beneath the loop

Below the active loop sits the accumulated memory band. This is not a throwaway session record but context that builds as the project advances—each pass draws from it and writes back to it.

SEC 02 · TEAM

A multi-agent team

The collaboration is built from six dedicated roles, each holding its place on the flow band and linked into one working group—rather than a single black-box model.

01 · PLANNERPlanBreaks down research goals, ordering tasks and their dependencies.

02 · RESEARCHERResearchRetrieves and gathers sources to ground the method.

03 · CODERCodeTurns the approach into runnable code and experiment scripts.

04 · DEBUGGERDebugLocates errors and fixes faults during execution.

05 · ANALYSTAnalyzeReads results, drawing out trends and conclusions.

06 · WRITERWriteCompiles conclusions into a clear research artifact.

collaborating body

A bench with a shared objective

Six roles divide the flow yet collaborate within one context. Built for researchers, data scientists, and AI developers, it suits scientific research, data analysis, and code generation where a division-of-labour team earns its keep.

ResearchersData scientistsAI developers

SEC 03 · MEMORY

From memory to reusable skills

AutoSkills distills reusable skills from the system's own memory, proposed for review via /autoskills—what accumulates is not raw chat but distilled method.

accumulated memory

Raw memory

The full context gathered over the loop's runs is the raw material for skill distillation. It grows over time, forming the inner basis on which the system understands problems and refines its method.

distilled skill

Distilling into skills

AutoSkills further compresses recurring, stable, reusable experience into skills and actively proposes them for review. Self-evolution is not instant magic—it is a build-up that pays off over time.

SEC 04 · VERIFY

A human-on-the-loop verification gate

At the top of the loop, the system hands this round's artifact to human review—human-on-the-loop, where the AI works as a research partner evolving with you rather than deciding alone.

This round's artifact (written and verified output)

When the gate lights, one cycle has handed its work to you: confirm, amend, or release it before the next ascent begins.

GATE OPEN · awaiting human review

A research partner, not a replacement. The system's value is not in deciding what is correct for you but in placing testable output in front of you, keeping you on the loop to steer. That is what sets it apart from a plain batch job.

SEC 05 · SURFACES

Multiple surfaces and channels

One control logic runs through desktop WebUI, CLI/TUI, and mobile, and reaches you through several LLM providers and everyday messaging channels.

SURFACEDesktop WebUIA graphical workbench in the browser, fit for full research orchestration.

SURFACECLI / TUIA command-line and terminal interface for scripted workflows.

SURFACEMobileA lightweight way to stay connected, review, and schedule.

channels

One logic, many entry points

Reach the same collaboration through messaging channels—pick the doorway you work from.

AnthropicOpenAIGoogleMiniMaxNVIDIA

TelegramSlackFeishuWeChat

SEC 06 · INTEGRATION

Scheduling and integrations

Support for cron-style scheduled tasks and MCP server integration lets research run on a timer and plug into a wider tool ecosystem.

Scheduled tasks /schedule Orchestrate recurring research with cron expressions and let the loop run on its own at the right time.

MCP server integration Connect external tools and data sources over MCP to extend the team's capabilities.

honest notes

Read before you start

Requires LLM API keys to operate.

Self-evolution depends on time; gains are not immediate.

Some capabilities rely on third-party services and providers.

This description draws on the project's repository description, feature set, and verified workflow configuration; capabilities are subject to their actual support.

24:T4a2,The project's defining, most evidence-backed mechanism is the six-station scientific workflow (intake → plan → execute → evaluate → write → verify) combined with self-evolution. Direction A renders that exact workflow as the page's structural spine-an ascending loop of six named nodes with a verification gate at the top-and it lets self-evolution be read as the loop never closing and distilling into sharper method, matching the auto-skills claim. Its composition is a genuine derived diagram (evidence-driven, content-bearing nodes) rather than decorative geometry, and it supports the human-on-the-loop point naturally at the verify gate. Direction B (growth rings) expresses accumulation/evolution well but is more abstract and risks reading as decoration; Direction C (constellation) foregrounds the multi-agent team but treats the workflow as a side note, which under-represents the project's flagship 'self-evolving scientific workflow'. Direction A best fits the evidence and produces the most distinctive composition, and because material.json declares no local assets it is implemented purely with type- and coordinate-driven stations in place of any media region.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","cmt2zgzvt00i46271hmxbiqvk:zh-CN:beautified",{"project":{"id":"cmt2zgzvt00i46271hmxbiqvk","slug":"evoscientist-evoscientist","name":"evoscientist","description":"自进化 AI 科学家，驾驭 Vibe Research","authorName":"evoscientist","authorVerification":{"verified":false,"verifiedAt":null},"category":"development_tools","subcategory":"data_ai_tools","experienceUrl":"https://evoscientist.ai/","qualityScore":87,"featured":false,"officialSite":null,"views":{"original":{"available":true},"official":{"available":false,"reason":"disabled_by_policy"},"beautified":{"available":true,"reason":null},"defaultView":"beautified"},"beautifiedView":{"revisionId":"cmt2zgzwe00ig6271j7xannkx","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":"research-loop-spiral","name":"The Research Loop","projectMetaphor":"A single, never-closing loop of the scientific workflow (intake → plan → execute → evaluate → write → verify). Each lap passes through the same six stations but never returns to its start; it advances, so every pass distills what was learned into sharper method. The machine is not a batch job-it is a bench that gets better each round.","signatureVisual":"An ascending spiral literally traced by the six workflow stations. Each station is a numbered lab node (intake, plan, execute, evaluate, write, verify) linked in a continuous band; the band does not close, it lifts with every cycle, and a short 'verification gate' ticks at the top where the lap hands its artifact to the human-on-the-loop.","heroGeometry":"A single ascending logarithmic spiral sweeping in from the lower left toward an upper-right focus; the six workflow stations sit on the spiral turns, the verify gate capping the ascent.","sectionSequence":["the-loop-stations","composition-of-team","memory-distillation","verification-gate","surfaces-and-channels","integration-options"],"surfaceModel":"Continuous planes: a near-field band (the active loop) sits on a far-field band (the accumulated memory behind it). Sections move along the loop's path rather than as static stacked blocks; each section is anchored to a station or to one of the spiral's two bands.","palette":"Deep ink laboratory ground; one warm accent (specimen amber) for the moving loop against cool anodized metal greys for memory/context; a single signal colour reserved for human verification. Faint faint-grey ruled tracing where the spiral passes.","typography":"Scientific-lab character: a precise mono for station numbers and coordinates, a tight grotesque for headings, modest measured body. Numeric tokens ('01'-'06') double as both phase labels and coordinate marks.","desktopComposition":"1280px: the hero is the spiral itself spanning the full viewport with stations positioned along its arc from below-left to upper-right, verify gate near the top. Later sections follow the loop's path left-to-right and bottom-to-top; the team and memory sections inherit the two-band surface, the integration section closes the loop path.","mobileTransformation":"390px: the spiral unrolls into a single vertical rail. Stations become sequential numbered checkpoints stacked one above the other, connected by a continuous vertical line-the loop is read as a descent-to-ascent ladder rather than a horizontal arc. Two-band depth becomes stacked plates.","mediaStrategy":"No local assets exist in the artifact manifest, so no media region is built. The signature visual is derived and type-and-coordinate driven: station nodes and numeric phase tokens carry the evidence; no product screenshot, dashboard, or remote image is simulated.","forbiddenPatterns":["no marketing hero with a tagline plus a row of three metric cards","no repeatable feature-card wall that could hold any project's copy","no fake terminal, browser, phone, dashboard, or editor shell","no decorative glow/particle/grid field standing in for the idea"],"compositionFingerprint":"spiral-of-six-station-nodes-ascending-with-verify-gate-on-one-anodized-surface"},{"id":"growth-ring-organism","name":"The Growth Ring","projectMetaphor":"A scientist that grows like an organism rather than running jobs. Each lap of experience deposits a new ring from the inside out: raw memory on the inner rings, distilled AutoSkills on the outer rings. Evolution is not a feature-it is the accumulation that visibly thickens the cross-section.","signatureVisual":"Concentric incised growth rings seen in cross-section, read from a small inner nucleus (memory) outward to broad outer rings (distilled skills). Thinner outer arcs are newly deposited skill layers; short 'check' notches mark where a layer was verified before it was allowed to harden.","heroGeometry":"A cluster of concentric circles with an off-centre focus-the nucleus sits toward one edge so the rings read as growing toward the viewer's field rather than a symmetric bullseye.","sectionSequence":["the-nucleus","auto-skill-distillation","the-agent-roster","scheduled-cycles","open-interfaces","human-on-the-loop"],"surfaceModel":"Radial depth: page surfaces are concentric plates that recede toward the nucleus focus. Content sits on ring-bands; interaction feels like moving inward (toward raw memory) or outward (toward shipped skill).","palette":"Monochrome sediment tones for the growth substrate, with each newly distilled skill ring scored in a single pigment so the newest layer visibly stands apart from aged inner rings. Sparse warm tint only on the active nucleus.","typography":"Field-log character: a wide, calm serif for the headings (a naturalist's ledger) paired with a small mono for ring index numbers and layer coordinates.","desktopComposition":"1280px: the hero is the concentric rings centred-left, filling the viewport; the nucleus near the bottom-left, outer rings sweeping up and right. Subsequent sections run as ring-band tracks, moving progressively from the nucleus outward across the viewport.","mobileTransformation":"390px: concentric composition re-centres vertically: rings collapse into stacked horizontal plates from bottom (nucleus/memory) to top (outer skill), progressing in a single upward scroll; the off-centre focus becomes a centred seed at the bottom.","mediaStrategy":"No local assets exist, so no media region is built. The derived cross-section (rings + check notches + ring indices) is a pure relationship diagram of memory-to-skill accumulation; no screenshot or remote image is constructed.","forbiddenPatterns":["no three-stat metric cards under a slogan","no feature-card wall that accepts any project's text","no phone/browser/terminal/dashboard/editor shell","no decorative gradient or particle glow replacing the concept"],"compositionFingerprint":"concentric-growth-rings-with-off-centre-nucleus-and-check-notch-verification"},{"id":"constellation-convergence","name":"The Constellation","projectMetaphor":"A distributed colony of specialist scientists converging on a shared finding. Six disciplines hold separate stations and are connected into one collaborating body; verifiable conclusions converge at the centre node, where the human stays on the loop.","signatureVisual":"A node-link constellation: six specialist nodes (plan, research, code, debug, analyze, write) arranged in orbit, their links weaving inward to a single central convergence node labelled the verified finding. The human keeps an opened link into the centre.","heroGeometry":"A centrality hub: an inner finding node with specialist nodes arranged around it, linked by tensioned chords rather than a strict circle.","sectionSequence":["the-convergent-heart","six-disciplines-in-orbit","shared-context-web","multi-signal-arrival","client-openings","published-findings"],"surfaceModel":"Connected fields: content lives on web/facet zones tethered to the node-graph; sections feel like faces of a shared workspace around the centre rather than a straight linear stack.","palette":"Cool connectivity palette: a dark field held by thin luminous-blue-tinted chords and neutral nodes; the centre conclusion node is the only warm element, drawing the eye inward.","typography":"Instrument-rack character: a crisp technical sans for headings, metrics, and chord metadata pairing; compact labels track discipline names and their signals.","desktopComposition":"1280px: the hero is the full node constellation across the desktop field, specialist nodes orbiting an inset central conclusion and the human link emphasised. Sections fan outward as constellation faces, ending at the published-findings node.","mobileTransformation":"390px: the orbiting constellation collapses into a hub-and-spoke vertical list-the convergence node on top, each specialist discipline a spoke row below draining into it; the field read becomes columnar rather than circular.","mediaStrategy":"No local assets exist, so no media region is built. The derived node-link diagram is a pure relationship diagram of team-role-to-conclusion; no screenshot, chart, or remote image is simulated.","forbiddenPatterns":["no slogan plus a row of three metric cards","no interchangeable feature-card grid","no terminal/browser/phone/dashboard/editor shell","no decorative particle or glow field standing in for the colony concept"],"compositionFingerprint":"orbit-node-constellation-collapsing-into-a-central-verified-finding-with-human-link"}],"selectedDirectionId":"research-loop-spiral","selectionRationale":"$24"},"designBrief":{"schemaVersion":"dodot-design-brief-v1","selectedDirectionId":"research-loop-spiral","projectMetaphor":"A single, never-closing loop of the scientific workflow (intake → plan → execute → evaluate → write → verify). Each lap passes through the same six stations but never returns to its start; it advances, so every pass distills what was learned into sharper method. The machine is not a batch job-it is a bench that gets better each round.","signatureVisual":"An ascending spiral literally traced by the six workflow stations. Each station is a numbered lab node (intake, plan, execute, evaluate, write, verify) linked in a continuous band; the band does not close, it lifts with every cycle, and a short 'verification gate' ticks at the top where the lap hands its artifact to the human-on-the-loop.","compositionFingerprint":"spiral-of-six-station-nodes-ascending-with-verify-gate-on-one-anodized-surface","heroGeometry":"A single ascending logarithmic spiral sweeping in from the lower left toward an upper-right focus; the six workflow stations sit on the spiral turns, the verify gate capping the ascent.","sectionSequence":["the-loop-stations","composition-of-team","memory-distillation","verification-gate","surfaces-and-channels","integration-options"],"surfaceModel":"Continuous planes: a near-field band (the active loop) sits on a far-field band (the accumulated memory behind it). Sections move along the loop's path rather than as static stacked blocks; each section is anchored to a station or to one of the spiral's two bands.","palette":"Deep ink laboratory ground; one warm accent (specimen amber) for the moving loop against cool anodized metal greys for memory/context; a single signal colour reserved for human verification. Faint faint-grey ruled tracing where the spiral passes.","typography":"Scientific-lab character: a precise mono for station numbers and coordinates, a tight grotesque for headings, modest measured body. Numeric tokens ('01'-'06') double as both phase labels and coordinate marks.","desktopComposition":"1280px: the hero is the spiral itself spanning the full viewport with stations positioned along its arc from below-left to upper-right, verify gate near the top. Later sections follow the loop's path left-to-right and bottom-to-top; the team and memory sections inherit the two-band surface, the integration section closes the loop path.","mobileTransformation":"390px: the spiral unrolls into a single vertical rail. Stations become sequential numbered checkpoints stacked one above the other, connected by a continuous vertical line-the loop is read as a descent-to-ascent ladder rather than a horizontal arc. Two-band depth becomes stacked plates.","mediaStrategy":"No local assets exist in the artifact manifest, so no media region is built. The signature visual is derived and type-and-coordinate driven: station nodes and numeric phase tokens carry the evidence; no product screenshot, dashboard, or remote image is simulated.","evidencePlan":["hero spiral carries the six named workflow stations as actual labelled nodes, sourced from the scientific-workflow claim (intake → plan → execute → evaluate → write → verify)","the composition-of-team section maps planner, researcher, coder, debugger, analyst, writer from the multi-agent-team claim","memory-distillation section renders the auto-skills memory-to-reusable-skill relationship from the AutoSkills claim; verification-gate and human-on-the-loop section are sourced from the human-on-the-loop claim","surfaces-and-channels section lists the desktop WebUI, CLI/TUI, mobile, and channels (Telegram, Slack, Feishu, WeChat) from the multi-surface and multi-channel claims","integration-options section names scheduled tasks (cron) and MCP server integration from the scheduled-tasks claim; caveats (LLM API keys required, evolution needs time) are shown honestly, not hidden"],"localeContract":{"zh-CN":"以科研工作台口吻陈述：'自进化 AI 科学家，驾驭 Vibe Research'。六个流程站（接收→规划→执行→评估→写作→验证）逐站展示；团队、记忆提炼、验证闸口、多界面、集成各配自然中文说明。语气客观克制、贴近实验室/数据科学家/研究者的视角。","en":"Instrument-rack tone: 'Self-evolving AI Scientist for Vibe Research'. Six workflow stations (intake → plan → execute → evaluate → write → verify) shown station by station; team, memory distillation, verification gate, surfaces, and integrations each get natural English copy. Voice is precise and lab-like, aimed at researchers, data scientists, and AI developers."},"forbiddenPatterns":["no marketing hero with a tagline plus a row of three metric cards","no repeatable feature-card wall that could hold any project's copy","no fake terminal, browser, phone, dashboard, or editor shell","no decorative glow/particle/grid field standing in for the idea","no fake screenshots, charts, logos, testimonials, or remote images"]},"visualReview":{"schemaVersion":"dodot-automated-design-review-v1","status":"approved","summary":"页面忠实实现了 design-brief 选中的 research-loop-spiral 方向。signature visual 为真实派生 SVG：六命名科研工作流站点（intake→plan→execute→evaluate→write→verify）沿上升不锈钢带排布、顶端以信号绿验证闸口封顶，双带表面模型（活动回路 + 底层记忆）、深墨实验地、琥珀活动/金属灰记忆/信号绿人类验证的调色板均由站点数字相位 token 承载，材料全部有据。六段章节与 brief sectionSequence 完全一致，文案逐条对应 material.json 真实 claim，双语等价且自然，未回退到通用 Hero/三卡/卡片墙/设备壳/装饰性网格，无假控件。移动端（820px 以内）正确实现项目专属重构：螺旋 SVG 隐藏，替换为 signal 绿 VERIFY GATE 于顶、01-06 数字检查点沿垂直连续线上下堆叠成升降梯、双带转为堆叠层板，符合 brief mobileTransformation。七项检查全部通过，无 blocking issue，判 approved。","checks":{"projectFit":{"passed":true,"evidence":"选中方向为 research-loop-spiral（design-directions.json selectedDirectionId 与 design-brief.json selectedDirectionId 一致）。页面采用上升螺旋 signature visual，六站点 intake→plan→execute→evaluate→write→verify 依弧线排布、顶端信号绿验证闸口封顶，并被 01-06 数字 token 标注（zh-CN.html 与 en.html 的 spiral SVG 同构）。sectionSequence 完整复现：the-loop-stations / composition-of-team / memory-distillation / verification-gate / surfaces-and-channels / integration-options，共六段（qa-report.json 两语均 sectionCount=6）。双带表面、调色板与 mono 站点数字/无衬线标题字体均对齐 brief，无偏离。"},"evidenceBound":{"passed":true,"evidence":"六站点工作流来自 scientific-workflow claim；六角色团队（PLANNER/RESEARCHER/CODER/DEBUGGER/ANALYST/WRITER）来自 multi-agent-team claim；AutoSkills 记忆→可复用技能来自 AutoSkills claim（含 /autoskills 审查）；验证闸口与 human-on-the-loop 来自对应 claim；界面（桌面 WebUI/CLI-TUI/移动端）与渠道（Telegram/Slack/飞书/微信）及提供商（Anthropic/OpenAI/Google/MiniMax/NVIDIA）来自 multi-surface/multi-channel claim；cron 定时任务与 MCP 来自 scheduled-tasks claim；caveat（需 LLM API 密钥、进化需时间、部分依赖第三方）如实展示未隐藏。仓库元数据（Python/Apache-2.0/evoscientist/evoscientist）及主页/仓库链接均取自 material.json，无误植无捏造。"},"signatureVisual":{"passed":true,"evidence":"螺旋为真实派生 SVG：含六个带编号（01-06）与命名标签的站点圆节点沿上升含氧带排布、琥珀色连续 band（不闭合）、金属灰底层记忆 band、顶端信号绿验证闸口节点与 VERIFY GATE 标签及 06 终点坐标。节点与坐标直接承载六个工作流站点证据，属 type-and-coordinate driven 的关系图而非装饰性几何。qa-report.json 两语桌面 brokenImageCount=0、无 consoleErrors、无 networkRequests，该视觉真实可渲染。"},"mobileTransformation":{"passed":true,"evidence":"max-width:820px 断点中 .spiral{display:none} 且 .rail-mobile{display:block}，移动端并未保留横向弧线。rail-mobile 将螺旋展开为单条垂直轨：.rail::before 提供连续垂直连线，信号绿 .rail-gate（VERIFY GATE · 交给人的审视）置于顶部，六个 .rail-stat 检查点（01-06）自上而下顺序编号堆列成升降梯；.rail-plates 以 .plate（active loop·近处带）与 .plate.mem（memory·远处带）偏移叠层把双带深度转为堆叠层板。与 brief mobileTransformation（螺旋退为垂直轨、站点按序编号检查点、两带堆叠为层板）逐点吻合。qa-report.json 390 视口 scrollWidth=390/clientWidth=390、无 overflow，两语均通过。"},"localeParity":{"passed":true,"evidence":"zh-CN.html 与 en.html 章节结构、数量与顺序完全一致：均 1 个 h1、6 个 h2、6 个 section（qa-report.json 两语 1280 视口 h2Count=6 sectionCount=6）。中文为科研工作台口吻（以科研工作台口吻）、英文为 instrument-rack 口吻，逐段语义对等且各自自然；六站点、六角色、界面/渠道/提供商、caveat 均一一对应，无缺页、错位或机翻生硬。"},"antiTemplate":{"passed":true,"evidence":"未出现 slogan+一行三张指标卡的 Hero：hero 顶部 chips 为仓库事实标签（项目名/语言/授权/分类），不含 stars/forks 等数字指标卡。无通用三卡或可替换 feature-card 墙：roster/surface/integration 卡片均承载具名证据（六个具名角色、三种具名界面、两种具名集成、具名渠道/提供商）。无假终端/浏览器/手机/仪表盘/编辑器设备壳（WebUI/CLI-TUI/Mobile 为文本描述非设备壳）。无装饰渐变/粒子/网格取代概念：.rule-trace 细网格与 .sheet 径向微光均为 brief palette 明确的 faint ruled tracing 与实验室地景辅助元素，真正的表意由螺旋与节点承担，主旨未由装饰场替代。"},"noFakeUi":{"passed":true,"evidence":"qa-report.json 全部四个视图（zh-CN/en × 1280/390）fakeControlCount=0、emptyLinkCount=0、brokenImageCount=0、missingAltCount=0。CTA 均为真实可点链接（cta 指向 https://evoscientist.ai/，cta-ghost 指向 GitHub 仓库）；VERIFY GATE · GATE OPEN 为状态标签非控件；无假按钮、假图表、假仪表盘或假终端位。networkRequests 为空，无远程模拟 UI。"}},"issues":[]},"repairEvidence":[{"attempt":1,"before":{"qaPassed":true,"designReviewApproved":false,"qaEvidence":{"zh-CN":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":2394,"cjkCount":1290,"latinLetterCount":596,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":5124,"h1Top":123.390625,"bottomInset":475.515625,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"fb33fb6b97904448a326bf143d70ecafbae57731e5ef1e4e743004ef90f3be7a","middle":"ef52927335aa991907cf8fcd3cc539b81e23b079a5391f1be016a7683475528f","bottom":"5d627a66f17e126883c240d485865001742afa292e3ca50202ff3b6a0b875ab6"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":2394,"cjkCount":1290,"latinLetterCount":596,"scrollWidth":390,"clientWidth":390,"scrollHeight":7301,"h1Top":118.34375,"bottomInset":460.859375,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"3655b04507991e97178d6d517395f4d449d0fb0d9a845419e714a401203d4495","middle":"b2ebbac9d0534d56aef596515c9ad570840e28c9836b8c924a7fd5639d8ee7de","bottom":"0f0f0e8399fbadb36cabfadb5f2a5d33a4b75a357818cbc6d3f2a8d40fdce200"},"consoleErrors":[],"networkRequests":[]}],"en":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":5478,"cjkCount":0,"latinLetterCount":4365,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":5549,"h1Top":123.390625,"bottomInset":569.203125,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"cb97a8cf92e433c77b877944f5bdd12aa56bac50815e87b7fa4fa332df56e854","middle":"cab476646cd0df7e3b0dfbde469af8663a6e609dda6bb8b7d6fbc5404b2655ed","bottom":"79b840256a67fbf5dbca9b746068b5ec0a57a452bb049141405e0e52296dd253"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":5478,"cjkCount":0,"latinLetterCount":4365,"scrollWidth":390,"clientWidth":390,"scrollHeight":8162,"h1Top":138.734375,"bottomInset":522.90625,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"38e23ead748f4c17637bf80cbb69a682f9f4b2ddcc6b2c04e083b90db622a6dc","middle":"e2b20c0f60bb59e7a6ba90d4a4079b583af61a054c064b671f705eb5a28c83b9","bottom":"79447708dbd4cefe44f80b746b5fb89aade0a36576f8873f1dd2bfb7580d34c2"},"consoleErrors":[],"networkRequests":[]}]},"visualReview":{"schemaVersion":"dodot-automated-design-review-v1","status":"revise","summary":"页面忠实实现了 'research-loop-spiral' 方向：以六个命名科研工作流站点组成的上升螺旋作为 signature visual，顶部验证闸口，双带表面模型（活动回路 + 底层记忆），琥珀色活动环与信号绿人类验证、冷却金属灰记忆的调色板，视觉全部由站点点位与数字相位 token 承载且材料有据。六段章节与 brief sectionSequence 完全一致，每段文案均对应 material.json 中的真实 claim，双语内容等价且自然，通用 Hero/三卡/卡片墙/设备壳/装饰性渐变网格均未出现，无假控件。唯一不符合项是移动端重构：brief 明确要求 390px 下 '螺旋展开为单条垂直轨，站点自上而下列叠为连续编号检查点，两带深度堆叠为层板'，但实现中螺旋 SVG 仅是等比缩放仍保持横向弧线排布，未展开成垂直编号检查点轨迹，未尽到项目专属的移动端重构。因此判为 revise。","checks":{"projectFit":{"passed":true,"evidence":"选中方向为 research-loop-spiral（design-directions.json selectedDirectionId 与 design-brief.json 一致）。页面采用上升螺旋 signature visual，六站点 intake→plan→execute→evaluate→write→verify 依弧线排布、顶端验证闸口（zh-CN.html svg 见 191-227 行，en.html 同构）。sectionSequence 完整复现：the-loop-stations / composition-of-team / memory-distillation / verification-gate / surfaces-and-channels / integration-options（zh-CN.html 230/255/283/305/325/357 行）。双带表面模型、调色板（深墨实验室底、琥珀活动环、金属灰记忆、信号绿人类验证）与字体（站点 mono 数字 token、无衬线标题）均对齐 brief，无偏离。"},"evidenceBound":{"passed":true,"evidence":"六站点工作流来自 scientific-workflow claim（material.json summary.claim[5]）；六角色团队（PLANNER/RESEARCHER/CODER/DEBUGGER/ANALYST/WRITER）来自 multi-agent-team claim；AutoSkills 记忆→可复用技能来自 AutoSkills claim；验证闸口/human-on-the-loop 来自相应 claim；界面（桌面 WebUI/CLI-TUI/移动端）与渠道（Telegram/Slack/飞书/微信）及提供商（Anthropic/OpenAI/Google/MiniMax/NVIDIA）来自 multi-surface/multi-channel claim；cron 定时任务与 MCP 来自 scheduled-tasks claim；caveat（需 LLM API 密钥、进化需时间、部分依赖第三方）如实展示未隐藏。仓库元数据（Python/Apache-2.0/evoscientist/evoscientist）与主页/仓库链接均取自 material.json，无误植。"},"signatureVisual":{"passed":true,"evidence":"螺旋为真实派生 SVG：含 six 个带编号（01-06）与命名标签的站点圆节点沿上升带排布、琥珀色连续 band（band 不闭合）、金属灰底层 songsoft 记忆带、顶端信号绿验证闸口节点与 'VERIFY GATE' 标签及 '06' 终点坐标。它直接承载六个工作流站点证据（type-and-coordinate driven），非装饰性几何。qa-report.json 桌面 brokenImageCount=0，Safari/Chrome 均无 consoleErrors 与 networkRequests，visual 真实可渲染。"},"mobileTransformation":{"passed":false,"evidence":"370/390 视口仅命中 max-width:820px 断点把 .two-band/.gate/.integ/.roster/.surf-list 折叠为单列、sec-head 与 hero-caption 纵向，但螺旋 SVG 仅以 width:100% 等比缩放仍保持 1080x440 的横向弧线六站点布局，未出现 brief 要求的 '螺旋展开为单条垂直轨，站点自上而下顺序编号检查点垂直连接，两带深度堆叠为层板' 的项目专属重构。mobileTransformation 未实现，页面在 390px 上退回为通用纵向堆叠。"},"localeParity":{"passed":true,"evidence":"zh-CN.html 与 en.html 章节结构、数量与顺序完全一致：均 1 个 h1、6 个 h2、6 个 section（qa-report.json zh-CN width 1280 h2Count=6 sectionCount=6；en width 1280 h2Count=6 sectionCount=6）。中文为科研工作台口吻（'以科研工作台口吻陈述'）、英文为 instrument-rack 口吻，逐段语义对等且各自自然；同一数据（六站点、六角色、界面/渠道/提供商、caveat）在两语中一一对应，无缺页或错位。"},"antiTemplate":{"passed":true,"evidence":"未出现 'slogan+一行三张指标卡' 的 Hero：hero 顶部 chips 为仓库事实标签（项目名/语言/授权/分类），不呈现数字指标或统计卡，无 stars/forks/metric 数字。无通用三卡或可替换 feature-card 墙：roster/surface/integration 的卡片均承载具名证据（六个具名角色、三种具名界面、两种具名集成方式、具名渠道），非替换任何项目的文案。无假终端/浏览器/手机/仪表盘/编辑器设备壳（'Desktop WebUI/CLI-TUI/Mobile' 为文本描述非设备壳）。无装饰性渐变/粒子/网格场取代概念：.rule-trace 细网格与 .sheet 径向微光均为 brief palette 明确的 'faint ruled tracing' 与实验室地景辅助元素，真正的表意由螺旋承担；feature 主旨未由装饰场替代。"},"noFakeUi":{"passed":true,"evidence":"qa-report.json 全部 4 个视图（zh-CN/en × 1280/390）fakeControlCount=0、emptyLinkCount=0、brokenImageCount=0。CTA 均为真实可点链接（cta 指向 https://evoscientist.ai/，cta-ghost 指向 GitHub 仓库）；'GATE OPEN · 等待人的审视' 为状态标签非控件；无假按钮、假图表、假 dashboard 或假终端。networkRequests 为空，无远程模拟 UI。"}},"issues":[{"code":"SPIRAL_NOT_UNROLLED_ON_MOBILE","severity":"blocking","locale":"all","description":"brief 的 mobileTransformation 明确要求 390px 下螺旋展开为单条垂直轨，六站点自上而下堆叠为连续编号检查点并由垂直连线相连，两带深度堆叠为层板。当前实现仅靠 max-width:820px 通用断点折列与缩放螺旋 SVG，螺旋仍保持横向弧线排布，未实现这一项目专属的移动端重构。","repairInstruction":"在移动端（390px）以移动专用表现重排螺旋：将两个 SVG 站点组（或新增移动专用轨道容器）改为一条垂直连线串联的站点列表，站点带编号（01-06）与命名标签自上而下纵向排列，顶端保留信号绿 VERIFY GATE；将视图中的 active-loop 与 memory 双层以堆叠 plate 呈现，保证 390px 读作垂直升降梯而非横向弧线。"}]}},"after":{"qaPassed":true,"designReviewApproved":true,"qaEvidence":{"zh-CN":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":2394,"cjkCount":1290,"latinLetterCount":596,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":5124,"h1Top":123.390625,"bottomInset":475.515625,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"fb33fb6b97904448a326bf143d70ecafbae57731e5ef1e4e743004ef90f3be7a","middle":"ef52927335aa991907cf8fcd3cc539b81e23b079a5391f1be016a7683475528f","bottom":"5d627a66f17e126883c240d485865001742afa292e3ca50202ff3b6a0b875ab6"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":2477,"cjkCount":1350,"latinLetterCount":618,"scrollWidth":390,"clientWidth":390,"scrollHeight":7710,"h1Top":118.34375,"bottomInset":460.796875,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"c50fe921906492a23fb8abdfa7c6daf4263da6e83d703fb61a7d194405afc298","middle":"abb6d02faa9ecaf82a6b6538f3fd6562ed59caa66c21fe7f7f76e1d79273d64c","bottom":"c2b78b53ba67b1a4c3288fb5dea21b3f41f0024256d568869359bd88797ada55"},"consoleErrors":[],"networkRequests":[]}],"en":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":5478,"cjkCount":0,"latinLetterCount":4365,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":5549,"h1Top":123.390625,"bottomInset":569.203125,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"cb97a8cf92e433c77b877944f5bdd12aa56bac50815e87b7fa4fa332df56e854","middle":"cab476646cd0df7e3b0dfbde469af8663a6e609dda6bb8b7d6fbc5404b2655ed","bottom":"79b840256a67fbf5dbca9b746068b5ec0a57a452bb049141405e0e52296dd253"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":5739,"cjkCount":0,"latinLetterCount":4597,"scrollWidth":390,"clientWidth":390,"scrollHeight":8644,"h1Top":138.734375,"bottomInset":522.78125,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"3038fd52788d703c9f2f83a3de6eb9b9ee7435c7b8a8ab7b8392d9808614c74f","middle":"6a954b2a9d6977abee0ec7358caeab911bfaee6517ae224ba5fd8bdd214ee23a","bottom":"4e424f4bcd9c7b0b174bdfa5531abdeea400bc1d9b7c8d79fe3cae3bfe997c48"},"consoleErrors":[],"networkRequests":[]}]},"visualReview":{"schemaVersion":"dodot-automated-design-review-v1","status":"approved","summary":"页面忠实实现了 design-brief 选中的 research-loop-spiral 方向。signature visual 为真实派生 SVG：六命名科研工作流站点（intake→plan→execute→evaluate→write→verify）沿上升不锈钢带排布、顶端以信号绿验证闸口封顶，双带表面模型（活动回路 + 底层记忆）、深墨实验地、琥珀活动/金属灰记忆/信号绿人类验证的调色板均由站点数字相位 token 承载，材料全部有据。六段章节与 brief sectionSequence 完全一致，文案逐条对应 material.json 真实 claim，双语等价且自然，未回退到通用 Hero/三卡/卡片墙/设备壳/装饰性网格，无假控件。移动端（820px 以内）正确实现项目专属重构：螺旋 SVG 隐藏，替换为 signal 绿 VERIFY GATE 于顶、01-06 数字检查点沿垂直连续线上下堆叠成升降梯、双带转为堆叠层板，符合 brief mobileTransformation。七项检查全部通过，无 blocking issue，判 approved。","checks":{"projectFit":{"passed":true,"evidence":"选中方向为 research-loop-spiral（design-directions.json selectedDirectionId 与 design-brief.json selectedDirectionId 一致）。页面采用上升螺旋 signature visual，六站点 intake→plan→execute→evaluate→write→verify 依弧线排布、顶端信号绿验证闸口封顶，并被 01-06 数字 token 标注（zh-CN.html 与 en.html 的 spiral SVG 同构）。sectionSequence 完整复现：the-loop-stations / composition-of-team / memory-distillation / verification-gate / surfaces-and-channels / integration-options，共六段（qa-report.json 两语均 sectionCount=6）。双带表面、调色板与 mono 站点数字/无衬线标题字体均对齐 brief，无偏离。"},"evidenceBound":{"passed":true,"evidence":"六站点工作流来自 scientific-workflow claim；六角色团队（PLANNER/RESEARCHER/CODER/DEBUGGER/ANALYST/WRITER）来自 multi-agent-team claim；AutoSkills 记忆→可复用技能来自 AutoSkills claim（含 /autoskills 审查）；验证闸口与 human-on-the-loop 来自对应 claim；界面（桌面 WebUI/CLI-TUI/移动端）与渠道（Telegram/Slack/飞书/微信）及提供商（Anthropic/OpenAI/Google/MiniMax/NVIDIA）来自 multi-surface/multi-channel claim；cron 定时任务与 MCP 来自 scheduled-tasks claim；caveat（需 LLM API 密钥、进化需时间、部分依赖第三方）如实展示未隐藏。仓库元数据（Python/Apache-2.0/evoscientist/evoscientist）及主页/仓库链接均取自 material.json，无误植无捏造。"},"signatureVisual":{"passed":true,"evidence":"螺旋为真实派生 SVG：含六个带编号（01-06）与命名标签的站点圆节点沿上升含氧带排布、琥珀色连续 band（不闭合）、金属灰底层记忆 band、顶端信号绿验证闸口节点与 VERIFY GATE 标签及 06 终点坐标。节点与坐标直接承载六个工作流站点证据，属 type-and-coordinate driven 的关系图而非装饰性几何。qa-report.json 两语桌面 brokenImageCount=0、无 consoleErrors、无 networkRequests，该视觉真实可渲染。"},"mobileTransformation":{"passed":true,"evidence":"max-width:820px 断点中 .spiral{display:none} 且 .rail-mobile{display:block}，移动端并未保留横向弧线。rail-mobile 将螺旋展开为单条垂直轨：.rail::before 提供连续垂直连线，信号绿 .rail-gate（VERIFY GATE · 交给人的审视）置于顶部，六个 .rail-stat 检查点（01-06）自上而下顺序编号堆列成升降梯；.rail-plates 以 .plate（active loop·近处带）与 .plate.mem（memory·远处带）偏移叠层把双带深度转为堆叠层板。与 brief mobileTransformation（螺旋退为垂直轨、站点按序编号检查点、两带堆叠为层板）逐点吻合。qa-report.json 390 视口 scrollWidth=390/clientWidth=390、无 overflow，两语均通过。"},"localeParity":{"passed":true,"evidence":"zh-CN.html 与 en.html 章节结构、数量与顺序完全一致：均 1 个 h1、6 个 h2、6 个 section（qa-report.json 两语 1280 视口 h2Count=6 sectionCount=6）。中文为科研工作台口吻（以科研工作台口吻）、英文为 instrument-rack 口吻，逐段语义对等且各自自然；六站点、六角色、界面/渠道/提供商、caveat 均一一对应，无缺页、错位或机翻生硬。"},"antiTemplate":{"passed":true,"evidence":"未出现 slogan+一行三张指标卡的 Hero：hero 顶部 chips 为仓库事实标签（项目名/语言/授权/分类），不含 stars/forks 等数字指标卡。无通用三卡或可替换 feature-card 墙：roster/surface/integration 卡片均承载具名证据（六个具名角色、三种具名界面、两种具名集成、具名渠道/提供商）。无假终端/浏览器/手机/仪表盘/编辑器设备壳（WebUI/CLI-TUI/Mobile 为文本描述非设备壳）。无装饰渐变/粒子/网格取代概念：.rule-trace 细网格与 .sheet 径向微光均为 brief palette 明确的 faint ruled tracing 与实验室地景辅助元素，真正的表意由螺旋与节点承担，主旨未由装饰场替代。"},"noFakeUi":{"passed":true,"evidence":"qa-report.json 全部四个视图（zh-CN/en × 1280/390）fakeControlCount=0、emptyLinkCount=0、brokenImageCount=0、missingAltCount=0。CTA 均为真实可点链接（cta 指向 https://evoscientist.ai/，cta-ghost 指向 GitHub 仓库）；VERIFY GATE · GATE OPEN 为状态标签非控件；无假按钮、假图表、假仪表盘或假终端位。networkRequests 为空，无远程模拟 UI。"}},"issues":[]}},"changedFiles":["artifact/zh-CN.html","artifact/en.html"]}],"provenance":{"executor":"opencode","model":"deepseek/deepseek-v4-flash","opencodeVersion":"1.1.58","promptVersion":"dodot-showcase-deepseek-v4","inputHash":"9995e10193f2721c51e5ded64c16384bf21ae6347f6f38838b6800af7caab3b5","artifactHash":"bcf248fcb1233a222d2b635b46d9cc17fc8477864db082ffbcda7a9f28484417","designSkill":"dodot-showcase-design"},"qaEvidence":{"zh-CN":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":2394,"cjkCount":1290,"latinLetterCount":596,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":5124,"h1Top":123.390625,"bottomInset":475.515625,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"fb33fb6b97904448a326bf143d70ecafbae57731e5ef1e4e743004ef90f3be7a","middle":"ef52927335aa991907cf8fcd3cc539b81e23b079a5391f1be016a7683475528f","bottom":"5d627a66f17e126883c240d485865001742afa292e3ca50202ff3b6a0b875ab6"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"zh-CN","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":2477,"cjkCount":1350,"latinLetterCount":618,"scrollWidth":390,"clientWidth":390,"scrollHeight":7710,"h1Top":118.34375,"bottomInset":460.796875,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"c50fe921906492a23fb8abdfa7c6daf4263da6e83d703fb61a7d194405afc298","middle":"abb6d02faa9ecaf82a6b6538f3fd6562ed59caa66c21fe7f7f76e1d79273d64c","bottom":"c2b78b53ba67b1a4c3288fb5dea21b3f41f0024256d568869359bd88797ada55"},"consoleErrors":[],"networkRequests":[]}],"en":[{"viewport":"desktop","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":5478,"cjkCount":0,"latinLetterCount":4365,"scrollWidth":1280,"clientWidth":1280,"scrollHeight":5549,"h1Top":123.390625,"bottomInset":569.203125,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"cb97a8cf92e433c77b877944f5bdd12aa56bac50815e87b7fa4fa332df56e854","middle":"cab476646cd0df7e3b0dfbde469af8663a6e609dda6bb8b7d6fbc5404b2655ed","bottom":"79b840256a67fbf5dbca9b746068b5ec0a57a452bb049141405e0e52296dd253"},"consoleErrors":[],"networkRequests":[]},{"viewport":"mobile","passed":true,"localePassed":true,"metrics":{"language":"en","h1Count":1,"h2Count":6,"sectionCount":6,"textLength":5739,"cjkCount":0,"latinLetterCount":4597,"scrollWidth":390,"clientWidth":390,"scrollHeight":8644,"h1Top":138.734375,"bottomInset":522.78125,"brokenImageCount":0,"missingAltCount":0,"fakeControlCount":0,"emptyLinkCount":0},"screenshotHashes":{"top":"3038fd52788d703c9f2f83a3de6eb9b9ee7435c7b8a8ab7b8392d9808614c74f","middle":"6a954b2a9d6977abee0ec7358caeab911bfaee6517ae224ba5fd8bdd214ee23a","bottom":"4e424f4bcd9c7b0b174bdfa5531abdeea400bc1d9b7c8d79fe3cae3bfe997c48"},"consoleErrors":[],"networkRequests":[]}]}},"generatorVersion":"opencode-1.1.58:deepseek/deepseek-v4-flash:dodot-showcase-deepseek-v4","publishedAt":"2026-08-17T15:18:17.462Z"},"rawExtraction":{"projectRevisionId":"cmt2zgzw400ia6271z3l5r7dh","projectVersion":1,"sources":[{"canonicalKey":"github:evoscientist/evoscientist","role":"primary","revisionVersion":1,"url":"https://github.com/EvoScientist/EvoScientist"}],"facts":[{"key":"repository_owner","value":"evoscientist","evidence":[{"path":"repository.owner","excerpt":"evoscientist"}]},{"key":"repository_name","value":"evoscientist","evidence":[{"path":"repository.name","excerpt":"evoscientist"}]},{"key":"description","value":"🔬 Harness Vibe Research with Self-evolving AI Scientists","evidence":[{"path":"repository.description","excerpt":"🔬 Harness Vibe Research with Self-evolving AI Scientists"}]},{"key":"primary_language","value":"Python","evidence":[{"path":"repository.language","excerpt":"Python"}]},{"key":"license","value":"Apache-2.0","evidence":[{"path":"repository.license","excerpt":"Apache-2.0"}]},{"key":"topics","value":"ai-agent, ai4science, multi-agent-system, vibe-research","evidence":[{"path":"repository.topics","excerpt":"ai-agent, ai4science, multi-agent-system, vibe-research"}]},{"key":"source_count","value":"1","evidence":[{"path":"projectRevision.sources","excerpt":"github:evoscientist/evoscientist"}]}],"taxonomy":{"category":"development_tools","subcategory":"data_ai_tools","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"🔬 Harness Vibe Research with Self-evolving AI Scientists"}]},"lifecycle":{"stage":"active","confidence":0.75,"evidence":[{"path":"repository.pushedAt","excerpt":"2026-08-17T12:10:40.000Z"}]},"userIntents":[{"slug":"build_software","label":"构建或改进软件","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"🔬 Harness Vibe Research with Self-evolving AI Scientists"}]},{"slug":"automate_work","label":"自动化工作流程","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"🔬 Harness Vibe Research with Self-evolving AI Scientists"}]},{"slug":"create_content","label":"创作内容或设计","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"🔬 Harness Vibe Research with Self-evolving AI Scientists"}]},{"slug":"analyze_data","label":"分析与理解数据","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"🔬 Harness Vibe Research with Self-evolving AI Scientists"}]},{"slug":"learn_explore","label":"学习与探索","confidence":0.65,"evidence":[{"path":"repository.description","excerpt":"🔬 Harness Vibe Research with Self-evolving AI Scientists"}]}],"readme":{"headings":["✨ Features","🔥 News","📖 Table of Contents","📦 Installation","Quick Install","Latest from GitHub","Development Install","🐳 Docker","🔑 Configuration","Pick one LLM provider","Web search (optional)","⚡ Quick Start","Per-session: auto-approve via CLI flag","Persistent: set in config (applies to all future sessions)","Or allow only specific command prefixes","Persistent: set in config","Re-enable","⏰ Scheduled Tasks","Add a schedule (cron expression required for /schedule add)","Manage schedules"],"commands":["pip install -e \".[dev]\"","pip install EvoScientist # quick install","pip install -e \".[dev]\" # development install","docker run -it --rm \\","docker compose run --rm evoscientist","docker build -t evoscientist:dev ."]},"changeSet":[{"field":"repository","beforeHash":"e8d9e9923fcac58271dfcc61eba74a59ab4dfdb5a89035688f83c06a7a637765","afterHash":"6bf0590cdc76806d6e21a34379f0ee8c014b473a9b2c75222b1bbdcff6e1a1a1"}]},"publishedAt":"2026-08-16T12:12:26.923Z","updatedAt":"2026-08-21T13:26:49.815Z","summary":{"id":"cmt2zgzw600ic6271goy9x61j","projectType":"agent","oneLine":"自进化 AI 科学家，驾驭 Vibe Research","overview":"EvoScientist 是一个自进化的 AI 科学家系统，采用多智能体团队（规划、研究、编码、调试、分析、写作）和自进化记忆，支持桌面 WebUI、CLI/TUI 和移动端。它采用 human-on-the-loop 范式，让 AI 作为研究伙伴与人类共同进化，适用于科学研究、数据分析和代码生成等场景。","highlights":["多智能体团队","自进化记忆","多界面支持","科学工作流"],"bestFor":["科研人员","数据科学家","AI 开发者"],"caveats":["需要配置 LLM API 密钥","自进化功能需时间积累","部分功能依赖第三方服务"],"categorySuggestion":"development_tools","subcategorySuggestion":"coding_agents","lifecycleStages":["idea_requirements","product_design","backend_data","testing_quality"],"audienceIntents":["discover","build","integrate","evaluate"],"confidence":0.85,"evidence":[{"claim":"EvoScientist 是一个自进化的 AI 科学家系统，采用多智能体团队（规划、研究、编码、调试、分析、写作）和自进化记忆。","sourceType":"readme","locator":"https://github.com/EvoScientist/EvoScientist"},{"claim":"支持多种交互界面：桌面 WebUI、CLI/TUI 和移动端，统一控制。","sourceType":"readme","locator":"https://github.com/EvoScientist/EvoScientist"},{"claim":"提供 AutoSkills 功能，从自身记忆中提炼可复用的技能，并通过 /autoskills 供用户审查。","sourceType":"readme","locator":"https://github.com/EvoScientist/EvoScientist"},{"claim":"支持多模型提供商（Anthropic、OpenAI、Google、MiniMax、NVIDIA）和多渠道（Telegram、Slack、飞书、微信等）。","sourceType":"readme","locator":"https://github.com/EvoScientist/EvoScientist"},{"claim":"采用科学工作流：接收 → 规划 → 执行 → 评估 → 写作 → 验证。","sourceType":"readme","locator":"https://github.com/EvoScientist/EvoScientist"},{"claim":"支持定时任务（cron 风格）和 MCP 服务器集成。","sourceType":"readme","locator":"https://github.com/EvoScientist/EvoScientist"},{"claim":"最新版本 v0.2.7 提升了启动性能，新增多个模型，并加固了 webhook 通道。","sourceType":"release","locator":"https://github.com/EvoScientist/EvoScientist/releases/tag/v0.2.7"}],"publishedAt":"2026-08-17T13:19:32.823Z"},"github":{"url":"https://github.com/EvoScientist/EvoScientist","ownerAvatarUrl":"https://avatars.githubusercontent.com/u/257286936?v=4","starsCount":4470,"forksCount":302,"language":"Python","syncedAt":"2026-08-17T13:17:18.061Z","license":"Apache-2.0","defaultBranch":"main","archived":false,"disabled":false,"lastChangedAt":"2026-08-17T12:10:40.000Z","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}}}]}],"$L25"]