[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7B04Y42XDHFRF4SSJNCWCKA":3,"skill-same:01M7B04Y42XDHFRF4SSJNCWCKA":48},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":29,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M7B04Y42XDHFRF4SSJNCWCKA","testing-dyb-pro-ui",null,"How to bring up and end-to-end test the DYB Pro frontend (three-zone workspace) against the local FastAPI backend on Windows, including running real design cycles without a Devin key.","yacine-baghli\u002FDYB-Pro","yacine-baghli","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F167008935?v=4","01M7B04V5Q4FWSQFAG0PBBYEM8","Yacine Baghli","01M7B04V6ANZ4C1SM8594KW0SG",27,"MIT","https:\u002F\u002Fgithub.com\u002Fyacine-baghli\u002FDYB-Pro\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Ftesting-dyb-pro-ui\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: testing-dyb-pro-ui\ndescription: How to bring up and end-to-end test the DYB Pro frontend (three-zone workspace) against the local FastAPI backend on Windows, including running real design cycles without a Devin key.\n---\n\n# Testing the DYB Pro web UI end-to-end\n\n## Bring up the stack (no credentials needed)\n\nBackend (SQLite fallback, port 8000), from `backend\u002F`:\n\n```powershell\n$env:ALLOW_LOCAL_SIMULATION=\"true\"      # lets a design cycle run without a Devin key (provider badge: local-simulation)\n$env:CELERY_TASK_ALWAYS_EAGER=\"true\"    # run agent tasks inline when Redis is unavailable\n$env:DAEMON_LITERATURE_NETWORK=\"false\"  # keep the lab daemon offline; literature calls otherwise hit OpenAlex\u002FSemantic Scholar and are slow\u002Frate-limited\npython -m uvicorn app.main:app --port 8000\n```\n\nInterpreter note: `python` \u002F `C:\\Python311\\python.exe` may not exist on the box even if a blueprint\nreferences them. The working interpreter has been `C:\\devin\\python\\python.exe` - check with\n`Get-Command python` \u002F `Test-Path` before assuming, and use PowerShell syntax (`;`, not `&&`).\n\n`npm run start` prints a warning that it \"does not work with output: standalone\"; it nonetheless serves\ncorrectly on :3000. After rebuilding you MUST hard-reload the browser (Ctrl+Shift+R) or you will keep\ntesting the previously served bundle - an easy way to report a stale UI as the new one.\n\nFrontend (port 3000), from `frontend\u002F`: `npm run dev` (or `npm run build; npm run start`).\nDefault API key `dyb-pro-demo-scientist` is sent automatically by the frontend; the other seeded keys\nare `dyb-pro-demo-admin` and `dyb-pro-demo-viewer` (viewer is read-only, RBAC is enforced).\n\nRedis is often unavailable in the sandbox - `CELERY_TASK_ALWAYS_EAGER=true` is the workaround.\nCaveat: with eager mode the `POST \u002Fcycles` request runs the whole cycle synchronously, so transient\nUI states (`cycle running...`, `queued`\u002F`running` agent pills) may flash by or never render; the run\nbutton just stays disabled until the cycle is already `committed`. To observe live agent-status\ntransitions, run a real Celery worker + Redis (`docker compose up`) instead of eager mode.\n\n## Seeding data\n\nClick `load demo project` in the left ASK pane - it seeds and selects \"GB1 stability + Fc binding (demo)\"\nwith a full commit history. Reset local state by deleting `backend\\dybpro.db` and `backend\\artifacts`.\nA cycle in local-simulation mode adds ~15 commits, so repeated cycles inflate the fold strip quickly\n(relevant to the layout bug below).\n\n## Useful selectors \u002F handles\n\n- Left pane: `aside[aria-label=\"ask and agent control\"]`, `textarea[aria-label=\"research brief\"]`,\n  `select[aria-label=\"project\"]`. Ctrl+Enter in the brief starts a cycle.\n- Center tabs are `button[aria-selected]` with text `protein viewer` \u002F `version DAG` \u002F `wet-lab shortlist` \u002F\n  `observation log`; fold strip thumbnails are `button[aria-label^=\"version \"]` inside `.thumbs`.\n- Right pane: `aside[aria-label=\"pixel agent swarm\"]`. It stays empty (\"Run a cycle to watch...\") until a\n  cycle has run; with `CELERY_TASK_ALWAYS_EAGER=true` the cycle completes synchronously (~30s) so every\n  agent lands on `done` and the intermediate `thinking`\u002F`searching`\u002F`writing` state colours never render.\n  Run a real Celery worker + Redis if those states must be verified.\n\n### Research lab tab (`\u002Fapi\u002Flab` subsystem)\n\n- Center tab labelled `research lab`; component tree `ResearchWorkspace` -> `DaemonPane`, `ResearchFeed`,\n  `LabelStudio`, `WetlabLoop`, `LearnedPane` via `frontend\u002Flib\u002Fresearch.ts`.\n- Handles: `[data-testid=campaign-summary]`, `daemon-refresh`, `daemon-tick`, `label-create`,\n  `wetlab-plan`, `wetlab-simulate`, `proposal-refresh`, `proposal-blocked`, `seed-campaign`.\n- `seed-campaign` takes ~1-2 min and is **disabled once `digest.result_count > 0`** (coded idempotency\n  guard, not a bug). Seed over the API first, as `frontend\u002Fe2e\u002Fdemo.spec.ts` does, then test the UI state:\n  `POST \u002Fapi\u002Flab\u002Fprojects\u002F{id}\u002Fresearch\u002Fseed-campaign` with header `X-API-Key: dyb-pro-demo-scientist`.\n- Cross-check the UI against the API rather than trusting the rendering: `\u002Fapi\u002Flab\u002Fprojects\u002F{id}\u002Fresearch`\n  (`digest.version_count|paper_count|result_count`, `learned.drift[]`, `learned.headline`),\n  `\u002Fapi\u002Flab\u002Fcommits\u002F{fullCommitId}\u002Fwetlab\u002Fplan|results`, `\u002Fapi\u002Flab\u002Fprojects\u002F{id}\u002Fresearch\u002Fproposal`.\n  Commit ids shown in the UI are TRUNCATED to 12 chars; the lab routes need the full 64-char id, which\n  you can get from `GET \u002Fapi\u002Fprojects\u002F{id}\u002Fcommits` (a short id returns `{\"detail\":\"version not found\"}`).\n- Running a design cycle also feeds the lab daemon (new commits queue `results_registered`\u002F`version_created`\n  tasks), so `queue`\u002F`events`\u002F`knowledge v` badges move - handy for proving the daemon reacts.\n- Command palette: Ctrl+K (input placeholder `run a command...`), type to filter, Enter to run, Escape closes.\n- Drag-to-compare needs a real mouse down, several intermediate moves, then up on the target thumbnail;\n  a single `left_click_drag` may not trigger the HTML5 drag handlers.\n\n## Things to check that have been broken before\n\n- **Horizontal overflow driven by the fold strip.** `.thumbs` is `display:flex; overflow-x:auto`, but if any\n  ancestor (`.fold-strip`, `.viewer`, `.center-body`, `main`) loses `min-width: 0` the row stretches the whole\n  page instead of scrolling: with 31 commits the document `scrollWidth` was 3884px at a 1280px viewport.\n  Verify with `document.documentElement.scrollWidth` vs `window.innerWidth` after loading a project with many\n  commits, at both ~1280px and ~1100px widths. Symptom: controls such as `exit compare` land far offscreen.\n- **Export links.** Shortlist `export CSV\u002FJSON\u002FFASTA` anchors point straight at the API URL, so the browser\n  navigation carries no `X-API-Key` header and the response is `{\"detail\":\"missing X-API-Key header\"}`.\n  Always click the export control in the UI rather than assuming the link works. (Known open bug.)\n- Responsive breakpoint: below 1180px the workspace should collapse to a single column\n  (`grid-template-columns` becomes one track).\n- **`MiniBars` charts collapsing in narrow columns.** `.barrow` is\n  `grid-template-columns: minmax(96px,148px) minmax(0,1fr) auto` with a `white-space: nowrap` mono readout,\n  so inside a narrow column (e.g. the 2-col `.center-body.lab` grid at a 1024px viewport) the fixed name\n  column and the readout eat the row and `.bartrack` collapses to ~10-30px, with bar fills under 3px -\n  the chart becomes unreadable even though the numbers are right. Worse, each `.barrow` is its own grid,\n  so track width varies per row and bar lengths are NOT comparable across rows. Always measure rather\n  than eyeball:\n  `[...document.querySelectorAll('.bartrack')].map(t=>t.getBoundingClientRect().width)`\n  and compare `.barfill` `style.width` percentages against the API values. Possible fixes to suggest:\n  a `min-width` on `.bartrack`, letting the readout wrap, shrinking the name column, or hoisting the\n  columns onto the shared `.barchart` grid so all rows align.\n- **`.center-body.lab` horizontal overflow.** The 2-column lab grid overflowed by ~61px\n  (`scrollWidth 762` vs `clientWidth 701`) at a 1024px viewport, producing a horizontal scrollbar.\n  Check `document.querySelector('.center-body.lab')` scrollWidth vs clientWidth.\n- **`.tip` tooltips carry prose that was deleted from the surface.** `globals.css` hides\n  `.tip::after` with `opacity:0; visibility:hidden` and reveals it on `:hover` and `:focus-visible`\n  (spans carry `tabIndex={0}`). If the tooltip fails, that explanatory text is simply gone, so verify\n  BOTH paths with screenshots: hover and screenshot while hovering; then click the element, move the\n  mouse far away, press `shift+Tab` then `Tab` (a plain click does not match `:focus-visible`) and\n  screenshot the focused state. Note `scrollWidth` of a `.tip` element is inflated by the tooltip\n  pseudo-element, so it produces false \"clipped text\" positives.\n\n## Window resizing on Windows for responsive checks\n\n`xdotool` is not available; resize Chrome via Win32 from PowerShell\n(`user32.dll` `MoveWindow` on `(Get-Process chrome | ? MainWindowHandle -ne 0).MainWindowHandle`),\nthen `ShowWindow(handle, 3)` to re-maximize.\n\n## Devin Secrets Needed\n\nNone - local simulation and the demo API key cover the whole golden path.\n","# Testing the DYB Pro web UI end-to-end\n\n## Bring up the stack (no credentials needed)\n\nBackend (SQLite fallback, port 8000), from `backend\u002F`:\n\n```powershell\n$env:ALLOW_LOCAL_SIMULATION=\"true\"      # lets a design cycle run without a Devin key (provider badge: local-simulation)\n$env:CELERY_TASK_ALWAYS_EAGER=\"true\"    # run agent tasks inline when Redis is unavailable\n$env:DAEMON_LITERATURE_NETWORK=\"false\"  # keep the lab daemon offline; literature calls otherwise hit OpenAlex\u002FSemantic Scholar and are slow\u002Frate-limited\npython -m uvicorn app.main:app --port 8000\n```\n\nInterpreter note: `python` \u002F `C:\\Python311\\python.exe` may not exist on the box even if a blueprint\nreferences them. The working interpreter has been `C:\\devin\\python\\python.exe` - check with\n`Get-Command python` \u002F `Test-Path` before assuming, and use PowerShell syntax (`;`, not `&&`).\n\n`npm run start` prints a warning that it \"does not work with output: standalone\"; it nonetheless serves\ncorrectly on :3000. After rebuilding you MUST hard-reload the browser (Ctrl+Shift+R) or you will keep\ntesting the previously served bundle - an easy way to report a stale UI as the new one.\n\nFrontend (port 3000), from `frontend\u002F`: `npm run dev` (or `npm run build; npm run start`).\nDefault API key `dyb-pro-demo-scientist` is sent automatically by the frontend; the other seeded keys\nare `dyb-pro-demo-admin` and `dyb-pro-demo-viewer` (viewer is read-only, RBAC is enforced).\n\nRedis is often unavailable in the sandbox - `CELERY_TASK_ALWAYS_EAGER=true` is the workaround.\nCaveat: with eager mode the `POST \u002Fcycles` request runs the whole cycle synchronously, so transient\nUI states (`cycle running...`, `queued`\u002F`running` agent pills) may flash by or never render; the run\nbutton just stays disabled until the cycle is already `committed`. To observe live agent-status\ntransitions, run a real Celery worker + Redis (`docker compose up`) instead of eager mode.\n\n## Seeding data\n\nClick `load demo project` in the left ASK pane - it seeds and selects \"GB1 stability + Fc binding (demo)\"\nwith a full commit history. Reset local state by deleting `backend\\dybpro.db` and `backend\\artifacts`.\nA cycle in local-simulation mode adds ~15 commits, so repeated cycles inflate the fold strip quickly\n(relevant to the layout bug below).\n\n## Useful selectors \u002F handles\n\n- Left pane: `aside[aria-label=\"ask and agent control\"]`, `textarea[aria-label=\"research brief\"]`,\n  `select[aria-label=\"project\"]`. Ctrl+Enter in the brief starts a cycle.\n- Center tabs are `button[aria-selected]` with text `protein viewer` \u002F `version DAG` \u002F `wet-lab shortlist` \u002F\n  `observation log`; fold strip thumbnails are `button[aria-label^=\"version \"]` inside `.thumbs`.\n- Right pane: `aside[aria-label=\"pixel agent swarm\"]`. It stays empty (\"Run a cycle to watch...\") until a\n  cycle has run; with `CELERY_TASK_ALWAYS_EAGER=true` the cycle completes synchronously (~30s) so every\n  agent lands on `done` and the intermediate `thinking`\u002F`searching`\u002F`writing` state colours never render.\n  Run a real Celery worker + Redis if those states must be verified.\n\n### Research lab tab (`\u002Fapi\u002Flab` subsystem)\n\n- Center tab labelled `research lab`; component tree `ResearchWorkspace` -> `DaemonPane`, `ResearchFeed`,\n  `LabelStudio`, `WetlabLoop`, `LearnedPane` via `frontend\u002Flib\u002Fresearch.ts`.\n- Handles: `[data-testid=campaign-summary]`, `daemon-refresh`, `daemon-tick`, `label-create`,\n  `wetlab-plan`, `wetlab-simulate`, `proposal-refresh`, `proposal-blocked`, `seed-campaign`.\n- `seed-campaign` takes ~1-2 min and is **disabled once `digest.result_count > 0`** (coded idempotency\n  guard, not a bug). Seed over the API first, as `frontend\u002Fe2e\u002Fdemo.spec.ts` does, then test the UI state:\n  `POST \u002Fapi\u002Flab\u002Fprojects\u002F{id}\u002Fresearch\u002Fseed-campaign` with header `X-API-Key: dyb-pro-demo-scientist`.\n- Cross-check the UI against the API rather than trusting the rendering: `\u002Fapi\u002Flab\u002Fprojects\u002F{id}\u002Fresearch`\n  (`digest.version_count|paper_count|result_count`, `learned.drift[]`, `learned.headline`),\n  `\u002Fapi\u002Flab\u002Fcommits\u002F{fullCommitId}\u002Fwetlab\u002Fplan|results`, `\u002Fapi\u002Flab\u002Fprojects\u002F{id}\u002Fresearch\u002Fproposal`.\n  Commit ids shown in the UI are TRUNCATED to 12 chars; the lab routes need the full 64-char id, which\n  you can get from `GET \u002Fapi\u002Fprojects\u002F{id}\u002Fcommits` (a short id returns `{\"detail\":\"version not found\"}`).\n- Running a design cycle also feeds the lab daemon (new commits queue `results_registered`\u002F`version_created`\n  tasks), so `queue`\u002F`events`\u002F`knowledge v` badges move - handy for proving the daemon reacts.\n- Command palette: Ctrl+K (input placeholder `run a command...`), type to filter, Enter to run, Escape closes.\n- Drag-to-compare needs a real mouse down, several intermediate moves, then up on the target thumbnail;\n  a single `left_click_drag` may not trigger the HTML5 drag handlers.\n\n## Things to check that have been broken before\n\n- **Horizontal overflow driven by the fold strip.** `.thumbs` is `display:flex; overflow-x:auto`, but if any\n  ancestor (`.fold-strip`, `.viewer`, `.center-body`, `main`) loses `min-width: 0` the row stretches the whole\n  page instead of scrolling: with 31 commits the document `scrollWidth` was 3884px at a 1280px viewport.\n  Verify with `document.documentElement.scrollWidth` vs `window.innerWidth` after loading a project with many\n  commits, at both ~1280px and ~1100px widths. Symptom: controls such as `exit compare` land far offscreen.\n- **Export links.** Shortlist `export CSV\u002FJSON\u002FFASTA` anchors point straight at the API URL, so the browser\n  navigation carries no `X-API-Key` header and the response is `{\"detail\":\"missing X-API-Key header\"}`.\n  Always click the export control in the UI rather than assuming the link works. (Known open bug.)\n- Responsive breakpoint: below 1180px the workspace should collapse to a single column\n  (`grid-template-columns` becomes one track).\n- **`MiniBars` charts collapsing in narrow columns.** `.barrow` is\n  `grid-template-columns: minmax(96px,148px) minmax(0,1fr) auto` with a `white-space: nowrap` mono readout,\n  so inside a narrow column (e.g. the 2-col `.center-body.lab` grid at a 1024px viewport) the fixed name\n  column and the readout eat the row and `.bartrack` collapses to ~10-30px, with bar fills under 3px -\n  the chart becomes unreadable even though the numbers are right. Worse, each `.barrow` is its own grid,\n  so track width varies per row and bar lengths are NOT comparable across rows. Always measure rather\n  than eyeball:\n  `[...document.querySelectorAll('.bartrack')].map(t=>t.getBoundingClientRect().width)`\n  and compare `.barfill` `style.width` percentages against the API values. Possible fixes to suggest:\n  a `min-width` on `.bartrack`, letting the readout wrap, shrinking the name column, or hoisting the\n  columns onto the shared `.barchart` grid so all rows align.\n- **`.center-body.lab` horizontal overflow.** The 2-column lab grid overflowed by ~61px\n  (`scrollWidth 762` vs `clientWidth 701`) at a 1024px viewport, producing a horizontal scrollbar.\n  Check `document.querySelector('.center-body.lab')` scrollWidth vs clientWidth.\n- **`.tip` tooltips carry prose that was deleted from the surface.** `globals.css` hides\n  `.tip::after` with `opacity:0; visibility:hidden` and reveals it on `:hover` and `:focus-visible`\n  (spans carry `tabIndex={0}`). If the tooltip fails, that explanatory text is simply gone, so verify\n  BOTH paths with screenshots: hover and screenshot while hovering; then click the element, move the\n  mouse far away, press `shift+Tab` then `Tab` (a plain click does not match `:focus-visible`) and\n  screenshot the focused state. Note `scrollWidth` of a `.tip` element is inflated by the tooltip\n  pseudo-element, so it produces false \"clipped text\" positives.\n\n## Window resizing on Windows for responsive checks\n\n`xdotool` is not available; resize Chrome via Win32 from PowerShell\n(`user32.dll` `MoveWindow` on `(Get-Process chrome | ? MainWindowHandle -ne 0).MainWindowHandle`),\nthen `ShowWindow(handle, 3)` to re-maximize.\n\n## Devin Secrets Needed\n\nNone - local simulation and the demo API key cover the whole golden path.\n",{"name":5,"description":7},"main","f5bab5d3269da1f12320f18df5dc850ab02bf6ef",".agents\u002Fskills\u002Ftesting-dyb-pro-ui",1,"https:\u002F\u002Fgithub.com\u002Fyacine-baghli\u002FDYB-Pro","https:\u002F\u002Fdyb-pro.vercel.app\u002F","https:\u002F\u002Fgithub.com\u002Fyacine-baghli","https:\u002F\u002Fyacine-baghli.github.io","MSc Researche @ CentraleSupélec","2026-08-23 17:52:13",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add yacine-baghli\u002FDYB-Pro --skill testing-dyb-pro-ui",{"tool":40,"command":41},"bunx","bunx skills add yacine-baghli\u002FDYB-Pro --skill testing-dyb-pro-ui",{"tool":43,"command":44},"pnpm","pnpm dlx skills add yacine-baghli\u002FDYB-Pro --skill testing-dyb-pro-ui","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fyacine-baghli\u002FDYB-Pro\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Ftesting-dyb-pro-ui\u002FSKILL.md",true,[],[49,55],{"skillNo":50,"name":51,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M7B053790JK2BE9DJ2D80MQ4","testing-foldsmith","https:\u002F\u002Fgithub.com\u002Fyacine-baghli\u002FDYB-Pro\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Ftesting-foldsmith\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":58,"categories":59,"keywords":60,"publishedAt":19},"01M7B0512MYN0P9R83RTAX5PYR","testing-foldsmith-demo","https:\u002F\u002Fgithub.com\u002Fyacine-baghli\u002FDYB-Pro\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Ftesting-foldsmith-demo\u002FSKILL.md",[],[]]