[{"data":1,"prerenderedAt":696},["ShallowReactive",2],{"infra-\u002Finfra\u002Fwebsites":3,"infra-children-\u002Finfra\u002Fwebsites":25},{"id":4,"title":5,"body":6,"contentId":13,"description":14,"ecosystem":15,"extension":16,"meta":17,"navigation":18,"part":15,"path":19,"rawbody":20,"repository":15,"robots":15,"schemaOrg":15,"seo":21,"sitemap":22,"stem":23,"url":15,"__hash__":24},"pages\u002Finfra\u002Fwebsites\u002Findex.md","Websites",{"type":7,"value":8,"toc":9},"minimark",[],{"title":10,"searchDepth":11,"depth":11,"links":12},"",2,[],"0497b82d-fdf9-4427-beca-c1002186cbfc","The public websites I build and maintain.",null,"md",{},true,"\u002Finfra\u002Fwebsites","---\ncontentId: 0497b82d-fdf9-4427-beca-c1002186cbfc\ntitle: Websites\ndescription: The public websites I build and maintain.\n---\n",{"title":5,"description":14},{"loc":19},"infra\u002Fwebsites\u002Findex","Z9CRAN2G7QVNdnSR9xvvGXEDKzO478XOH8Mrk2RNScU",[26,206,452],{"id":27,"title":28,"body":29,"contentId":140,"description":141,"ecosystem":142,"extension":16,"meta":199,"navigation":18,"part":15,"path":200,"rawbody":201,"repository":116,"robots":15,"schemaOrg":15,"seo":202,"sitemap":203,"stem":204,"url":37,"__hash__":205},"pages\u002Finfra\u002Fwebsites\u002Fcode-soubiran-dev.md","code.soubiran.dev",{"type":7,"value":30,"toc":133},[31,41,49,54,57,60,63,67,70,73,80,84,95,101,104,107,111,123,126,130],[32,33,34,40],"p",{},[35,36,28],"a",{"href":37,"rel":38},"https:\u002F\u002Fcode.soubiran.dev",[39],"nofollow"," turns source code into syntax-highlighted PNG images. It combines a browser editor for people with a public MCP endpoint for assistants, and both paths use the same page to render the final image.",[32,42,43,44,48],{},"The browser exposes URL-backed controls for the code, language, canvas size, gradient, title, and watermark before downloading the result. An MCP client can send those fields to ",[45,46,47],"code",{},"https:\u002F\u002Fcode.soubiran.dev\u002Fmcp"," and receive the PNG as image content.",[50,51,53],"h2",{"id":52},"why-the-url-holds-the-editor-state","Why the URL holds the editor state",[32,55,56],{},"I wanted shared links and automated captures to reproduce the same image without storing documents on a server. The app therefore keeps its editor state in URL search parameters. It encodes the source code as Base64 over UTF-8 bytes and stores the other options as plain parameter values.",[32,58,59],{},"This URL is the contract between the two ways to use the product. Opening it reconstructs the editor in a browser. The MCP Worker can build the same URL from tool arguments and pass it to a remote browser without maintaining a second rendering implementation.",[32,61,62],{},"The approach fits a small rendering tool whose state can travel with the link and does not need persistence. Base64 turns the UTF-8 source bytes into text that the query parameter can carry. It does not encrypt them, so these URLs are not intended for secrets.",[50,64,66],{"id":65},"how-the-browser-creates-an-image","How the browser creates an image",[32,68,69],{},"The Vue app reads the URL when it starts. Each URL-backed control writes its new value back to the query string, while empty and default values stay out of the URL where possible. Copying the current address is enough to preserve the image configuration.",[32,71,72],{},"The editor renders two aligned layers. Shiki produces highlighted HTML underneath a transparent textarea, so users edit normal text while seeing highlighted code. A wrapper around the editor, title, background, and watermark marks the exact DOM element used for capture.",[32,74,75,76,79],{},"When the user selects Capture, ",[45,77,78],{},"modern-screenshot"," converts that element into a PNG data URL and the browser downloads it. This capture path does not call the Worker or an image-generation API.",[50,81,83],{"id":82},"how-the-mcp-route-reuses-the-page","How the MCP route reuses the page",[32,85,86,87,90,91,94],{},"The Cloudflare Worker handles ",[45,88,89],{},"\u002Fmcp"," and exposes one tool, ",[45,92,93],{},"generate_code_image",". Its input uses the URL-backed image fields: code, language, size, gradient, title, and watermark.",[32,96,97,98,100],{},"For each call, the Worker validates the arguments and creates a ",[45,99,28],{}," URL. It omits options that match the defaults and uses the same Base64 encoder as the browser for source code. The Worker then asks Cloudflare Browser Run to open that public URL, wait for the marked editor element, and capture it as a PNG.",[32,102,103],{},"Browser Run returns the image bytes to the Worker. The MCP response carries those bytes as Base64-encoded image content with its MIME type. Reusing the public page keeps browser downloads and MCP results on the same Vue, Shiki, and CSS rendering path.",[32,105,106],{},"The MCP endpoint is public, but its Browser Run credentials remain in Worker environment bindings. Tool logs record option choices, source length, outcome, duration, and image metadata without recording the source code, title, or watermark.",[50,108,110],{"id":109},"how-it-runs","How it runs",[32,112,113,114,119,120,122],{},"Vite builds the Vue app and the Worker from the ",[35,115,118],{"href":116,"rel":117},"https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Fcode.soubiran.dev",[39],"same repository",". Cloudflare Workers Static Assets serves the app, while the asset configuration sends ",[45,121,89],{}," to the Worker before applying the single-page application fallback. The editor and its automation endpoint therefore share one deployment and one domain.",[32,124,125],{},"The deployment command builds the project before Wrangler publishes it. GitHub Actions checks pushes and pull requests with linting, a production build, type checking, and tests. Cloudflare logs and traces cover Worker requests, including MCP tool outcomes.",[50,127,129],{"id":128},"ecosystem","Ecosystem",[131,132],"infra-ecosystem",{},{"title":10,"searchDepth":11,"depth":11,"links":134},[135,136,137,138,139],{"id":52,"depth":11,"text":53},{"id":65,"depth":11,"text":66},{"id":82,"depth":11,"text":83},{"id":109,"depth":11,"text":110},{"id":128,"depth":11,"text":129},"2e9e7543-2215-49a6-b37d-a531f779b2da","The browser editor and public MCP endpoint that turn URL-based editor state into syntax-highlighted PNG images.",[143,195],{"type":144,"id":145,"name":146,"description":147,"ecosystem":148},"deployment","code-soubiran-dev","Cloudflare Workers","Serve the browser app and handle the public `\u002Fmcp` endpoint.",[149,191],{"type":150,"id":28,"name":151,"description":152,"href":116,"ecosystem":153},"repository","GitHub","Host the editor, Worker, shared image contract, and deployment configuration.",[154,159,163,167,171,174,178,182,186],{"type":155,"name":156,"description":157,"href":158},"stack","Vite","Build the browser app and Worker from one project.","https:\u002F\u002Fvite.dev",{"type":155,"name":160,"description":161,"href":162},"Vue","Bind the editor controls to URL search parameters.","https:\u002F\u002Fvuejs.org",{"type":155,"name":164,"description":165,"href":166},"Nuxt UI","Provide the editor controls and visual components.","https:\u002F\u002Fui.nuxt.com",{"type":155,"name":168,"description":169,"href":170},"Shiki","Render the syntax-highlighted code layer in the browser.","https:\u002F\u002Fshiki.style",{"type":155,"name":78,"description":172,"href":173},"Capture the rendered editor element as a PNG in the browser.","https:\u002F\u002Fgithub.com\u002Fqq15725\u002Fmodern-screenshot",{"type":155,"name":175,"description":176,"href":177},"Model Context Protocol SDK","Define the image-generation tool and its image response.","https:\u002F\u002Fmodelcontextprotocol.io",{"type":155,"name":179,"description":180,"href":181},"Cloudflare Agents","Expose the MCP server through the Worker's `\u002Fmcp` route.","https:\u002F\u002Fdevelopers.cloudflare.com\u002Fagents\u002F",{"type":155,"name":183,"description":184,"href":185},"Evlog","Record structured MCP tool telemetry without logging image content.","https:\u002F\u002Fevlog.dev",{"type":187,"relationship":188,"name":189,"description":190},"ci\u002Fcd","consumer","GitHub Actions","Lint, build, type-check, and test changes pushed to the repository.",{"type":155,"name":192,"description":193,"href":194},"Cloudflare Browser Run","Load a URL that describes an image and capture its rendered editor element.","https:\u002F\u002Fdevelopers.cloudflare.com\u002Fbrowser-rendering\u002F",{"type":196,"name":197,"description":198},"domain","Cloudflare Domains","Route `code.soubiran.dev` to the Worker deployment.",{},"\u002Finfra\u002Fwebsites\u002Fcode-soubiran-dev","---\ncontentId: 2e9e7543-2215-49a6-b37d-a531f779b2da\ntitle: code.soubiran.dev\ndescription: >-\n  The browser editor and public MCP endpoint that turn URL-based editor state\n  into syntax-highlighted PNG images.\nurl: 'https:\u002F\u002Fcode.soubiran.dev'\nrepository: 'https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Fcode.soubiran.dev'\necosystem:\n  - type: deployment\n    id: code-soubiran-dev\n    name: Cloudflare Workers\n    description: Serve the browser app and handle the public `\u002Fmcp` endpoint.\n    ecosystem:\n      - type: repository\n        id: code.soubiran.dev\n        name: GitHub\n        description: Host the editor, Worker, shared image contract, and deployment configuration.\n        href: 'https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Fcode.soubiran.dev'\n        ecosystem:\n          - type: stack\n            name: Vite\n            description: Build the browser app and Worker from one project.\n            href: 'https:\u002F\u002Fvite.dev'\n          - type: stack\n            name: Vue\n            description: Bind the editor controls to URL search parameters.\n            href: 'https:\u002F\u002Fvuejs.org'\n          - type: stack\n            name: Nuxt UI\n            description: Provide the editor controls and visual components.\n            href: 'https:\u002F\u002Fui.nuxt.com'\n          - type: stack\n            name: Shiki\n            description: Render the syntax-highlighted code layer in the browser.\n            href: 'https:\u002F\u002Fshiki.style'\n          - type: stack\n            name: modern-screenshot\n            description: Capture the rendered editor element as a PNG in the browser.\n            href: 'https:\u002F\u002Fgithub.com\u002Fqq15725\u002Fmodern-screenshot'\n          - type: stack\n            name: Model Context Protocol SDK\n            description: Define the image-generation tool and its image response.\n            href: 'https:\u002F\u002Fmodelcontextprotocol.io'\n          - type: stack\n            name: Cloudflare Agents\n            description: Expose the MCP server through the Worker's `\u002Fmcp` route.\n            href: 'https:\u002F\u002Fdevelopers.cloudflare.com\u002Fagents\u002F'\n          - type: stack\n            name: Evlog\n            description: Record structured MCP tool telemetry without logging image content.\n            href: 'https:\u002F\u002Fevlog.dev'\n          - type: ci\u002Fcd\n            relationship: consumer\n            name: GitHub Actions\n            description: Lint, build, type-check, and test changes pushed to the repository.\n      - type: stack\n        name: Cloudflare Browser Run\n        description: Load a URL that describes an image and capture its rendered editor element.\n        href: 'https:\u002F\u002Fdevelopers.cloudflare.com\u002Fbrowser-rendering\u002F'\n  - type: domain\n    name: Cloudflare Domains\n    description: Route `code.soubiran.dev` to the Worker deployment.\n---\n\n[code.soubiran.dev](https:\u002F\u002Fcode.soubiran.dev) turns source code into syntax-highlighted PNG images. It combines a browser editor for people with a public MCP endpoint for assistants, and both paths use the same page to render the final image.\n\nThe browser exposes URL-backed controls for the code, language, canvas size, gradient, title, and watermark before downloading the result. An MCP client can send those fields to `https:\u002F\u002Fcode.soubiran.dev\u002Fmcp` and receive the PNG as image content.\n\n## Why the URL holds the editor state\n\nI wanted shared links and automated captures to reproduce the same image without storing documents on a server. The app therefore keeps its editor state in URL search parameters. It encodes the source code as Base64 over UTF-8 bytes and stores the other options as plain parameter values.\n\nThis URL is the contract between the two ways to use the product. Opening it reconstructs the editor in a browser. The MCP Worker can build the same URL from tool arguments and pass it to a remote browser without maintaining a second rendering implementation.\n\nThe approach fits a small rendering tool whose state can travel with the link and does not need persistence. Base64 turns the UTF-8 source bytes into text that the query parameter can carry. It does not encrypt them, so these URLs are not intended for secrets.\n\n## How the browser creates an image\n\nThe Vue app reads the URL when it starts. Each URL-backed control writes its new value back to the query string, while empty and default values stay out of the URL where possible. Copying the current address is enough to preserve the image configuration.\n\nThe editor renders two aligned layers. Shiki produces highlighted HTML underneath a transparent textarea, so users edit normal text while seeing highlighted code. A wrapper around the editor, title, background, and watermark marks the exact DOM element used for capture.\n\nWhen the user selects Capture, `modern-screenshot` converts that element into a PNG data URL and the browser downloads it. This capture path does not call the Worker or an image-generation API.\n\n## How the MCP route reuses the page\n\nThe Cloudflare Worker handles `\u002Fmcp` and exposes one tool, `generate_code_image`. Its input uses the URL-backed image fields: code, language, size, gradient, title, and watermark.\n\nFor each call, the Worker validates the arguments and creates a `code.soubiran.dev` URL. It omits options that match the defaults and uses the same Base64 encoder as the browser for source code. The Worker then asks Cloudflare Browser Run to open that public URL, wait for the marked editor element, and capture it as a PNG.\n\nBrowser Run returns the image bytes to the Worker. The MCP response carries those bytes as Base64-encoded image content with its MIME type. Reusing the public page keeps browser downloads and MCP results on the same Vue, Shiki, and CSS rendering path.\n\nThe MCP endpoint is public, but its Browser Run credentials remain in Worker environment bindings. Tool logs record option choices, source length, outcome, duration, and image metadata without recording the source code, title, or watermark.\n\n## How it runs\n\nVite builds the Vue app and the Worker from the [same repository](https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Fcode.soubiran.dev). Cloudflare Workers Static Assets serves the app, while the asset configuration sends `\u002Fmcp` to the Worker before applying the single-page application fallback. The editor and its automation endpoint therefore share one deployment and one domain.\n\nThe deployment command builds the project before Wrangler publishes it. GitHub Actions checks pushes and pull requests with linting, a production build, type checking, and tests. Cloudflare logs and traces cover Worker requests, including MCP tool outcomes.\n\n## Ecosystem\n\n::infra-ecosystem\n::\n",{"title":28,"description":141},{"loc":200},"infra\u002Fwebsites\u002Fcode-soubiran-dev","k7DAtWtOXlBf89HINRnZzBBEM062q8JSV1rdXVwG7yw",{"id":207,"title":208,"body":209,"contentId":365,"description":366,"ecosystem":367,"extension":16,"meta":445,"navigation":18,"part":15,"path":446,"rawbody":447,"repository":220,"robots":15,"schemaOrg":15,"seo":448,"sitemap":449,"stem":450,"url":215,"__hash__":451},"pages\u002Finfra\u002Fwebsites\u002Fsoubiran-dev.md","soubiran.dev",{"type":7,"value":210,"toc":358},[211,224,228,238,245,253,256,260,275,285,296,303,307,315,321,324,331,335,348,354,356],[32,212,213,217,218,223],{},[35,214,208],{"href":215,"rel":216},"https:\u002F\u002Fsoubiran.dev",[39]," is my personal website. It brings posts, series, projects, talks, and infrastructure documentation into one Nuxt 4 application. The ",[35,219,222],{"href":220,"rel":221},"https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Fbeta.soubiran.dev",[39],"source repository"," contains the content and the local Nuxt modules that implement the site's features.",[50,225,227],{"id":226},"static-pages-build-time-content","Static pages, build-time content",[32,229,230,233,234,237],{},[45,231,232],{},"nuxt generate"," produces the website in ",[45,235,236],{},".output\u002Fpublic",". Wrangler deploys that directory to Cloudflare Workers Static Assets, which serves the generated pages without running Nuxt for each visit.",[32,239,240,241,244],{},"Nuxt Content reads English and French Markdown into collections for pages, posts, and series. The infrastructure documentation lives in ",[45,242,243],{},"content\u002Finfra\u002F",", stays English-only, and uses the same page collection and shared layout. Its ecosystem and dataflow diagrams use Vue Flow.",[32,246,247,248,252],{},"The build also imports remote data. The projects section fetches GitHub repository metadata. The talks module reads both localized catalogs from ",[35,249,251],{"href":250},"\u002Finfra\u002Fwebsites\u002Ftalks-soubiran-dev","talks.soubiran.dev",", fetches each talk's metadata, and imports summaries and available transcripts into separate Nuxt Content collections. Talk pages combine those documents with recordings, slides, resource links, and event locations. The talk index includes a map and aggregate statistics.",[32,254,255],{},"Nuxt UI supplies interface components, with Vue and Tailwind CSS for the application code and styles. Local modules organize content, authentication, comments, reactions, feedback, analytics, and the assistant.",[50,257,259],{"id":258},"a-catalog-for-readers-and-tools","A catalog for readers and tools",[32,261,262,263,266,267,270,271,274],{},"The website publishes ",[45,264,265],{},"\u002Fpages.en.json"," and ",[45,268,269],{},"\u002Fpages.fr.json",". Each catalog is an array of records with a stable ID, locale, content type, title, description, published URL, and explicit ",[45,272,273],{},"markdownUrl",". Dated content includes a publication date. Translation links connect records available in another locale.",[32,276,277,278,281,282,284],{},"The catalogs include ordinary pages, infrastructure documentation, posts, series, series articles, and talks with summaries. Infrastructure documents use the ",[45,279,280],{},"page"," type. Consumers follow ",[45,283,273],{}," to retrieve a document rather than infer its location from the HTML URL.",[32,286,287,288,291,292,295],{},"The ",[45,289,290],{},"\u002Fraw\u002F"," endpoints export readable Markdown. A talk document includes its event metadata, resource links, summary, and transcript when available. Generated documents for index pages include the corresponding content lists. The site also publishes RSS feeds and LLM-oriented content through ",[45,293,294],{},"nuxt-llms",".",[32,297,298,302],{},[35,299,301],{"href":300},"\u002Finfra\u002Fservices\u002Fmcp-soubiran-dev","mcp.soubiran.dev"," reads both page catalogs. The website therefore publishes the discovery metadata and Markdown used by the MCP tools, including the talks imported from the archive.",[50,304,306],{"id":305},"interactions-belong-to-the-api","Interactions belong to the API",[32,308,309,310,314],{},"The generated pages do not own mutable user data. ",[35,311,313],{"href":312},"\u002Finfra\u002Fservices\u002Fapi-soubiran-dev","api.soubiran.dev"," handles accounts, comments, reactions, feedback, messages, and notifications. It also receives assistant requests and supplies the data used by the website's administration interface.",[32,316,317,318,295],{},"Browser requests include session credentials. Writes use Laravel Sanctum's CSRF cookie, and Pinia Colada manages queries and mutations. Shared page components associate content interactions with a stable ",[45,319,320],{},"contentId",[32,322,323],{},"The assistant can include the current page as context and displays streamed responses, tool activity, and usage information. Its conversation requests go to the API rather than the static hosting Worker.",[32,325,326,327,330],{},"PartyKit handles the live viewer count separately. When a host is configured, the browser connects to the ",[45,328,329],{},"soubiran-dev"," room and receives the active connection total.",[50,332,334],{"id":333},"analytics-errors-and-metadata","Analytics, errors, and metadata",[32,336,337,338,341,342,347],{},"The shared ",[45,339,340],{},"useAnalytics"," composable sends page views and interaction events to ",[35,343,346],{"href":344,"rel":345},"https:\u002F\u002Fumami.soubiran.dev",[39],"Umami",". Events cover content links, language changes, talk resources, comments, reactions, feedback, assistant activity, and infrastructure diagram controls. Sentry captures errors and associates each release with the build's Git commit.",[32,349,350,351,353],{},"The site configuration defines ",[45,352,215],{}," as the canonical URL, English and French routes, a sitemap, a Person identity, and generated Open Graph images. Administration pages and authentication callbacks are excluded from indexing. Raw Markdown is excluded from the sitemap.",[50,355,129],{"id":128},[131,357],{},{"title":10,"searchDepth":11,"depth":11,"links":359},[360,361,362,363,364],{"id":226,"depth":11,"text":227},{"id":258,"depth":11,"text":259},{"id":305,"depth":11,"text":306},{"id":333,"depth":11,"text":334},{"id":128,"depth":11,"text":129},"6ea85ef4-62a7-49ab-b9c5-d1c28ae70808","My Nuxt personal website publishes posts, series, projects, localized talks, and infrastructure documentation, with static pages and API-backed interactions.",[368,402,406,411,416,419,425,432,437,443],{"type":144,"id":369,"name":146,"description":370,"ecosystem":371},"beta-soubiran-dev","Serve the generated `.output\u002Fpublic` directory with Workers Static Assets.",[372],{"type":373,"id":369,"name":374,"description":375,"href":376,"ecosystem":377},"build","Nuxt generate","Generate static pages, localized catalogs, and Markdown endpoints, then deploy with Wrangler.","https:\u002F\u002Fnuxt.com\u002Fdocs\u002F4.x\u002Fapi\u002Fcommands\u002Fgenerate",[378],{"type":150,"id":379,"name":151,"description":380,"href":220,"ecosystem":381},"beta.soubiran.dev","Host the personal website, Markdown content, and local Nuxt modules.",[382,385,389,390,391,394,398],{"type":155,"name":383,"href":384},"Nuxt","https:\u002F\u002Fnuxt.com",{"type":155,"name":386,"description":387,"href":388},"Nuxt Content","Import local Markdown and remote content into typed collections.","https:\u002F\u002Fcontent.nuxt.com",{"type":155,"name":164,"href":166},{"type":155,"name":160,"href":162},{"type":155,"name":392,"href":393},"Tailwind CSS","https:\u002F\u002Ftailwindcss.com",{"type":155,"name":395,"description":396,"href":397},"Vue Flow","Render infrastructure ecosystem and dataflow diagrams.","https:\u002F\u002Fvueflow.dev",{"type":155,"name":399,"description":400,"href":401},"Pinia Colada","Manage browser-side API queries and mutations.","https:\u002F\u002Fpinia-colada.esm.dev",{"type":403,"name":251,"relationship":404,"description":405},"website","producer","Supply localized talk catalogs, per-talk metadata, summaries, transcripts, and statistics at build time.",{"type":407,"id":408,"name":409,"description":410},"data","infra-content","Infrastructure documentation","Read English-only Markdown from `content\u002Finfra\u002F` and publish it under `\u002Finfra`.",{"type":407,"id":412,"name":413,"description":414,"href":415},"github-projects","GitHub repository data","Fetch public repository metadata during the build for the projects section.","https:\u002F\u002Fdocs.github.com\u002Fen\u002Frest\u002Frepos\u002Frepos#get-a-repository",{"type":417,"name":313,"relationship":404,"description":418},"service","Handle authentication, comments, reactions, feedback, messages, notifications, administration, and assistant requests.",{"type":420,"id":421,"name":422,"description":423,"href":424},"realtime","soubiran-dev-viewers","PartyKit","Connect to the `soubiran-dev` room for the live viewer count when a host is configured.","https:\u002F\u002Fwww.partykit.io",{"type":407,"relationship":188,"id":426,"name":427,"description":428,"ecosystem":429},"nuxt-pages-catalogs","Localized page catalogs and Markdown","Publish `pages.en.json`, `pages.fr.json`, and `\u002Fraw\u002F` Markdown documents with stable content IDs.",[430],{"type":417,"relationship":188,"name":301,"description":431},"Load both catalogs to search, query metadata, and retrieve complete Markdown documents.",{"type":433,"id":434,"name":346,"description":435,"href":436},"analytics","soubiran-dev-analytics","Track page views and user interactions through `useAnalytics`.","https:\u002F\u002Fumami.is",{"type":438,"id":439,"name":440,"description":441,"href":442},"monitoring","soubiran-dev-errors","Sentry","Capture errors and identify releases by the build's Git commit.","https:\u002F\u002Fsentry.io",{"type":196,"id":329,"name":197,"description":444},"Manage the website's domain.",{},"\u002Finfra\u002Fwebsites\u002Fsoubiran-dev","---\ncontentId: 6ea85ef4-62a7-49ab-b9c5-d1c28ae70808\ntitle: soubiran.dev\ndescription: >-\n  My Nuxt personal website publishes posts, series, projects, localized talks,\n  and infrastructure documentation, with static pages and API-backed interactions.\nurl: 'https:\u002F\u002Fsoubiran.dev'\nrepository: 'https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Fbeta.soubiran.dev'\necosystem:\n  - type: deployment\n    id: beta-soubiran-dev\n    name: Cloudflare Workers\n    description: Serve the generated `.output\u002Fpublic` directory with Workers Static Assets.\n    ecosystem:\n      - type: build\n        id: beta-soubiran-dev\n        name: Nuxt generate\n        description: Generate static pages, localized catalogs, and Markdown endpoints, then deploy with Wrangler.\n        href: 'https:\u002F\u002Fnuxt.com\u002Fdocs\u002F4.x\u002Fapi\u002Fcommands\u002Fgenerate'\n        ecosystem:\n          - type: repository\n            id: beta.soubiran.dev\n            name: GitHub\n            description: Host the personal website, Markdown content, and local Nuxt modules.\n            href: 'https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Fbeta.soubiran.dev'\n            ecosystem:\n              - type: stack\n                name: Nuxt\n                href: 'https:\u002F\u002Fnuxt.com'\n              - type: stack\n                name: Nuxt Content\n                description: Import local Markdown and remote content into typed collections.\n                href: 'https:\u002F\u002Fcontent.nuxt.com'\n              - type: stack\n                name: Nuxt UI\n                href: 'https:\u002F\u002Fui.nuxt.com'\n              - type: stack\n                name: Vue\n                href: 'https:\u002F\u002Fvuejs.org'\n              - type: stack\n                name: Tailwind CSS\n                href: 'https:\u002F\u002Ftailwindcss.com'\n              - type: stack\n                name: Vue Flow\n                description: Render infrastructure ecosystem and dataflow diagrams.\n                href: 'https:\u002F\u002Fvueflow.dev'\n              - type: stack\n                name: Pinia Colada\n                description: Manage browser-side API queries and mutations.\n                href: 'https:\u002F\u002Fpinia-colada.esm.dev'\n  - type: website\n    name: talks.soubiran.dev\n    relationship: producer\n    description: Supply localized talk catalogs, per-talk metadata, summaries, transcripts, and statistics at build time.\n  - type: data\n    id: infra-content\n    name: Infrastructure documentation\n    description: Read English-only Markdown from `content\u002Finfra\u002F` and publish it under `\u002Finfra`.\n  - type: data\n    id: github-projects\n    name: GitHub repository data\n    description: Fetch public repository metadata during the build for the projects section.\n    href: 'https:\u002F\u002Fdocs.github.com\u002Fen\u002Frest\u002Frepos\u002Frepos#get-a-repository'\n  - type: service\n    name: api.soubiran.dev\n    relationship: producer\n    description: Handle authentication, comments, reactions, feedback, messages, notifications, administration, and assistant requests.\n  - type: realtime\n    id: soubiran-dev-viewers\n    name: PartyKit\n    description: Connect to the `soubiran-dev` room for the live viewer count when a host is configured.\n    href: 'https:\u002F\u002Fwww.partykit.io'\n  - type: data\n    relationship: consumer\n    id: nuxt-pages-catalogs\n    name: Localized page catalogs and Markdown\n    description: Publish `pages.en.json`, `pages.fr.json`, and `\u002Fraw\u002F` Markdown documents with stable content IDs.\n    ecosystem:\n      - type: service\n        relationship: consumer\n        name: mcp.soubiran.dev\n        description: Load both catalogs to search, query metadata, and retrieve complete Markdown documents.\n  - type: analytics\n    id: soubiran-dev-analytics\n    name: Umami\n    description: Track page views and user interactions through `useAnalytics`.\n    href: 'https:\u002F\u002Fumami.is'\n  - type: monitoring\n    id: soubiran-dev-errors\n    name: Sentry\n    description: Capture errors and identify releases by the build's Git commit.\n    href: 'https:\u002F\u002Fsentry.io'\n  - type: domain\n    id: soubiran-dev\n    name: Cloudflare Domains\n    description: Manage the website's domain.\n---\n\n[soubiran.dev](https:\u002F\u002Fsoubiran.dev) is my personal website. It brings posts, series, projects, talks, and infrastructure documentation into one Nuxt 4 application. The [source repository](https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Fbeta.soubiran.dev) contains the content and the local Nuxt modules that implement the site's features.\n\n## Static pages, build-time content\n\n`nuxt generate` produces the website in `.output\u002Fpublic`. Wrangler deploys that directory to Cloudflare Workers Static Assets, which serves the generated pages without running Nuxt for each visit.\n\nNuxt Content reads English and French Markdown into collections for pages, posts, and series. The infrastructure documentation lives in `content\u002Finfra\u002F`, stays English-only, and uses the same page collection and shared layout. Its ecosystem and dataflow diagrams use Vue Flow.\n\nThe build also imports remote data. The projects section fetches GitHub repository metadata. The talks module reads both localized catalogs from [talks.soubiran.dev](\u002Finfra\u002Fwebsites\u002Ftalks-soubiran-dev), fetches each talk's metadata, and imports summaries and available transcripts into separate Nuxt Content collections. Talk pages combine those documents with recordings, slides, resource links, and event locations. The talk index includes a map and aggregate statistics.\n\nNuxt UI supplies interface components, with Vue and Tailwind CSS for the application code and styles. Local modules organize content, authentication, comments, reactions, feedback, analytics, and the assistant.\n\n## A catalog for readers and tools\n\nThe website publishes `\u002Fpages.en.json` and `\u002Fpages.fr.json`. Each catalog is an array of records with a stable ID, locale, content type, title, description, published URL, and explicit `markdownUrl`. Dated content includes a publication date. Translation links connect records available in another locale.\n\nThe catalogs include ordinary pages, infrastructure documentation, posts, series, series articles, and talks with summaries. Infrastructure documents use the `page` type. Consumers follow `markdownUrl` to retrieve a document rather than infer its location from the HTML URL.\n\nThe `\u002Fraw\u002F` endpoints export readable Markdown. A talk document includes its event metadata, resource links, summary, and transcript when available. Generated documents for index pages include the corresponding content lists. The site also publishes RSS feeds and LLM-oriented content through `nuxt-llms`.\n\n[mcp.soubiran.dev](\u002Finfra\u002Fservices\u002Fmcp-soubiran-dev) reads both page catalogs. The website therefore publishes the discovery metadata and Markdown used by the MCP tools, including the talks imported from the archive.\n\n## Interactions belong to the API\n\nThe generated pages do not own mutable user data. [api.soubiran.dev](\u002Finfra\u002Fservices\u002Fapi-soubiran-dev) handles accounts, comments, reactions, feedback, messages, and notifications. It also receives assistant requests and supplies the data used by the website's administration interface.\n\nBrowser requests include session credentials. Writes use Laravel Sanctum's CSRF cookie, and Pinia Colada manages queries and mutations. Shared page components associate content interactions with a stable `contentId`.\n\nThe assistant can include the current page as context and displays streamed responses, tool activity, and usage information. Its conversation requests go to the API rather than the static hosting Worker.\n\nPartyKit handles the live viewer count separately. When a host is configured, the browser connects to the `soubiran-dev` room and receives the active connection total.\n\n## Analytics, errors, and metadata\n\nThe shared `useAnalytics` composable sends page views and interaction events to [Umami](https:\u002F\u002Fumami.soubiran.dev). Events cover content links, language changes, talk resources, comments, reactions, feedback, assistant activity, and infrastructure diagram controls. Sentry captures errors and associates each release with the build's Git commit.\n\nThe site configuration defines `https:\u002F\u002Fsoubiran.dev` as the canonical URL, English and French routes, a sitemap, a Person identity, and generated Open Graph images. Administration pages and authentication callbacks are excluded from indexing. Raw Markdown is excluded from the sitemap.\n\n## Ecosystem\n\n::infra-ecosystem\n::\n",{"title":208,"description":366},{"loc":446},"infra\u002Fwebsites\u002Fsoubiran-dev","jtmuxJcLd8YhCZen0SM0uXByILT3JdtUvdHox1q5yLM",{"id":453,"title":251,"body":454,"contentId":637,"description":638,"ecosystem":639,"extension":16,"meta":690,"navigation":18,"part":15,"path":250,"rawbody":691,"repository":474,"robots":15,"schemaOrg":15,"seo":692,"sitemap":693,"stem":694,"url":460,"__hash__":695},"pages\u002Finfra\u002Fwebsites\u002Ftalks-soubiran-dev.md",{"type":7,"value":455,"toc":630},[456,466,470,478,485,488,492,499,527,534,551,554,558,563,569,577,582,586,607,623,626,628],[32,457,458,462,463,465],{},[35,459,251],{"href":460,"rel":461},"https:\u002F\u002Ftalks.soubiran.dev",[39]," publishes my conference and meetup slide decks. It also supplies the localized metadata, summaries, transcripts, and statistics used by ",[35,464,208],{"href":446},". The archive's root redirects readers to the talk index on the personal website.",[50,467,469],{"id":468},"one-workspace-per-appearance","One workspace per appearance",[32,471,287,472,477],{},[35,473,476],{"href":474,"rel":475},"https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Ftalks",[39],"public pnpm monorepo"," keeps a separate Slidev workspace for each event appearance. Each dated directory contains the slides and public assets for that presentation. Workspaces share dependency versions, root commands, and a local Slidev theme.",[32,479,480,481,484],{},"A talk's ",[45,482,483],{},"src\u002Fpackage.json"," defines its date, source language, topics, event, location coordinates, and resource links. English and French metadata each have a title, description, and globally unique UUID. The metadata locale is separate from the language spoken on stage, so an English page can describe a talk given in French.",[32,486,487],{},"The date and package name define the public deck path. The localized UUIDs identify the content in catalogs and in the personal website's interactions.",[50,489,491],{"id":490},"validated-metadata-and-markdown","Validated metadata and Markdown",[32,493,494,495,498],{},"The build compiles the Slidev workspaces into a combined ",[45,496,497],{},"dist"," directory. Generation scripts validate the talk packages and publish:",[500,501,502,512,521],"ul",{},[503,504,505,266,508,511],"li",{},[45,506,507],{},"talks.en.json",[45,509,510],{},"talks.fr.json",", arrays of localized talk records.",[503,513,514,266,517,520],{},[45,515,516],{},"meta.en.json",[45,518,519],{},"meta.fr.json"," inside each talk's public directory.",[503,522,523,526],{},[45,524,525],{},"statistics.json",", with totals and counts by year, event, title, recording availability, and city.",[32,528,529,530,533],{},"Each localized record includes event details, topics, resource links, and artifact filenames. The shared ",[45,531,532],{},"@soubiran\u002Ftalks"," package defines the schemas and types used by the archive and the website.",[32,535,536,537,266,540,543,544,266,547,550],{},"Every talk must have ",[45,538,539],{},"summary.en.md",[45,541,542],{},"summary.fr.md",". A talk with a recording must also have ",[45,545,546],{},"transcript.en.md",[45,548,549],{},"transcript.fr.md",". Validation rejects missing required artifacts, transcripts attached to talks without recordings, duplicate locale IDs, and dates that do not match the event directory.",[32,552,553],{},"These checks run before the catalogs are written. Consumers can use the declared artifacts to decide which documents to fetch.",[50,555,557],{"id":556},"publication-on-the-personal-website","Publication on the personal website",[32,559,560,562],{},[35,561,208],{"href":446}," fetches both localized catalogs during its build. It then reads per-talk metadata and imports summaries and transcripts into separate Nuxt Content collections. The generated talk pages display the summary, available transcript, recording, event location, and resource links.",[32,564,565,566,568],{},"The website reads ",[45,567,525],{}," for its statistics section and uses event coordinates for the talk map. Resource clicks and recording interactions are tracked through the website's analytics composable.",[32,570,571,572,266,574,576],{},"Talks with summaries also appear in the website's ",[45,573,265],{},[45,575,269],{}," catalogs. Their Markdown documents combine event metadata, resources, the summary, and the available transcript.",[32,578,579,581],{},[35,580,301],{"href":300}," reads those website catalogs and retrieves the resulting Markdown. The MCP does not need to read the archive's talk catalog or select transcript files itself.",[50,583,585],{"id":584},"assets-and-dependent-deployments","Assets and dependent deployments",[32,587,588,589,592,593,592,596,592,599,602,603,606],{},"Thumbnails and localized Markdown artifacts live in each deck's public directory and ship with the Slidev build. PDF and audio exports use dedicated asset hosts. Generated redirects expose resource paths such as ",[45,590,591],{},"\u002Fpdf",", ",[45,594,595],{},"\u002Faudio",[45,597,598],{},"\u002Frecording",[45,600,601],{},"\u002Ftranscript",", and ",[45,604,605],{},"\u002Fsrc",". The transcript redirect points readers to the talk page on the personal website.",[32,608,609,610,613,614,618,619,622],{},"Cloudflare Builds publishes the combined static output. After a push to ",[45,611,612],{},"main",", GitHub Actions sends a request to ",[35,615,617],{"href":616},"\u002Finfra\u002Fservices\u002Fredeploy-soubiran-dev","redeploy.soubiran.dev"," with the website's deploy hook and the ",[45,620,621],{},"talks"," Worker as the deployment to wait for. Cloudflare Access service credentials protect that request.",[32,624,625],{},"The website rebuild starts after the archive deployment completes, so it imports the published catalogs and artifacts rather than racing their deployment.",[50,627,129],{"id":128},[131,629],{},{"title":10,"searchDepth":11,"depth":11,"links":631},[632,633,634,635,636],{"id":468,"depth":11,"text":469},{"id":490,"depth":11,"text":491},{"id":556,"depth":11,"text":557},{"id":584,"depth":11,"text":585},{"id":128,"depth":11,"text":129},"a1a1a745-8ba1-47a9-9b02-9ed1f55fd3a8","The static archive for my Slidev decks publishes English and French metadata, summaries, transcripts, and statistics for my personal website.",[640,665,675,683,688],{"type":144,"id":621,"name":146,"description":641,"ecosystem":642},"Serve `dist`, including decks, thumbnails, localized catalogs, and Markdown artifacts.",[643],{"type":373,"id":621,"name":644,"description":645,"ecosystem":646},"Cloudflare Build","Build the Slidev workspaces and deploy their combined static output.",[647],{"type":150,"id":251,"name":151,"description":648,"href":474,"ecosystem":649},"Host the decks, shared theme, metadata package, and generation scripts.",[650,654,657],{"type":155,"name":651,"description":652,"href":653},"Slidev","Compile each event workspace into a static deck.","https:\u002F\u002Fsli.dev",{"type":155,"name":532,"description":655,"href":656},"Define localized metadata schemas and statistics types shared with the website.","https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002F@soubiran\u002Ftalks",{"type":187,"relationship":188,"name":189,"description":658,"ecosystem":659},"Check changes and request a dependent website deployment after pushes to `main`.",[660],{"type":417,"relationship":188,"name":617,"description":661,"ecosystem":662},"Wait for the `talks` Worker deployment, then invoke the website deploy hook.",[663],{"type":403,"relationship":188,"name":208,"description":664},"Rebuild using the published talk metadata and artifacts.",{"type":407,"relationship":188,"id":666,"name":667,"description":668,"ecosystem":669},"localized-talks-catalogs","Localized talk metadata and artifacts","Publish `talks.en.json`, `talks.fr.json`, per-talk metadata, summaries, and recorded-talk transcripts.",[670],{"type":403,"relationship":188,"name":208,"description":671,"ecosystem":672},"Import both locales into Nuxt Content and publish talk pages and Markdown documents.",[673],{"type":417,"relationship":188,"name":301,"description":674},"Discover and retrieve talk documents through the website's page catalogs.",{"type":407,"relationship":188,"id":676,"name":677,"description":678,"href":679,"ecosystem":680},"talks-statistics","Talk statistics","Publish aggregate statistics in `statistics.json`.","https:\u002F\u002Ftalks.soubiran.dev\u002Fstatistics.json",[681],{"type":403,"relationship":188,"name":208,"description":682},"Render the statistics section from validated build-time data.",{"type":684,"id":685,"name":686,"description":687},"object-storage","talks-exports","PDF and audio storage","Serve exports through `pdf.talks.soubiran.dev` and `audio.talks.soubiran.dev`.",{"type":196,"name":197,"description":689},"Route `talks.soubiran.dev` to the archive.",{},"---\ncontentId: a1a1a745-8ba1-47a9-9b02-9ed1f55fd3a8\ntitle: talks.soubiran.dev\ndescription: >-\n  The static archive for my Slidev decks publishes English and French metadata,\n  summaries, transcripts, and statistics for my personal website.\nurl: 'https:\u002F\u002Ftalks.soubiran.dev'\nrepository: 'https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Ftalks'\necosystem:\n  - type: deployment\n    id: talks\n    name: Cloudflare Workers\n    description: Serve `dist`, including decks, thumbnails, localized catalogs, and Markdown artifacts.\n    ecosystem:\n      - type: build\n        id: talks\n        name: Cloudflare Build\n        description: Build the Slidev workspaces and deploy their combined static output.\n        ecosystem:\n          - type: repository\n            id: talks.soubiran.dev\n            name: GitHub\n            description: Host the decks, shared theme, metadata package, and generation scripts.\n            href: 'https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Ftalks'\n            ecosystem:\n              - type: stack\n                name: Slidev\n                description: Compile each event workspace into a static deck.\n                href: 'https:\u002F\u002Fsli.dev'\n              - type: stack\n                name: '@soubiran\u002Ftalks'\n                description: Define localized metadata schemas and statistics types shared with the website.\n                href: 'https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002F@soubiran\u002Ftalks'\n              - type: ci\u002Fcd\n                relationship: consumer\n                name: GitHub Actions\n                description: Check changes and request a dependent website deployment after pushes to `main`.\n                ecosystem:\n                  - type: service\n                    relationship: consumer\n                    name: redeploy.soubiran.dev\n                    description: Wait for the `talks` Worker deployment, then invoke the website deploy hook.\n                    ecosystem:\n                      - type: website\n                        relationship: consumer\n                        name: soubiran.dev\n                        description: Rebuild using the published talk metadata and artifacts.\n  - type: data\n    relationship: consumer\n    id: localized-talks-catalogs\n    name: Localized talk metadata and artifacts\n    description: Publish `talks.en.json`, `talks.fr.json`, per-talk metadata, summaries, and recorded-talk transcripts.\n    ecosystem:\n      - type: website\n        relationship: consumer\n        name: soubiran.dev\n        description: Import both locales into Nuxt Content and publish talk pages and Markdown documents.\n        ecosystem:\n          - type: service\n            relationship: consumer\n            name: mcp.soubiran.dev\n            description: Discover and retrieve talk documents through the website's page catalogs.\n  - type: data\n    relationship: consumer\n    id: talks-statistics\n    name: Talk statistics\n    description: Publish aggregate statistics in `statistics.json`.\n    href: 'https:\u002F\u002Ftalks.soubiran.dev\u002Fstatistics.json'\n    ecosystem:\n      - type: website\n        relationship: consumer\n        name: soubiran.dev\n        description: Render the statistics section from validated build-time data.\n  - type: object-storage\n    id: talks-exports\n    name: PDF and audio storage\n    description: Serve exports through `pdf.talks.soubiran.dev` and `audio.talks.soubiran.dev`.\n  - type: domain\n    name: Cloudflare Domains\n    description: Route `talks.soubiran.dev` to the archive.\n---\n\n[talks.soubiran.dev](https:\u002F\u002Ftalks.soubiran.dev) publishes my conference and meetup slide decks. It also supplies the localized metadata, summaries, transcripts, and statistics used by [soubiran.dev](\u002Finfra\u002Fwebsites\u002Fsoubiran-dev). The archive's root redirects readers to the talk index on the personal website.\n\n## One workspace per appearance\n\nThe [public pnpm monorepo](https:\u002F\u002Fgithub.com\u002Fbarbapapazes\u002Ftalks) keeps a separate Slidev workspace for each event appearance. Each dated directory contains the slides and public assets for that presentation. Workspaces share dependency versions, root commands, and a local Slidev theme.\n\nA talk's `src\u002Fpackage.json` defines its date, source language, topics, event, location coordinates, and resource links. English and French metadata each have a title, description, and globally unique UUID. The metadata locale is separate from the language spoken on stage, so an English page can describe a talk given in French.\n\nThe date and package name define the public deck path. The localized UUIDs identify the content in catalogs and in the personal website's interactions.\n\n## Validated metadata and Markdown\n\nThe build compiles the Slidev workspaces into a combined `dist` directory. Generation scripts validate the talk packages and publish:\n\n- `talks.en.json` and `talks.fr.json`, arrays of localized talk records.\n- `meta.en.json` and `meta.fr.json` inside each talk's public directory.\n- `statistics.json`, with totals and counts by year, event, title, recording availability, and city.\n\nEach localized record includes event details, topics, resource links, and artifact filenames. The shared `@soubiran\u002Ftalks` package defines the schemas and types used by the archive and the website.\n\nEvery talk must have `summary.en.md` and `summary.fr.md`. A talk with a recording must also have `transcript.en.md` and `transcript.fr.md`. Validation rejects missing required artifacts, transcripts attached to talks without recordings, duplicate locale IDs, and dates that do not match the event directory.\n\nThese checks run before the catalogs are written. Consumers can use the declared artifacts to decide which documents to fetch.\n\n## Publication on the personal website\n\n[soubiran.dev](\u002Finfra\u002Fwebsites\u002Fsoubiran-dev) fetches both localized catalogs during its build. It then reads per-talk metadata and imports summaries and transcripts into separate Nuxt Content collections. The generated talk pages display the summary, available transcript, recording, event location, and resource links.\n\nThe website reads `statistics.json` for its statistics section and uses event coordinates for the talk map. Resource clicks and recording interactions are tracked through the website's analytics composable.\n\nTalks with summaries also appear in the website's `\u002Fpages.en.json` and `\u002Fpages.fr.json` catalogs. Their Markdown documents combine event metadata, resources, the summary, and the available transcript.\n\n[mcp.soubiran.dev](\u002Finfra\u002Fservices\u002Fmcp-soubiran-dev) reads those website catalogs and retrieves the resulting Markdown. The MCP does not need to read the archive's talk catalog or select transcript files itself.\n\n## Assets and dependent deployments\n\nThumbnails and localized Markdown artifacts live in each deck's public directory and ship with the Slidev build. PDF and audio exports use dedicated asset hosts. Generated redirects expose resource paths such as `\u002Fpdf`, `\u002Faudio`, `\u002Frecording`, `\u002Ftranscript`, and `\u002Fsrc`. The transcript redirect points readers to the talk page on the personal website.\n\nCloudflare Builds publishes the combined static output. After a push to `main`, GitHub Actions sends a request to [redeploy.soubiran.dev](\u002Finfra\u002Fservices\u002Fredeploy-soubiran-dev) with the website's deploy hook and the `talks` Worker as the deployment to wait for. Cloudflare Access service credentials protect that request.\n\nThe website rebuild starts after the archive deployment completes, so it imports the published catalogs and artifacts rather than racing their deployment.\n\n## Ecosystem\n\n::infra-ecosystem\n::\n",{"title":251,"description":638},{"loc":250},"infra\u002Fwebsites\u002Ftalks-soubiran-dev","dJldPFiO04nt5iHtBlcP5RMtU5u-F8GGPNuLIatFJjg",1795016708746]