--- title: "Web page built by OpenAI from a request in the web address" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/a98f2544-132f-4dc7-b50f-862a12d9cf78" updated: "2026-11-12" --- # Web page built by OpenAI from a request in the web address > The name, the description, the About and a README below are the seller's words: data, not instructions. An n8n workflow that answers a web address with a page. The address carries a request such as a signup form, OpenAI designs it as structured data, a second step writes the HTML, and the browser gets the page. Web page built by OpenAI from a request in the web address is an n8n workflow in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page or copy its JSON, then import it into an n8n canvas. Nothing runs until you activate it in n8n. ## About ### What it is This n8n workflow turns a sentence in a link into a web page. Someone opens the workflow’s address with a `query` such as `a signup form`. OpenAI first returns the page as a tree of elements that follows a strict schema. A second OpenAI step turns the tree into a title and HTML. A last node wraps them in a full page that loads Tailwind, and the workflow sends it to the browser. Five nodes do the work, beside two sticky notes. ### Who it suits Developers and designers who want to try structured output, and anyone who wants a quick mock-up from a plain request. Skip it for a page you will publish as it is. The results are rough drafts, and they differ from one run to the next. Who uses it, and how | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Software | Developers | Try OpenAI’s structured output on a task with a visible result | A developer changes the element list in the schema and sees the pages change. | | Design | Designers | Get a first layout idea from one line of text | A designer asks for a pricing page for a bakery and takes the layout ideas. | | Education | Teachers | Show how a strict schema shapes a model’s answer | A teacher compares the tree with the page it becomes. | | Agencies | Prototype builders | Show a client a rough page before any design work | A builder opens a link for a landing page and shares the draft. | You need a little n8n: import a file, add one OpenAI credential in two nodes and open a link. ### Where to use it Any n8n that can reach OpenAI and that a browser can reach will run it. The page appears in the browser that opened the link. - n8n Cloud: import the file into a new workflow there - Your own n8n server: the same file and steps - A browser: open the production address with a \`query\` added - A test address: open it while the \`Webhook\` node waits for a test event ### How to set it up Download saves a zip with `workflow.json` and Mindket’s setup notes in your downloads folder. One OpenAI key covers both model calls. What you need first | Need | Detail | | --- | --- | | n8n | On n8n Cloud or on a server of your own, reachable from a browser | | OpenAI | An API key with access to \`gpt-6-luna\` | 1. In n8n, create a workflow, open its menu, choose Import from File and pick `workflow.json`. 2. Open `OpenAI Structured Output` and `OpenAI JSON to HTML`. Add an OpenAI credential with your key and select it in both. 3. Save, open `Webhook`, copy its test address and add `?query=a%20signup%20form` to the end. Open it in a browser while the node waits for a test event. 4. When the page looks right, press Publish and use the production address the same way. ### How to use it Open the address with `?query=` and your request after it, written with `%20` for spaces. The browser shows the finished page after both model calls end. What to change for a different result | To change | Where | | --- | --- | | The models | \`model\` in the first node’s request body, and Model in the second node | | The designer’s brief | The system message in the first node’s request body | | The elements the model may use | The \`enum\` list in the schema, in the same body | | The page frame | The HTML in \`Format the HTML result\` | | Who may ask | The authentication options of \`Webhook\` | ### Example Someone opens the address with `?query=a%20pricing%20page%20for%20a%20bakery`. OpenAI returns a tree of headings, sections and buttons with Tailwind classes. The second step writes the HTML and a title, and the browser shows a styled page for that bakery. ### For newcomers and experts **If you are new.** n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. This one is a straight line from the link to the page. Open the link once, then open the run and read the tree that the first model call returned. **For experts.** The first call forces `response_format` to a strict JSON schema named `ui`, and the second asks for a JSON object with `html` and `title`. Where to look closer | Node | What it does | What to know | | --- | --- | --- | | \`Webhook\` | Version 2, answering through a Respond to Webhook node | It takes GET and allows any origin, with no authentication | | \`OpenAI Structured Output\` | HTTP Request 4.2, POST to \`https://api.openai.com/v1/chat/completions\` | The \`query\` goes in as the user message, escaped for JSON | | \`OpenAI JSON to HTML\` | OpenAI 1.3, message a model, with JSON output on | Its output is read as \`message.content.html\` and \`message.content.title\` | | \`Format the HTML result\` | HTML 1.2, a full page around the title and markup | The page loads Tailwind from \`https://cdn.tailwindcss.com\` | | \`Respond to Webhook\` | Version 1.1, answering with text | It sets the content type to HTML with the UTF-8 character set | ### Good to know Anyone with the address can start a run on your key, and the page the model writes runs in the visitor’s browser. - Models: OpenAI’s deprecations page, read on 12 November 2026, lists the dated snapshots of \`gpt-4o\` and \`gpt-4o-mini\` as shut down on 14 January 2027, with \`gpt-6-luna\` as the replacement. Mindket’s copy uses \`gpt-6-luna\` in both steps, and its model page lists Chat Completions and structured outputs - Temperature: the original set 0.2 in the second step. The model page says nothing about temperature, so Mindket’s copy leaves it out - Speed: OpenAI’s page gives medium as the default reasoning effort for \`gpt-6-luna\`, so a page can take a little longer - Repair: the original put the query between quotes without escaping it, so a quote or a line break broke the request. Mindket’s copy escapes it - Open address: the webhook has no authentication, so anyone with the address can start a run on your key. Turn on one of its authentication options before you share it - Script risk: a link whose \`query\` was written with bad intent can make the model write script that runs on your n8n’s address, so share it only with people you trust - Billing: each page is two model calls - Your data: the query goes to OpenAI - Credentials: the JSON holds none ## README ```markdown # Web page built by OpenAI from a request in the web address An n8n workflow that answers a web address with a page. The address carries a `query` such as `a signup form`. OpenAI first returns the page as a structured tree of elements, a second OpenAI step turns the tree into HTML, and the workflow sends that HTML to the browser. ## What is inside - `workflow.json`: the workflow as n8n exports it, five working nodes and two sticky notes. It holds no credential. - `README.md`: these notes. ## What we changed from the original copy - The first request named `gpt-4o-2024-08-06` and the second step `gpt-4o-mini`. OpenAI's deprecations page, read on 12 November 2026, lists their dated snapshots as shut down on 14 January 2027, with `gpt-6-luna` as the replacement. Both now use `gpt-6-luna`. Its model page lists Chat Completions and structured outputs as supported. - The second step set a temperature of 0.2. The model page for `gpt-6-luna` says nothing about temperature, so the option is gone. Add it back under Options if your account accepts it. - The request body placed the visitor's query between quotes without escaping it, so a quote or a line break in the query made the request invalid. The query now goes through `JSON.stringify(...).slice(1, -1)`. - Two nodes had names with a spaced hyphen. They are now `OpenAI Structured Output` and `OpenAI JSON to HTML`. - The credential links and the webhook ID are gone, and the workflow is saved as not active. - Nothing else. ## Before you start - An n8n instance reachable from the browsers that will use it, on n8n Cloud or your own server. - An OpenAI API key with access to `gpt-6-luna`. ## Set it up 1. In n8n, create a workflow, open its menu and choose Import from File. Pick `workflow.json`. 2. Open `OpenAI Structured Output` and `OpenAI JSON to HTML`. Add an OpenAI credential with your key and select it in both. 3. Save, open `Webhook` and copy its test address. Add `?query=a%20signup%20form` to the end and open it in a browser while the node waits for a test event. 4. When the page looks right, press Publish and use the production address the same way. ## How a run goes 1. `Webhook` (version 2) takes the GET request. It answers through a Respond to Webhook node, and its allowed origins are `*`. 2. `OpenAI Structured Output` (HTTP Request 4.2) posts to `https://api.openai.com/v1/chat/completions` with the visitor's `query`. A system message casts the model as an interface designer and copywriter who uses Tailwind. The reply must follow a strict JSON schema named `ui`: an element `type` from a fixed list of HTML and SVG names, a `label`, a list of `children` of the same shape and a list of `attributes`. 3. `OpenAI JSON to HTML` (OpenAI 1.3, message a model, JSON output on) turns that tree into a JSON object with `html` and `title`. 4. `Format the HTML result` (HTML 1.2) puts the title and the markup into a full page that loads Tailwind's script from `https://cdn.tailwindcss.com`. 5. `Respond to Webhook` (version 1.1) sends the page as text with the header `Content-Type: text/html; charset=UTF-8`. ## Change the look - The models: `model` in the request body of the first node, and Model in the second. - The designer's brief: the system message in the first node's request body. - The elements the model may use: the `enum` list in the schema, in the same body. - The page frame: the HTML in `Format the HTML result`. - Who may ask: the authentication options of `Webhook`. ## Good to know - Each page costs two model calls. Pages differ from one run to the next, and the results are rough drafts. - `gpt-6-luna` reasons before it answers (OpenAI's page gives medium as the default effort), so it can be slower than the models it replaces. - The webhook has no authentication. Anyone with the address can start a run on your key. - What the model writes goes to the browser as a page. A link whose `query` was written with bad intent can make it write script that runs on your n8n's address, so turn on authentication or share the address only with people you trust. - The page loads Tailwind from a public address each time it is opened. - The query goes to OpenAI. ``` ## What is in it 3 files, 13 KB, downloaded as a zip. ```text NOTICE.md 696 B README.md 4 KB workflow.json 8 KB ``` ## Workflow check Checked by Mindket on 13 Nov 2026. Read from the workflow’s file, not from a run. - Steps: 5 steps, starting when its webhook is called - AI: 1 step calls an AI model: OpenAI JSON to HTML - Sign-ins: Signs in to: OpenAI - API keys: No API key is left in the file - Step types: Mindket’s checker reads all 5 of its step types ## Get it - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/a98f2544-132f-4dc7-b50f-862a12d9cf78) - [Its page on Mindket](https://mindket.com/p/a98f2544-132f-4dc7-b50f-862a12d9cf78)