/ai: one agent picker in the hero, Expert first - #5670
Conversation
…t section Splits the Connect capability group into inbound and outbound so the two MCP directions cannot be confused, scopes the approval-card governance claim to Expert and Platform Automations, and replaces the hero prompt widget with a two-path chooser. The widget's components and script are left in place.
✅ Deploy Preview for flowforge-website ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
|
Preview affected pages:
from the changelog, |
…wn agent Restructures the page so the reader understands the capabilities before being asked to choose. Adds a FlowFuse Expert section built the same way as the connect section, so both paths have a destination and a CTA. Also fixes the diagram rendering at zero height, and stops the disabled Expert modal and its script from shipping: an HTML comment does not prevent Nunjucks executing an include.
dbad24d to
d6fe87d
Compare
dimitrieh
changed the title
Replaces the two path sections and the fork cards with a single tabbed picker in the hero, FlowFuse Expert first, so the choice is made once. Adds the five agent marks, an auto-rotating showcase well with per-tab timing, and media wells on the capability items. Moves AI Governance directly below the hero as shared ground. Also drops a bespoke copy button in favour of the existing ff-command component, which has the non-secure-context fallback the bespoke one lacked.
dimitrieh
changed the title
Expert nested inside the platform, an external agent overlapping it from outside. Replaces the placehold.co images, so the page makes no external image requests.
| externalNote: "Anyone can build them. For external agents, whether they can call them depends on the agent." | ||
| status: "Available" | ||
| statusKind: "live" | ||
| description: "Build your own MCP servers and let Insights-mode agents call your tools and services as part of a workflow. Wrap an internal API or system as a tool once, then let agents use it wherever it fits. The agent stays inside the workflow you designed, calling only the tools you register." |
There was a problem hiding this comment.
Another important thing I’d recommend is adding relevant blog posts from the FlowFuse AI blog to this section. There are several articles that support the topics covered on the page and can help build topical authority around AI for industrial applications. If we can’t add them directly to the section, I’d at least recommend linking to the AI blog section naturally within the copy.
There was a problem hiding this comment.
Added a further reading block: the AI blog plus the four posts closest to what the page claims.
| title: "Industrial AI: FlowFuse Expert, Governed and Built-In" | ||
| description: "Industrial AI with FlowFuse: build industrial applications with FlowFuse Expert, act on live operations behind approvals, and give your company-sanctioned AI governed access. Every AI capability on one platform, with the permissions, RBAC, and audit your plants require." | ||
| title: "Industrial AI: Our Agent, or Bring Your Own" | ||
| description: "Industrial AI with FlowFuse: build with FlowFuse Expert, or connect the AI agent your company already approved. Microsoft Copilot, ChatGPT, Claude and local models can query your operations and build Node-RED applications, with the permissions, RBAC and audit your plants require." |
There was a problem hiding this comment.
| description: "Industrial AI with FlowFuse: build with FlowFuse Expert, or connect the AI agent your company already approved. Microsoft Copilot, ChatGPT, Claude and local models can query your operations and build Node-RED applications, with the permissions, RBAC and audit your plants require." | |
| description: "AI for industrial applications, without losing control: use FlowFuse Expert, or connect Copilot, ChatGPT, Claude or a local model to your operations." |
Current one is 280 chars and gets truncated by Google around 155-160, so the RBAC/audit trust signal at the end never actually shows in the SERP snippet. This version leads with the target keyword phrase and fits fully within the limit
| meta: | ||
| title: "Industrial AI: FlowFuse Expert, Governed and Built-In" | ||
| description: "Industrial AI with FlowFuse: build industrial applications with FlowFuse Expert, act on live operations behind approvals, and give your company-sanctioned AI governed access. Every AI capability on one platform, with the permissions, RBAC, and audit your plants require." | ||
| title: "Industrial AI: Our Agent, or Bring Your Own" |
There was a problem hiding this comment.
| title: "Industrial AI: Our Agent, or Bring Your Own" | |
| title: "AI for Industrial Applications: Governed AI for Industry" |
Let's target “AI for industrial applications” on this page, as it has a much lower keyword difficulty. We can target “industrial AI” separately through blog content, since its KD is higher and it will require more topical authority over time. This gives us a more realistic path to ranking instead of competing head-on for a term we’re not yet positioned to win.
One more point: to rank for a keyword, the page needs to actually address that topic. Otherwise, it is unlikely to rank well for that term. Where possible, I’d recommend incorporating relevant content around the keyword. In particular, I think the “Why Edge AI for Industrial Applications” section would be a good place to naturally address it. However, I wouldn’t force this page to include informational content that doesn’t align with its conversion-focused intent.
There was a problem hiding this comment.
Taken verbatim. The H1 now says "AI for Industrial Applications" too, since targeting a phrase the largest heading does not say was the gap you raised. Push back if you would rather keep the H1 short.
| @@ -3,8 +3,8 @@ layout: page | |||
| nohero: true | |||
| imageDescription: "Plant worker interacting with an HMI" | |||
| meta: | |||
There was a problem hiding this comment.
I think we should add FAQs that target LLMs as well. The FAQ section is a good place to naturally incorporate this content, so I’d recommend adding the following alongside the ones you think are most relevant:
- What is AI for industrial applications?
AI for industrial applications means using AI to work with industrial data, equipment, and processes. Common uses include helping engineers build automation applications, analyzing machine data, inspecting products with computer vision, and allowing people or AI agents to interact with production systems. FlowFuse provides a platform for connecting these AI capabilities to industrial data and applications.
- How is AI used in industrial applications?
AI can support different parts of an industrial operation, from engineering and quality inspection to production monitoring and automation. For example, it can help generate application logic, analyze live machine data, identify defects from camera feeds, or let an AI agent interact with industrial systems. FlowFuse brings these capabilities into the same environment as the industrial applications and data they need to work with.
- What are examples of AI for industrial applications?
Examples include AI-assisted engineering, automated visual inspection, natural-language access to production data, AI agents that interact with industrial systems, and AI models running locally at the edge. With FlowFuse, teams can use FlowFuse Expert to build and explain applications, connect external AI agents through MCP, run ONNX vision models, and use cloud or local LLMs within industrial workflows.
There was a problem hiding this comment.
Added, your three plus two on the launch itself: whether you can use your own agent, and how changes are kept from happening unapproved.
Uses the homepage meta.faq shape, so jsonld.njk emits FAQPage off the frontmatter. Answers are plain text, since that partial interpolates them into a JSON string and markup would land in the schema.
| step1: | ||
| title: "Copy the FlowFuse connector URL" | ||
| description: "You will paste this into your agent in the next step." | ||
| step3: |
There was a problem hiding this comment.
There are also several opportunities throughout the page to link to relevant existing pages, such as RTSP, Node-RED, MCP, and ONNX. We should also consider adding links to this page from relevant pages and blog posts across the website to help build its authority and strengthen its internal linking.
There was a problem hiding this comment.
Done. Node-RED, ONNX, RTSP, certified LLM nodes and MCP server nodes, linked where the copy already named them. All verified 200 against production.
Inbound links from other pages are worth a separate pass.
Title and meta description are the reviewer's suggestions verbatim. The description was 280 characters and truncated around 155, so the trust signal at the end never reached the snippet. The H1 now carries "AI for industrial applications" too. Targeting a phrase the largest heading does not say is the gap the reviewer raised. Adds meta.faq with the three questions from the review plus two specific to this launch. It uses the homepage's faq.njk and meta.faq shape, so jsonld.njk emits FAQPage structured data off the frontmatter with no extra wiring. Answers are plain text on purpose: jsonld.njk interpolates them into a JSON string, so markup in an answer would land in the schema. Internal links added where the copy already named the thing: Node-RED, ONNX, RTSP, the certified LLM nodes and MCP server nodes. Plus a further reading block pointing at the AI blog and the four posts closest to what this page claims. All ten targets verified 200 against production.
| </div> | ||
| {% endfor %} | ||
| </div> | ||
| <p class="text-sm text-gray-500 mt-5 mb-0">{{ connect.codingNote }} <a href="/docs/user/expert/third-party-agents/" class="text-indigo-600 font-semibold">See the documentation</a>.</p> |
There was a problem hiding this comment.
Flagging this link: what ships first is not settled yet, and gets decided later today.
/ai and the docs page are landing on different tracks. /ai is #5670 (11ty) or #5681 (Nuxt), still an open renderer question, and we aligned in the Slack thread that it may land the week after release. The docs page is FlowFuse/flowfuse#8241, which is green.
Default for now: point at /ai. It carries the address to copy and the three steps per agent, which is what a reader at this point actually needs. The docs link stays for the detail.
If /ai slips past release day, this line needs the order flipped, so leave the thread open until that call is made.
|
Deprecated in favour of #5681 |
3 participants
Description
Rebuilds
/aiaround a single agent picker instead of a page that asked the visitor to choose twice.Hero carries a tabbed picker: FlowFuse Expert, Microsoft Copilot, ChatGPT, Claude, Gemini, Local and Custom Agents. Three steps per tab. Expert's are "sign in / or use the Device Agent / ask for what you need"; the rest are "copy the URL / add it in your agent / sign in and choose scopes". Expert has no endpoint box, the others do.
Then a showcase well alternating between two schematics on a timer, AI Governance, and the capability band.
Removed: the two path sections, the fork cards, the use-cases section, and the hero CTAs. All were saying what the picker now says.
Notable
adoption.microsoft.com, Google's CDN. Sources and the trademark position are insrc/images/ai/agents/README.md. Microsoft still needs a legal check. A tab with no mark falls back to its name as text.media:in frontmatter.Fixes to my earlier revision on this branch
{% include %}inside an HTML comment; only{# #}prevents it.ff-commandcomponent, which has the non-secure-context fallback and also stops the URL truncating.sign-up-url.njkand housecapture()names rather than hardcoded URLs and invented events.Alternative
#5681 is the same page rebuilt on Nuxt, with
src/ai.njkdeleted. Content review happens here; that one is the renderer question. Whichever we take, the other closes.Related Issue(s)
FlowFuse/flowfuse#8192
Depends on
/docs/user/expert/third-party-agents/lands with FlowFuse/flowfuse#8241 (green). The link check fails until it merges; Netlify's build does not run it, so the preview builds. Changelog and catalog entry are #5672, blog post #5671.Checklist