Recently Updated Topics

I think today represents an important milestone in running a good model locally. See what Antirez (creator of Redis) has released. Tweet embedded below. Does anyone have an Apple M3 with 128gb RAM? I'd be interested to hear how it goes if you run it.

ProcessWire is more alive than ever.

Fix is confirmed. Installed the latest version from Github and it's working as before. 🎉 GLM is pretty nice. Feel free to test it a bit more using that key.

Working with AI in PageGrid Inspired by projects like AgentTools, I began investigating how well an AI could handle PageGrid’s native PW structure (Pages, Templates, and Fields). The results were surprisingly good and with a few targeted optimizations, the workflow has become remarkably solid. With the latest updates, an AI agent is reliably able to create and design PageGrid layouts, create new block templates, or perform content updates. To teach AI how to speak "PageGrid, I created a small AGENTS.md file that acts as a central hub. I’ve then added specific "skills" in separate .md files for various tasks ensuring the AI only loads the documentation it actually needs. This approach is highly optimized to minimize token consumption, allowing even "smaller" models like GPT-4o mini or Claude Haiku to produce error-free migrations. To make this possible, I also introduced dedicated Migration Functions that allow the AI to programmatically add items or set styles. For CLI-based projects, I highly recommend the AgentTools Module to streamline the integration. Beyond CLI support, AgentTools also provides a native AI interface directly within the ProcessWire backend editor. Getting Started Install the lastest version of the FieldtypePageGrid module (try for free). Tell your AI agent to read the PageGrid agent guide first: That file gives the agent everything it needs to understand PageGrid and routes it to the right documentation for your task. What You Can Ask the AI to Do Build or modify a layout Create pages with blocks, apply styles, set up responsive layouts using CSS grid columns. Example prompts: "Create a landing page with a full-width hero section and a 3-column feature grid below it." "Add a text block and an image block side by side inside the first group on my homepage." "Make the hero block have a dark background and white text, with 60px padding on desktop and 24px on mobile." The core advantage here is that the AI doesn't write your frontend code from scratch. The HTML and logic are already defined in your PageGrid Block Templates. The AI simply acts as an orchestrator, assembling these blocks and applying styles. This ensures the output remains clean, semantic, and easy to maintain via drag-and-drop later on. Create a custom block template Define a new block type with custom fields and register it with a PageGrid field. Example prompts: "Create a custom block called pg_testimonial with a quote text field and an author name field." "Create a custom card block with a title, description, and link field, and add it to my homepage PageGrid field." Write a site template Render a PageGrid field inside your own PHP template file. Example prompts: "Show me how to render my PageGrid field inside my home.php template using markup regions." "Generate a site template for pagegrid-page that includes the PageGrid output between the header and footer." Docs Documentaion

I've just updated the AdminNeo core and included two new plugins: GeminiSqlPlugin and OpenWebUiPlugin for inline prompt based SQL command generation. I haven't tested the OpenWebUiPlugin one, so if anyone has any problems with it, please let me know.

Thank you. This could have saved me a lot of time, I appreciate your input. I will try this out.

Hello @rushy, I think it only matters if you want the correct created and modified timestamp when creating or editing pages. I don't think there are any further implications. You could try this script in your config.php. It worked for me:

Hi @maximus, I've been trying it out this module too and really like it! I have a couple of suggestions, if that's OK? Pages that don't have template view files (<templatename>.php) still show a view icon in the collection list, resulting in a 404 when clicked. This could be solved by adding a $page->viewable() check to the $showViewLink boolean. When a template's "Can this template be used for new pages?" setting ("noParents") is set to -1 (only one page) or 1 (no new pages), the "Add [collectionItem]" button still shows. Maybe a canAddNew() method of the Collection class to hide the button in those circumstances? [this worked for me, though there may be a more elegant way, or that suits your coding style]: /** * Whether the admin "Add" button should be shown for this collection. * Mirrors ProcessPageAdd's logic on Template::noParents: * 1 → template disallows new pages * -1 → singleton; only allowed while no page using it exists yet */ public function canAddNew(): bool { $template = wire('templates')->get($this->template); if (!$template || !$template->id) return false; $noParents = (int) $template->noParents; if ($noParents === 1) return false; if ($noParents === -1) { return $template->getNumPages() === 0; } return true; } //and then in collection-list.php: <?php if ($canCreate && $collection->canAddNew()): $addTemplate = $wire->templates->get($collection->template); $addUrl = $adminUrl . 'page/add/?template_id=' . $addTemplate->id; ?> Anyway, thanks for the great module!

Thx for the kind words 😊 – and thx for the catch, I really wonder why I never noticed this before, it's kinda obvious 😅 Should be fixed in new version 1.6.1

I did not know that, never clicked a link in the title column, always on an upgrade link, if available...

I do not strive for balance. I think that's just not a reality to achieve that. Maybe if you join some monks, then you can find it, but you'll need to give up everything else then. I focus on goals, and when I achieve one, well, that is satisfying. That might sound simple, but it's not. Of course, it's hard to stay focused, and it is also hard to keep the priorities. But I try not to miss the most important: family, health/sport, work, friends (in that order, and that's the order of importance, and not the order I spend time with/on, which is a different story..., but generally, I am just trying to be flexible with time as much as I can.) And when I am feeling blue, I just need to remind myself: https://youtu.be/uAzwji5Cfmk?list=RDuAzwji5Cfmk&t=24

Hello, everyone. I was wondering if it wouldn't just be better to switch TinyMCE with SunEditor instead as the "official" WYSIWYG editor of ProcessWire. It's licensed under MIT and it outputs cleaner code. I'm not a developer, so I'm afraid I cannot create a module and implement it myself. I don't need it personally, but I'm sure there are people (clients) who do need a WYSIWYG editor.

Thank you for your advises and corrections. I think we're done with this issue.

I not owned module. I think not good idea storage 3 files 🙂

@AndZyk Uiii, very nice. Was just wondering if it could get official that the core files are inside the vendor directory.

@ryan the AgentTools module looks really promising. Really interested to see how this develops. I wanted to share that I tried it out today with a fresh Claude API token and the response was quite strange and weirdly formatted. (See screenshots). This was in a local dev environment using MAMP/localhost

Hi everyone, Built this for a financial news site that needed live stock quotes embedded inline in editorial content. Drop a ticker into any text field and it renders as a live badge with price, change, and direction - clicking opens a full detail popup. GitHub: https://github.com/mxmsmnv/Stocks What it does: Live price badges - ticker, price, change amount/%, direction arrow; click opens a popup with open/high/low, 52-week range, P/E, volume, market state Three data providers - Yahoo Finance (free, no key needed), Finnhub, Alpha Vantage TextFormatter with three parse modes: Explicit [stock:AAPL] tags Cashtag/hashtag - $AAPL, #TSLA Auto-detection by company name and aliases in text Company Manager - tracked companies with names, aliases, enable/disable toggles, bulk CSV import Four CSS frameworks - Vanilla CSS (built-in), Tailwind, Bootstrap 5, UIkit 3; auto-detected or manually set File-based cache with configurable TTL and per-ticker clear from admin Circuit breaker - pauses API calls after repeated failures, serves stale cache with ~ marker Custom provider API - add any data source by extending StocksProviderBase $stocks = $modules->get('Stocks'); echo $stocks->renderBadge('AAPL'); echo $stocks->renderBadgeAs('TSLA', 'bootstrap'); Requirements: ProcessWire 3.0+, PHP 8.2+ MIT License.

https://demo.processwire.com/

Hey @ryan, is this a side effect of your recent restructuring of the /wire/core/ classes?

Hello @Peter Knight To be clear, FrontendForms doesn't send any form data by default – it just checks whether a form is valid or not. Inside the isValid() method, you can send form data — or do whatever you want (e.g., store the values on a page). So it's not a "Email Form" module by default. Informing about what went wrong So in this case, it's a bit more complicated. The validator can only set the form validation to true or false (without sending information to an administrator). Of course, it could happen that a real person uses some characteristics of a SPAM text and the text is therefore declared as SPAM. This risk always exists, but it is very low. For this reason, it is necessary to inform the user about which mechanisms were responsible for the failure of validation. But I agree with you: too much detailed information (e.g., word 1 and word 2 are SPAM words) could give a human spammer too much information. My Opinion: In this case you have to make a compromise. You can't stop SPAMMERS completely. If it's a bot, it should post a fixed text in a text box – it will fail and can't react the way it changes the SPAM text it's going to pass. When you have real human SPAMMERS, it becomes much harder to stop them, because they can react much different than a bot. At the moment, the information in the error message is very general and doesn't contain too much detailed information. Yes the danger is always there, especially with AI - so I guess I leave it general. You can also change the error message to your own by asking the user to call you by phone if the message fails validation. Another option would be to set the threshold lower (e.g. from 50 to 30) so that more SPAM properties are accepted and the entered text is more likely to pass validation. I don't know what the "best practice" scenario is in this case. Maybe someone has an idea how best to deal with messages that have been falsely declared as SPAM, even though they are not.

Hi @nbcommunication - I've actually set up my own report endpoint which logs to PW log files (csp-report and csp-report-only). I also ping via slack for any enforced violations. The only thing I still have in report-only mode now is "require-trusted-types-for". I have actually reported to a few open source projects I integrate about the changes to support "require-trusted-types-for" and they've actually been very responsive so I am hoping to be able to enforce this option very soon also. Note that I do exclude violations from browser plugins and the like, so my csp-report log is actually almost completely silent now. I do have separate policies for frontend vs PW admin which helps because PW still has various inline scripts and no nonce (yet). I am using Cloudfront but all my CSP and nonce logic is handled locally in my /site/init.php file - it's very different and much stricter than what I posted earlier above. I have Cloudfront enabled caching for all assets, but I am not using Procache for a variety of reasons, but performance with Cloudfront asset caching is a pretty good middle-ground.

Context v1.5.0 — AI Gateway Hi, just pushed a new release. The main addition is a centralized AI Gateway — a ContextAI class that any ProcessWire module can use to make AI requests through a single shared configuration: $ai = wire('context')->ai(); // Simple $text = $ai->complete('Summarize: ' . $page->title); // Full options $result = $ai->chat([ 'messages' => [['role' => 'user', 'content' => '...']], 'model' => 'openai/gpt-4o-mini', 'caller' => 'MyModule', ]); // Gateway entry point for third-party modules $result = $ai->gateway(['caller' => 'MyModule', 'messages' => [...]]); Supported providers: OpenRouter (200+ models via one key), OpenAI, or any OpenAI-compatible endpoint. New settings fieldset: provider, API key, default model, temperature, max tokens, timeout, global system prompt, and a Test Connection button that fires a live request and shows the response time. The idea is that modules don't need to implement their own HTTP clients or manage API keys — they just call wire('context')->ai() and everything is configured in one place. GitHub: https://github.com/mxmsmnv/Context

Hey NEO in this instance is more aligned to “new” than any acronym or AI /LLM technology. I guess the upper case letters suggests an acronym but, honestly, I just liked the visual rhythms of both words. Not sure if this answers your question or if I misunderstood?

I’m using the core FieldtypeComments in ProcessWire and everything works correctly on the frontend and in the database. Comments are saved properly and displayed without issues. However, in the admin I get a one-time SQL error after a new comment is submitted: SQLSTATE[HY000]: General error: 1525 Incorrect TIMESTAMP value: '1777649215' Observations: Difference between “error value” and stored value is consistently +3600 seconds No data corruption or functional issues occur Environment: PHP timezone and $config->timezone both set to Europe/Brussels The warning only appears once per comment submission and does not affect saving or rendering. Has anyone seen this behavior before with FieldtypeComments?

Yes, I know it well — great Dashboard module! I've used it on almost every project to keep the admin from feeling bare. But as projects grew, I kept hitting the same wall: customizing Dashboard meant writing PHP for each install, and the widget lists got unwieldy. Start (and its companion module Collections) came out of that frustration — the goal was a fully visual, no-code editor where you just drag, drop, and pick icons without touching any config files.