robert

Published Version 2.4 with support for custom user role permissions: promptai allows to use the prompts in page edit promptai-config allows to add/edit new prompts

Haha, yeah, those $50 in Opus credits pushed me quite a bit further too while writing this update … It’s really fascinating how fast you can move with a single prompt sometimes – only to suddenly going in circles for 30 minutes for a simple bug. Wild times 😅

What a coincidence! I’d just released a big update to my PromptAI module when I saw your post 🤯 As far as I can tell from your examples, our use cases differ slightly. Looks really good!

PromptAI v2.3 — Inline Mode, Streaming, Placeholders, and more Hi everyone, It's been a while since the last update for PromptAI, and quite a lot has changed. Here's a summary of what's new since the last public release. Inline Mode The biggest addition is a new Inline Mode that adds magic wand buttons directly next to your fields. Instead of the "Save + Send to AI" workflow, you can now trigger AI processing on individual fields without saving the page first. The response streams in live and replaces the field content — but nothing is saved until you hit Save. Both modes (Page Mode and Inline Mode) can be mixed freely in the same configuration. Each prompt is configured as either "page" or "inline". Inline mode works with all supported field types: text fields, TinyMCE/CKEditor fields, image descriptions, file descriptions, and custom subfields — including inside repeaters. Streaming Responses AI responses now stream in real-time via Server-Sent Events (SSE). You can watch the text appear token by token instead of waiting for the full response. This obviously only works in inline mode. Placeholder Support Prompts can now reference other field values from the page using `{page.fieldname}` syntax. Inside repeaters, use `{item.fieldname}` to access the current repeater item's fields. Examples: - Summarize the following text to 400 characters: {page.body} - Create an SEO meta description for a page titled '{page.title}' - Write a caption for this gallery item titled '{item.title}' from the {page.gallery_name} collection This makes prompts context-aware without having to send field content manually. Multi-Field Selection Instead of configuring a single source field per prompt, you can now select multiple fields. The same prompt will be applied to each selected field. This replaces the old source/target field concept — the "target" is now only used for subfield options in file/image fields. AI Tools (experimental) The module now ships with three experimental tools that allow the AI to query your ProcessWire installation: - getPages — find pages by selector - getPage — get detailed info about a single page - getFields — list available templates and their fields These are disabled by default and must be enabled in module configuration. They're meant as starting points — their usefulness depends heavily on your use case and I recommend to create your own for each projects specific demand. Other Changes - PHP 8.3 is now required to use this module Important: Backup Before Updating The internal configuration structure has changed significantly. The old `sourceField`/`targetField` concept has been replaced with a multi-field `fields` array and a separate `targetSubfield` option. There is no automatic migration for this change. I'm sorry for the inconvenience — the old structure simply didn't support the new features. I attached two short screen videos showing the prompt config screen with example prompts and one page edit screen where I showcase some of the predefined prompts. As always, feedback and bug reports are welcome — either here or on GitHub. promptai-showcase.mp4 promptai-config.mp4

@ngrmm I think I found the bug (again), could you try and change TranslateGlossary.php:34 to the following: $entryArr = self::convertGlossaryStringToArray($language->translate_glossary ?? ''); The change is the ?? '' after translate_glossary. This handles empty glossary fields that return null instead of an empty string. Hopefully this will do it!

@ngrmmHmm, that might be possible, but unfortunately I can't replicate the issue here. Could you double-check that TranslateGlossary.php:118 has a default value for the $glossaryString parameter in the source code? The line should look like this: private static function convertGlossaryStringToArray(string $glossaryString = ''): array { Also, a few questions to help investigate: – Do you have any entries in your language-specific translate_glossary field? – Are you using the DeepL API Free or Pro plan? Thanks for helping with the investigation! 😅

@ngrmm I think I found the bug, thanks for reporting! Version 1.3 should work now and can be found in the ProcessWire modules directory. Since you mentioned your Fluency version in your comment: it's no longer required for this module to run (but it's still a great module nevertheless 😄)

@update AG Thanks for reporting. I pushed a new version 1.3 to Github which supports PHP 8.1. The DeepL API client uses Symfony HTTP Client as a dependency, which requires PHP 8.2 in version 7.x. I downgraded it to Symfony 6.x, so now PHP 8.1 is the minimum supported version. Regarding ProcessWire compatibility with PHP 8.2: I checked client websites and found one installation running PW 3.0.230 without any errors/warnings, and another with 3.0.213 that has some deprecation warnings but works fine. Edit: I forgot to mention that Fluency is no longer necessary for ProcessTranslatePage to run, both modules work independently.

Hi Ivan, I haven’t had time yet to verify what (if anything) needs to be done on my side, but I’m fairly sure ProcessTranslatePage already uses the DeepL API v2. It relies on the official PHP client library and was updated in mid‑2025, so we should be in good shape. If any updates are required to keep the module working, I’ll take care of them – thanks for the heads up!

Funny, I just implemented almost the same feature as a Hanna Code for a client 😄 <script type="module" src="https://unpkg.com/@le-pepe/snow-effect"></script> <snow-effect flakes="500" color="#e5f5ff"></snow-effect> It seems it’s Christmas time again! 🎄

I like this idea – thanks! I’ll explore this further and see if I can come up with a clean implementation when I have time.

Hi! I’ve just released v1.7 with native DeepSeek support as an AI provider. Note that, due to limitations in the Neuron AI framework’s DeepSeek integration, file and image uploads aren’t supported at the moment. Regarding SEO Maestro: I haven’t worked with it before, and it looks like it uses custom subfields which aren't compatible with my current config form. It will take a bit more engineering to support properly. I’ll take a closer look about how to add subfield support in general, but I don’t expect to ship this in the near future.

Hi @Jorge, I think I found the bug causing your error, I mixed up the native _() and the Processwire __() function 😅 Fixed version is on Github, does this resolve your issue? https://github.com/robertweiss/PromptAI

Hi @Jorge, I‘m afk till friday but will have a look when I‘m back!

v1.1 has been released with some bug fixes and a ListerPro Page Action added to batch translate pages using the ListerPro module. Keep in mind that the Page Action module needs to be installed separately.