[andy-stark-redis]

Summary

Adds the groundwork for replacing the image shortcode with native Markdown images. Unit 0 of DOC-7128, continuing the DOC-6909 render-hook migration series. This PR converts no content. The rollout units run the codemod.

Authors will write:

![Alt text](/images/x.png)
{width="300px" class="inline"}

Files (4):

• layouts/_default/_markup/render-image.html (new): renders Markdown images with the width/class the shortcode accepted. It drops the shortcode's <a sdata-lightbox> wrapper, because nothing reads sdata-lightbox and static/js/index.js already makes any unwrapped <img> click-to-open (it skips #no-click images, as before). It warns at build time when an image isn't in static/.
• config.toml: turns on markup.goldmark.parser.attribute.block and turns off wrapStandAloneImageWithinParagraph. Hugo needs both to pass an attribute line to the image hook, and only for an image that is its own paragraph.
• build/migrate_image_shortcodes.py (new): the codemod. It converts standalone image shortcodes and rewrites every path to site-root /images/.... Anything it can't convert safely stays a shortcode and is reported.
• layouts/partials/process-markdown-content.html: keeps the AI Markdown/JSON feed correct for converted images by using absolute URLs and dropping the attribute line.

Rendered changes from this PR alone: 7 pages with existing native standalone images lose the <p> around those images. That matches how shortcode images already render (they were never in a <p>). 3 pages with page-relative native images now get absolute paths to the same files.

Not convertible, left as shortcodes: Markdown can't attach width/class to an image inside a sentence, so the ~123 inline button icons (#no-click, class="inline") stay as shortcodes pending a decision on the ticket.

Verification

• Full-site Hugo builds of origin/main and this branch: both succeeded, 19,661 pages each. The two parser settings capture no existing {...} line anywhere in the corpus.
• Trial conversion of the whole corpus (not included here), built under the production baseURL (/docs/latest/) and compared image by image (src/width/alt/class) across every page. 1,646 of 1,829 shortcodes convert. The only rendered image changes are:
  • 40 alt texts lose Markdown emphasis (**Delete** becomes Delete), which reads better for screen readers.
  • 34 page-relative native images resolve to their permalinks.
  • The one dot-relative image, develop/tools/insight/insight-stream-consumer.md, is fixed. It currently renders /docs/images/... in production, which is broken.
• The codemod is idempotent: a second run changes nothing. Every changed file has exactly one deleted line per converted shortcode.
• AI feed: after the trial conversion, no attribute lines and no relative /images/ links remain in any .md output.
• Not checked: the spacing of the 7 native-image pages in a browser.

🤖 Generated with Claude Code

---

Note

Medium Risk
Site-wide Goldmark parser and a new image render hook change how every Markdown image is parsed and emitted (including a few existing native standalone images losing a wrapping <p>); AI feed regexes must stay aligned with author syntax.

Overview
Unit 0 groundwork to replace the {{< image >}} shortcode with native Markdown plus a Hugo image render hook. No content is converted in this PR; later rollout runs build/migrate_image_shortcodes.py.

Authors will use a standalone image line plus an optional attribute line ({width="..." class="..."}), which config.toml enables via wrapStandAloneImageWithinParagraph = false and markup.goldmark.parser.attribute.block = true.

layouts/_default/_markup/render-image.html is new: it maps /images/... and /static/images/... through relURL, resolves page resources for relative paths, applies width/class from block attributes, omits the old shortcode lightbox <a> wrapper (click-to-open stays in site JS), and warnfs when a root path is missing under static/.

build/migrate_image_shortcodes.py is a conservative, idempotent codemod that only rewrites shortcodes on their own paragraph, normalizes paths to site-root form when the file exists in static/, and reports skips (inline/table, bad attrs, missing files, etc.).

process-markdown-content.html strips attribute lines and rewrites /images/ links to absolute https://redis.io/docs/latest/images/ so AI Markdown/JSON output stays clean after migration.

^{Reviewed by Cursor Bugbot for commit f7bab35. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

DOC-7128

[github-actions]

🧠 Redis Memory

Found 5 related items from repository history (5 new this commit):

• PR — DOC-7059: Migrate develop/reference+setup to render hooks 🆕

  Summary

• PR — DOC-7089: Migrate content/operate/oss_and_stack/install/ to render hooks 🆕

  Summary

• PR — DOC-7059: Migrate develop/ai/context-engine+featureform+agent-builder to render hooks 🆕

  Summary

• PR — DOC-7080: Migrate content/commands/ probabilistic data structure and time series commands to render hooks 🆕

  Summary

• PR — DOC-7059: Migrate develop/ai/search-and-query to render hooks 🆕

  Summary

Memory updated at f7bab35

[andy-stark-redis]

Thanks @dwdougherty !