fix: infer image, cta, and rtf types in local-editor

[benlife5]

We infer the types of entity document fields in the local-editor, but RTF, Image, and CTA fields were just being picked up as generic objects. This means that they wouldn't show up in the correct dropdowns (particularly for createItemSource). This checks for their subfields and narrows the type if there's a match.

Also removes the c_ prefix when generating the display name.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 80699899-3199-4dc7-b079-fbef9f3764c4

📥 Commits

Reviewing files that changed from the base of the PR and between c1feb58 and 11a679c.

📒 Files selected for processing (3)

• packages/visual-editor/src/vite-plugin/local-editor/data.ts
• packages/visual-editor/src/vite-plugin/local-editor/entityFields.test.ts
• packages/visual-editor/src/vite-plugin/local-editor/entityFields.ts

🚧 Files skipped from review as they are similar to previous changes (2)

• packages/visual-editor/src/vite-plugin/local-editor/data.ts
• packages/visual-editor/src/vite-plugin/local-editor/entityFields.ts

---

Walkthrough

inferEntityFields now accepts either a single snapshot document or an array of snapshot documents, iterating over all entries and merging inferred field nodes. A new inferStructuredObjectType helper and shape-checking predicates detect rich text v2, image, and CTA object shapes, assigning specific typeRegistryId values instead of the generic "type.object". toDisplayName strips the "c_" prefix. getLocalEditorDocument now reads every snapshot JSON for the selected template and passes the full array to inferEntityFields. A new Vitest suite covers single-document structured-type inference, nested image detection, and multi-document schema merging.

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly matches the main change: inferring image, CTA, and RTF types in the local editor.
Description check	✅ Passed	The description is directly related to the changeset and mentions the same field type inference and display-name update.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch local-editor-types

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 3

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@packages/visual-editor/src/vite-plugin/local-editor/data.ts`:
- Around line 115-127: The template-sibling snapshot loading in
getLocalEditorDocument is too brittle because readJsonFile errors from
non-selected documents can abort the whole request. Update the templateDocuments
collection logic to handle parse failures per sibling snapshot, emit a
diagnostic for the malformed file, and continue building entityFields from only
the snapshots that were read successfully. Keep the selected document path
unchanged and use the existing getLocalEditorDocument, readJsonFile, and
inferEntityFields flow to locate the fix.

In `@packages/visual-editor/src/vite-plugin/local-editor/entityFields.ts`:
- Around line 218-223: The isRichTextValue predicate is too broad and is
reclassifying unrelated objects as rich text. Tighten the check in
isRichTextValue so it only returns true for the actual rich-text shape, such as
requiring a json object with the expected rich-text structure (for example
json.root) or a valid html string. Keep the change localized to isRichTextValue
in entityFields.ts so the type inference no longer promotes false positives to
type.rich_text_v2.
- Around line 277-279: The toDisplayName helper only strips the first “c_” from
the full path, so nested custom fields can still display incorrectly or lose
non-prefix matches. Update toDisplayName in entityFields.ts to process each path
segment separately, removing the “c_” prefix from every segment before
formatting the display name, and keep the rest of the path-to-label conversion
logic unchanged.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 9128723e-fe2d-43ed-96c2-5cf9f4821fb8

📥 Commits

Reviewing files that changed from the base of the PR and between d605f45 and c1feb58.

📒 Files selected for processing (3)

• packages/visual-editor/src/vite-plugin/local-editor/data.ts
• packages/visual-editor/src/vite-plugin/local-editor/entityFields.test.ts
• packages/visual-editor/src/vite-plugin/local-editor/entityFields.ts

[mkilpatrick]

What happens when this returns false? Anything detrimental?

Would it be better to do linkType && (label || link)? Or even JUST linkType?

[benlife5]

ooh yeah linkType && (label || link) is probably good. If it doesn't match, then it doesn't show up in type.cta dropdowns... We might eventually need to pull the full schema so that the local-editor is more functional with all data types, but this will get us the main types for now