# New summary ## Description The Load Document widget rejects `.md` / `.markdown` uploads because `md` and `markdown` are never added to the standard Document media field (`field.field.media.document.field_media_document`). `ai_context_document_loader` already has an idempotent helper, `ai_context_document_loader_extend_document_media_extensions()`, called from `hook_install()`. That helper no-ops when the document field does not exist yet. A later field-config import can also overwrite the extensions after they were added. This is **not** caused by config import or recipes skipping `hook_install()`. Drupal still runs `hook_install()` when a module is enabled that way. The same gap can happen on a normal install if the document field is created after the module is enabled. The extraction pipeline itself works (`drush document-loader:load` on a `.md` file succeeds). The block is media field extension validation. ## Steps to reproduce 1. Go to `/admin/config/ai/context/items/add`. 2. In the Content field (MDX editor), click **Load Document**. 3. Select the **File** category → **Add media**. 4. Upload a `.md` file (for example `sample.md`). ## Expected behavior The `.md` file is accepted. Its text is extracted and copied into the Content editor, same as `.txt`. ## Actual behavior The upload is rejected before Document Loader runs: > Only files with the following extensions are allowed: txt rtf doc docx > ppt pptx xls xlsx pdf odf odg odp ods odt fodt fods fodp fodg key > numbers pages. Confirmed in field config and `$media->validate()`. ## Proposed resolution - Keep the existing idempotent helper and the `hook_install()` call. - Add an update hook that calls the helper so already-enabled sites get the extensions via `drush updb`. - Also run the helper when the document field is created after module install (FieldConfig insert or config-save for that field). A one-shot `hook_update_N` is not enough: new installs skip it because schema is set to the highest update number. - Kernel test for the real race: enable the module with no document field, create the field without `md`/`markdown`, assert the extensions are added, and assert a second run is idempotent. - Update `docs/features/importers.md`: install can miss the field; existing sites need `drush updb`; sites using config management must export the field or the next `drush cim` can remove the extensions. --- # Original info **Original title** Uploading a .md/.markdown file via "Load Document" is rejected — extensions were never registered on the "document" media type **Original summary** ## **Description:** The `ai_context_document_loader` submodule ships a hook (`ai_context_document_loader_extend_document_media_extensions()`, invoked in `hook_install()`) whose purpose is to add the `md` and `markdown` extensions to the allowed-extensions list on the "document" media type's file field (`field.field.media.document.field_media_document`), since the module itself exists to let users import Markdown into the Context Item editor. In this install, that hook apparently never ran (common when a site is assembled via recipe/config-import rather than the standard "install module" flow), and the allowed-extensions list is still only `txt rtf doc docx ppt pptx xls xlsx pdf odf odg odp ods odt fodt fods fodp fodg key numbers pages` — no `md`/`markdown`. This blocks exactly the scenario the docs describe: "Import a small markdown or txt file." ## **Steps to reproduce:** 1. Go to `/admin/config/ai/context/items/add`. 2. In the Content field (MDX editor), click **Load Document**. 3. Select the **File** category → **Add media**. 4. Try to upload a `.md` file (e.g. `sample.md`). **Expected behavior:** The `.md` file should be accepted, its text extracted and copied into the Content editor, ready for editing before saving as draft — same as with `.txt`. **Actual behavior:** The `.md` file upload is rejected by the media field's extension validation before the Document Loader ever gets to process the content. Confirmed at two levels: * Field config: allowed extensions do not include `md`/`markdown`. * Entity validation (`$media->validate()`) on a `document`-bundle Media entity pointing at a `.md` file: returns a violation — _"Only files with the following extensions are allowed: txt rtf doc docx ppt pptx xls xlsx pdf odf odg odp ods odt fodt fods fodp fodg key numbers pages."_ {width=900 height=545} Important: this is not a parser/loader failure — I tested the extraction pipeline directly (`drush document-loader:load`) with a real `.md` file and it extracted the content perfectly, with no PHP error. The block is purely at the media-type upload/extension-validation step. **Evidence (CLI):** ``` $ drush cget field.field.media.document.field_media_document --format=yaml | grep extensionfile_extensions: 'txt rtf doc docx ppt pptx xls xlsx pdf odf odg odp ods odt fodt fods fodp fodg key numbers pages' ``` ``` $ drush eval "... $media->validate() ..."Violations: 1 - field_media_document.0: Only files with the following extensions are allowed: txt rtf doc docx ppt pptx xls xlsx pdf odf odg odp ods odt fodt fods fodp fodg key numbers pages. ``` ``` $ drush document-loader:load --input file_input=public://zoocha_sample.md --output-format=markdown# Zoocha QA Markdown SampleThis is a **markdown** file for Document Loader test case 10.- item one- item two(extraction succeeded, no error) ```