EphemeraAdmin: Import: Help:Ephemera Agent/Bulk Import

2026-04-01T04:29:32Z

Import: Help:Ephemera Agent/Bulk Import

New page

{{DISPLAYTITLE:Ephemera Agent — Bulk Import}}
__TOC__

The '''WIKI''' tab contains a '''BULK IMPORT''' section for loading many pages at once without making AI calls (except optionally for formatting). This is the primary way to populate the <code>Instructions:</code> namespace.

== JSON Import ==

=== Supported Formats ===

The importer accepts three JSON structures automatically:

; Flat array (simplest)
:<syntaxhighlight lang="json">
[
{ "title": "Instructions:Foo", "content": "raw wikitext here" },
{ "title": "Instructions:Bar", "content": "more wikitext", "summary": "Initial import" }
]
</syntaxhighlight>

; Instruction namespace schema
:<syntaxhighlight lang="json">
{
"pages": [
{
"title": "Instructions:Create/Source (Base Workflow)",
"meta": { "id": "create-source-base", "type": "workflow" },
"sections": {
"Summary": "Base workflow for creating a new source page.",
"Procedure": ["Step one", "Step two", "Step three"],
"Output": ["Sources page", "Talk page"]
}
}
]
}
</syntaxhighlight>

The <code>sections</code> object is automatically converted to wikitext:
* String values → paragraphs
* Arrays → bullet lists (<code>*</code>)
* Nested objects → definition lists (<code>;</code> / <code>:</code>)

No AI call is needed for this conversion.

=== Conflict Handling ===

The '''On Conflict''' dropdown controls what happens when a page already exists:

{| class="wikitable"
! Option !! Behaviour
|-
| Skip existing pages || The page is left unchanged and marked as skipped in the log
|-
| Overwrite existing pages || The page is replaced silently
|-
| Ask for each conflict || A browser dialog prompts you to confirm or skip each conflict individually
|}

=== AI Formatting ===

If a page has no <code>content</code> field and no <code>sections</code> object, it cannot be imported without formatting help. Enable '''Use AI to format wikitext''' to have the generator model convert the raw JSON structure to wikitext. This costs tokens and requires an API key to be loaded.

=== Running an Import ===

# Click '''📄 Choose JSON file''' and select your file
# A preview shows how many pages were found and flags any that need AI formatting
# Set your conflict preference
# Click '''IMPORT PAGES'''
# The progress bar and log show status for each page as it is created

A CSRF token is fetched once and reused for the whole batch. A 300ms pause between pages respects MediaWiki rate limits.

== XML Dump Import ==

=== Overview ===

Accepts standard MediaWiki XML export format — the same file produced by [[Special:Export]] or any MediaWiki dump tool. The XML is posted as a multipart file upload to <code>proxy.php</code>, which forwards it to the wiki's <code>action=import</code> API.

=== Required Permission ===

The <code>importupload</code> right is required on your bot account. This is typically restricted to sysops. To grant it:

Add to <code>LocalSettings.php</code>:
<syntaxhighlight lang="php">
$wgGroupPermissions['bot']['importupload'] = true;
</syntaxhighlight>

Then grant '''Import pages from file uploads''' at [[Special:BotPasswords]] when creating or editing your bot password.

=== Options ===

; Import full revision history
: Checked — all revisions from the XML are imported, preserving the full edit history. Unchecked — only the latest revision of each page is imported.

; Warning
: XML import can overwrite existing pages and histories depending on the dump. Test with a small file first.

=== Running an XML Import ===

# Click '''🗂 Choose XML dump''' and select your <code>.xml</code> file
# The tool sniffs the first 512 bytes to confirm it looks like a MediaWiki export
# The file size is displayed — confirm before proceeding for large dumps
# Set the revision history option
# Click '''IMPORT XML DUMP'''
# The log shows each imported page and its revision count

[[Category:Help]]
[[Category:EphemeraAgent]]

Help:Ephemera Agent/Bulk Import - Revision history

EphemeraAdmin: Import: Help:Ephemera Agent/Bulk Import