Knowledge Article Translation

Lingora supports translating Salesforce Knowledge Articles (__kav records) alongside traditional metadata translation. Knowledge Articles are rich content — HTML bodies, summaries, titles — that require a different workflow from string-by-string metadata.

Overview

Knowledge Article translation differs from metadata translation in several ways:

• Document-level — you're translating full articles with rich text HTML, not isolated strings
• Per-record-type fields — different record types may use different fields, so you select which fields to translate per record type
• REST API — articles are queried and written back via the Salesforce REST API (not the Metadata API used for XLIFF/STF)
• HTML-aware engines — translation engines receive raw HTML and preserve structure natively (DeepL uses tag_handling: "html", Google uses format: "html", Lara uses contentType: "text/html")

Requirements

• Studio or Partner plan — Knowledge Article translation requires an active Salesforce connection
• Salesforce Knowledge enabled — your org must have Salesforce Knowledge configured with at least one article type
• API permissions — the connected user needs API access and Knowledge article permissions

Creating a Knowledge Article Project

From the New Project Page

1. Navigate to Projects → New Project
2. Select Knowledge Article Translation
3. Choose your entry method:
   • Live Connection — query articles directly from your connected Salesforce org via REST API
   • Upload Export — upload a .zip file exported from Salesforce Translation Workbench

Live Connection Flow

1. Select an Org — if you have multiple connected orgs, choose which one to query
2. Select Article Type — choose the __kav object type (e.g., Knowledge__kav, FAQ__kav)
3. Select Fields per Record Type — a matrix shows all translatable fields (Title, Summary, custom RTA fields) by record type. Check which fields to translate for each record type
4. Configure Project — set the project name, source/target languages, and article publish statuses to include (Published, Draft, or both)
5. Create — Lingora queries all matching articles and creates translation units for each article × field × target language

File Upload Flow

1. Upload ZIP — drag and drop the .zip file exported from Salesforce Translation Workbench
2. The parser reads export.properties (source/target language, article type) and the CSV manifest with inline fields and HTML file references
3. Review and Confirm — verify the parsed articles and field counts, set the project name
4. Create — translation units are created from the parsed content

The Document Workspace

The Knowledge Article workspace uses a side-by-side layout designed for document-level content:

Layout

• Left sidebar — article list with search and status filters (All, Pending, Translated, Review, Approved, Rejected)
• Field tabs — one tab per translatable field (Title, Summary, Article_Body__c, etc.) with status dots
• Side-by-side editor — source content on the left, translation on the right

Rich Text Fields (RTA)

For HTML content fields like Article Body:

• Source pane — rendered HTML in a sandboxed iframe (images, formatting, links all display correctly)
• Target pane — shows translation candidates or an editing area
• Candidates — when translation completes, candidates appear with rendered HTML previews. The recommended candidate auto-expands
• HTML editing — you can write custom translations as raw HTML with a preview toggle

Inline Fields

For short text fields like Title and Summary:

• Simple two-column layout with source text and target input
• Character limit indicator when applicable

Keyboard Shortcuts

Translating Articles

Single Field

Click the Translate button next to any field's "Translation" header, or press t. This queues a translation job for just that field.

Single Article

Click Translate Article (N) in the header bar, or press T. This translates all pending fields in the currently selected article.

Bulk Translation

Click Translate All (N) to translate all pending units across all articles in the project.

Engine Routing

Knowledge Articles use the KnowledgeArticle resource type for engine routing. You can configure a specific engine for articles in Settings → Engine Routing Rules — for example, route Knowledge Articles to DeepL while using the Lingora LLM for metadata.

Reviewing and Approving

• 3 candidates — when using the Lingora LLM, articles always generate 3 candidates (Literal, Idiomatic, Formal) for review
• BYOK engines (DeepL, Google, Lara) generate a single candidate that can be directly approved
• Approve a candidate by clicking the Approve button or pressing a for the recommended one
• Reject a field to flag it for re-translation
• Custom text — write your own translation (supports raw HTML editing for RTA fields)

Write-Back to Salesforce

Once articles are approved, click Write Back to push translations to Salesforce:

1. Lingora submits the article for translation in Salesforce (creates an empty draft in the target language)
2. Queries for the newly created draft version ID
3. Updates the draft with your approved translations via the sObject API

Important: Lingora creates draft translations only. You publish the articles from within Salesforce when ready.

Job Monitoring

The workspace shows a job monitor bar below the header:

• Real-time progress bars with unit counts and ETA
• Separate tracking for "Translation" and "Write Back" jobs
• Clickable error counts open a dialog with full error details
• Jobs can be dismissed individually or cleared all at once

Knowledge-Specific Languages

Knowledge Articles have their own language configuration in Salesforce, separate from metadata translation languages. The article project wizard queries KnowledgeArticleVersion for available languages, so you'll see the correct language options even if they differ from your Translation Workbench setup.