klbr/n8nworkflows.xyz

Fork 0

mirror of https://github.com/khoaliber/n8nworkflows.xyz.git synced 2026-04-28 00:29:22 +00:00

Files

T

nusquama ed2d8e9b46 creation

2026-02-22 12:02:26 +08:00

27 KiB

Raw Blame History

Translate WordPress posts and ACF fields using DeepL and OpenAI

https://n8nworkflows.xyz/workflows/translate-wordpress-posts-and-acf-fields-using-deepl-and-openai-13543

Translate WordPress posts and ACF fields using DeepL and OpenAI

1. Workflow Overview

Purpose:
This workflow receives a WordPress post event (via webhook), fetches the full post data from WordPress, detects the source language using OpenAI, determines which target translations are missing, optionally corrects the original post’s language flag in WordPress, then translates Gutenberg content + selected ACF text fields using DeepL, reconstructs the translated post payload while preserving ACF structure, and finally creates translated draft posts in WordPress linked to the original.

Typical use cases:

Multilingual WordPress sites (e.g., Polylang/WPML-like setups) where posts must be replicated per language.
Sites using Gutenberg content and Advanced Custom Fields (ACF), including nested structures (repeaters/flexible content).
Avoiding duplicate translations by checking existing translations metadata from WordPress.

1.1 Trigger & Fetch Raw WP Post Data

Receives a POST webhook containing a post identifier/type, then fetches the canonical WP REST API post object.

1.2 AI Language Detection & Smart Routing

Builds a short text snippet, detects language via OpenAI, decides whether the WP language flag must be fixed, and computes which target languages still need translations.

1.3 Flatten ACF Structure & Translate with DeepL

Extracts translatable strings from post title/content + ACF (based on configured keys), splits execution by each target language, and sends a batch translation request to DeepL (HTML-aware).

1.4 Rebuild ACF Object & Push to WordPress

Reapplies translated strings into the correct title/content/ACF paths, cleans the ACF JSON, and creates translated WordPress posts as drafts linked to the original.

2. Block-by-Block Analysis

Block 1.1 — Trigger & Fetch Raw WP Post Data

Overview:
This block starts the workflow via webhook and retrieves the complete post JSON from the WordPress REST API using Basic Auth.

Nodes Involved:

Webhook
Get Post Data

Node: Webhook

Type / Role: n8n-nodes-base.webhook — Entry point (HTTP trigger).
Config (interpreted):
- Method: POST
- Path: translate-post (endpoint becomes .../webhook/translate-post depending on n8n setup)
Inputs/Outputs:
- Output → Get Post Data
Key data expected in input:
- Uses expressions later like: $json.body.post.ID and $json.body.post.post_type
Edge cases / failures:
- Missing/invalid payload structure (no body.post.ID or body.post.post_type) will break URL construction downstream.
- If webhook is publicly accessible, consider authentication/secret header validation (not implemented here).
Version notes: Node typeVersion 1.

Node: Get Post Data

Type / Role: n8n-nodes-base.httpRequest — Fetch post from WP REST API.
Config (interpreted):
- URL is dynamic:
  - Base: https://your-wordpress-site.com/wp-json/wp/v2/
  - Endpoint chosen by post_type:
    - if post → posts
    - if solution → solutions
    - else → uses the raw post_type directly
  - Then appends /{ID}
- Authentication: HTTP Basic Auth via n8n “generic credential type”
Key expressions:
- {{$json.body.post.post_type}}, {{$json.body.post.ID}}
Inputs/Outputs:
- Input ← Webhook
- Output → Prepare Text Snippet
Edge cases / failures:
- 401/403 if Application Password / user lacks permissions.
- 404 if post type endpoint does not exist or ID invalid.
- If WP returns a different schema (missing content.rendered, _links.self[0].href, etc.), later nodes may fail.
Sticky notes applying:
- “## 1. Trigger & Fetch Raw WP Post Data”
- “### URL and WP APP Password\nUpdate the URL to your WP domain and select your WP App Password credentials.”
Version notes: typeVersion 3.

Block 1.2 — AI Language Detection & Smart Routing

Overview:
Creates a short plain-text snippet, detects its language using OpenAI, determines required target translations, and optionally corrects the original post’s language flag in WordPress. A merge node reunifies the “fix language” and “no fix” paths.

Nodes Involved:

Prepare Text Snippet
OpenAI Language Detect
Smart Router & Targets
Needs WP Lang Fix?
Fix WP Language Flag
Merge
Needs Translations?

Node: Prepare Text Snippet

Type / Role: n8n-nodes-base.function — Build a short sample text for detection.
Config (interpreted):
- Creates textSnippet from:
  - post.title.rendered
  - plus stripped HTML from post.content.rendered if present
  - else falls back to post.acf.intro if present
- Truncates to 1000 chars to reduce OpenAI token usage.
Output:
- Passes through the full post JSON plus textSnippet.
Inputs/Outputs:
- Input ← Get Post Data
- Output → OpenAI Language Detect
Edge cases / failures:
- If title.rendered missing, may throw.
- If content is extremely large, still safe due to substring but HTML stripping can be expensive (usually fine).
Version notes: typeVersion 1.

Node: OpenAI Language Detect

Type / Role: n8n-nodes-base.openAi — Chat completion for language detection.
Config (interpreted):
- Resource: chat
- Model: gpt-4o-mini
- System message instructs: output ONLY 2-letter ISO 639-1 code.
- Note: The node as shown includes only a system message; in a working setup you typically also include a user message containing {{$json.textSnippet}}. (If not configured in the UI, detection will not work.)
Inputs/Outputs:
- Input ← Prepare Text Snippet
- Output → Smart Router & Targets
Edge cases / failures:
- OpenAI credential/auth errors, rate limits, transient 5xx.
- Model may output unexpected whitespace/newline; downstream .trim().toLowerCase() mitigates.
- If prompt does not include the snippet, output is undefined/unreliable.
Sticky notes applying:
- “## 2. AI Language Detection & Smart Routing”
Version notes: typeVersion 1.

Node: Smart Router & Targets

Type / Role: n8n-nodes-base.function — Decide detected language, targets, and whether to fix WP language flag.
Config (interpreted):
- Reads:
  - originalPost from $items("Prepare Text Snippet")[0].json
  - detectedLang from items[0].json.message.content
  - currentWpLang from originalPost.lang (defaults to en)
- Supported languages configured in allSupported = ['en', 'fr', 'de'] (must be updated to match site).
- Computes languagesNeeded:
  - targets = allSupported excluding detectedLang
  - then removes languages already present in originalPost.translations map
- Computes needsLangFix if WP language flag differs from detected language.
- Outputs postTypeUrl from originalPost._links.self[0].href
Inputs/Outputs:
- Input ← OpenAI Language Detect
- Output → Needs WP Lang Fix?
Edge cases / failures:
- If OpenAI output not ISO code, toLowerCase() still works but logic may mis-route.
- If _links.self[0].href missing, language fix and base URL reconstruction later will fail.
- If translations structure differs (not a {lang: id} map), duplicate-prevention fails.
Sticky notes applying:
- “## 2. AI Language Detection & Smart Routing”
- “### Configuration:\nOpen this node and update the "allSupported" array to match your website's languages.”
Version notes: typeVersion 1.

Node: Needs WP Lang Fix?

Type / Role: n8n-nodes-base.if — Conditional branch.
Condition:
- {{$json.needsLangFix}} == true
Routing:
- True → Fix WP Language Flag
- False → Merge (input 2 / index 1)
Inputs/Outputs:
- Input ← Smart Router & Targets
Edge cases:
- If needsLangFix is missing/non-boolean, condition may evaluate unexpectedly.
Version notes: typeVersion 1.

Node: Fix WP Language Flag

Type / Role: n8n-nodes-base.httpRequest — Updates original post language in WP.
Config (interpreted):
- URL: {{$json.postTypeUrl}} (self link of the original post)
- Method: POST
- Body parameter: lang = {{$json.detectedLang}}
- Authentication: HTTP Basic Auth (WordPress Application Password)
Inputs/Outputs:
- Input ← Needs WP Lang Fix? (true branch)
- Output → Merge (input 1 / index 0)
Edge cases / failures:
- WP may require PUT depending on configuration; many WP REST endpoints accept POST for updates, but not all.
- If multilingual plugin uses different field name than lang, update will have no effect.
- Auth/permission errors, or REST API disabled.
Version notes: typeVersion 3.

Node: Merge

Type / Role: n8n-nodes-base.merge — Joins the “fixed” and “not fixed” branches back into a single flow.
Config:
- Default merge behavior (no explicit mode shown).
Inputs/Outputs:
- Input 0 ← Fix WP Language Flag
- Input 1 ← Needs WP Lang Fix? (false path)
- Output → Needs Translations?
Edge cases:
- Depending on merge mode, you may get unexpected output if both branches run (here only one runs).
Version notes: typeVersion 2.1.

Node: Needs Translations?

Type / Role: n8n-nodes-base.if — Stops early if nothing to translate.
Condition:
- {{$json.languagesNeeded.length > 0}} == true
Routing:
- True → Extract Content
- False → flow ends (no further nodes)
Inputs/Outputs:
- Input ← Merge
Edge cases:
- If languagesNeeded missing/not an array, expression errors.
Version notes: typeVersion 1.

Block 1.3 — Flatten ACF Structure & Translate with DeepL

Overview:
Builds an ordered list of strings to translate and a matching list of “map paths” indicating where each translated string should be written back. Then splits the execution into one item per target language and sends the translation request to DeepL.

Nodes Involved:

Extract Content
Split by Target Lang
DeepL Translate

Node: Extract Content

Type / Role: n8n-nodes-base.code — Extracts translatable strings from Gutenberg + ACF, preserving mapping.
Config (interpreted):
- Builds:
  - stringsToTranslate: ordered array
  - mapPaths: same length; entries like post_title, post_content, acf.intro, acf.product_flexible[0].featureTitle, etc.
- Always includes post.title.rendered mapped to post_title
- Includes post.content.rendered mapped to post_content if non-empty
- Adds acf.intro if present
- Walks nested ACF objects/arrays to find keys listed in textKeys (must be customized to your ACF schema)
- Special handling: if acf.product_flexible exists, it recursively walks it
Inputs/Outputs:
- Input ← Needs Translations? (true branch)
- Output → Split by Target Lang
Edge cases / failures:
- If ACF field names differ from textKeys, important text won’t be translated.
- If ACF contains rich structures with unexpected types, walk() is defensive but can still traverse huge objects (performance).
- Gutenberg content is translated as HTML; later DeepL uses tag_handling=html.
Sticky notes applying:
- “## 3. Flatten ACF Structure & Translate with DeepL”
- “### ACF Keys\nOpen this node and update the "textKeys" array to match the actual ACF field names used on your WordPress site”
Version notes: typeVersion 2 (Code node, JS).

Node: Split by Target Lang

Type / Role: n8n-nodes-base.itemLists — Creates one item per needed target language.
Config (interpreted):
- Splits field: languagesNeeded
- Include: “allOtherFields” (keeps the rest of the JSON on each split item)
- alwaysOutputData: true (prevents empty output in some scenarios)
Resulting shape:
- Each item has languagesNeeded set to a single language (string), plus original data.
Inputs/Outputs:
- Input ← Extract Content
- Output → DeepL Translate
Edge cases:
- If languagesNeeded is empty, behavior depends on node internals; upstream IF prevents this.
Version notes: typeVersion 1.

Node: DeepL Translate

Type / Role: n8n-nodes-base.httpRequest — Calls DeepL translate API.
Config (interpreted):
- URL: https://api-free.deepl.com/v2/translate (free endpoint; paid is api.deepl.com)
- Method: POST
- Headers:
  - Authorization: DeepL-Auth-Key YOUR_DEEPL_API_KEY_HERE (must be replaced)
- Body parameters:
  - text = {{$json.stringsToTranslate}} (array; DeepL supports repeated text parameters—n8n will serialize)
  - target_lang = {{$json.languagesNeeded.toUpperCase()}}
  - source_lang = {{$json.detectedLang.toUpperCase()}}
  - tag_handling = html
Inputs/Outputs:
- Input ← Split by Target Lang
- Output → RECONSTRUCT JSON
Edge cases / failures:
- 403 if key invalid/wrong endpoint (free vs paid mismatch).
- DeepL target language codes differ from ISO639-1 in some cases (e.g., EN-GB, PT-PT); this workflow assumes simple 2-letter codes.
- HTML handling can still mangle Gutenberg blocks in edge cases; consider preserve_formatting or glossary if needed.
Sticky notes applying:
- “## 3. Flatten ACF Structure & Translate with DeepL”
- “### DEEPL Translate\nAdd your DeepL API Key in the Header Parameters.”
Version notes: typeVersion 3.

Block 1.4 — Rebuild ACF Object & Push to WordPress

Overview:
Matches DeepL’s returned translations to the original mapPaths, writes them back into a cloned post object (title/content/ACF), cleans empty strings, then POSTs a new translated draft post to WordPress with translation linkage.

Nodes Involved:

RECONSTRUCT JSON
Create Translated Post

Node: RECONSTRUCT JSON

Type / Role: n8n-nodes-base.code — Rebuilds a translated WP payload per language.
Config (interpreted):
- Pulls the original split items via $items("Split by Target Lang") to align translation results with the correct language and map paths.
- For each translated item:
  - translatedTexts = items[i].json.translations (expects DeepL response shape: array of { text: "..." })
  - Clones originalPost
  - For each mapPaths[index]:
    - post_title → writes newPost.title.raw + newPost.title.rendered
    - post_content → writes newPost.content.raw + newPost.content.rendered
    - acf.* → uses a custom set() to write nested values (supports arrays via [index])
  - Runs cleanAcf():
    - converts empty strings "" to null recursively
  - Derives createUrl:
    - takes originalData.postTypeUrl and removes the trailing /{id} segment
- Outputs per language:
  - lang (target language)
  - originalId
  - newTitle, newAcf, newContent
  - createUrl (collection endpoint)
Inputs/Outputs:
- Input ← DeepL Translate
- Output → Create Translated Post
Edge cases / failures:
- Assumes DeepL response is items[i].json.translations and index-aligned with stringsToTranslate; if DeepL returns errors/partial results, mapping breaks.
- set() assumes intermediate objects/arrays exist; if ACF structure differs between original and expected path, may throw.
- splitItems[i] alignment relies on n8n preserving item order between Split → DeepL → this node; usually true, but consider adding explicit correlation IDs if you later parallelize.
Sticky notes applying:
- “## 4. Rebuild ACF Object & Push to WordPress”
Version notes: typeVersion 2.

Node: Create Translated Post

Type / Role: n8n-nodes-base.httpRequest — Creates translated WP post (draft).
Config (interpreted):
- URL: {{$json.createUrl}} (e.g., https://.../wp-json/wp/v2/posts)
- Method: POST
- Auth: HTTP Basic Auth (WordPress Application Password)
- Body parameters:
  - title = {{$json.newTitle}}
  - acf = {{$json.newAcf}}
  - status = draft
  - lang = {{$json.lang}}
  - translation_of = {{$json.originalId}}
  - content expression is set as =={{ $json.newContent }} (note the double ==, likely a mistake; should be ={{ $json.newContent }})
Inputs/Outputs:
- Input ← RECONSTRUCT JSON
- Output: ends workflow
Edge cases / failures:
- Expression bug risk: content uses =={{ ... }} which may post literal text or fail; should be corrected.
- WP may reject acf unless ACF-to-REST is enabled and user has permissions.
- translation_of and lang fields depend on multilingual plugin; if unsupported, these fields will be ignored or cause validation errors.
Sticky notes applying:
- “## 4. Rebuild ACF Object & Push to WordPress”
- “### URL and WP APP Password\nUpdate the URL to your WP domain and select your WP App Password credentials.”
Version notes: typeVersion 3.

3. Summary Table

Node Name	Node Type	Functional Role	Input Node(s)	Output Node(s)	Sticky Note
Webhook	Webhook Trigger	Receives WP post event payload	—	Get Post Data	## 1. Trigger & Fetch Raw WP Post Data
Get Post Data	HTTP Request	Fetches full post JSON from WP REST API	Webhook	Prepare Text Snippet	## 1. Trigger & Fetch Raw WP Post Data; ### URL and WP APP Password\nUpdate the URL to your WP domain and select your WP App Password credentials.
Prepare Text Snippet	Function	Builds 1000-char snippet for language detection	Get Post Data	OpenAI Language Detect	## 2. AI Language Detection & Smart Routing
OpenAI Language Detect	OpenAI (Chat)	Detects source language ISO code	Prepare Text Snippet	Smart Router & Targets	## 2. AI Language Detection & Smart Routing
Smart Router & Targets	Function	Computes needed target languages + whether to fix WP lang	OpenAI Language Detect	Needs WP Lang Fix?	## 2. AI Language Detection & Smart Routing; ### Configuration:\nOpen this node and update the "allSupported" array to match your website's languages.
Needs WP Lang Fix?	IF	Branch: update WP language flag or skip	Smart Router & Targets	Fix WP Language Flag; Merge	## 2. AI Language Detection & Smart Routing
Fix WP Language Flag	HTTP Request	Updates original post `lang` in WP	Needs WP Lang Fix?	Merge	## 2. AI Language Detection & Smart Routing
Merge	Merge	Rejoins “fixed” and “not fixed” paths	Fix WP Language Flag; Needs WP Lang Fix?	Needs Translations?	## 2. AI Language Detection & Smart Routing
Needs Translations?	IF	Stops if no missing translations	Merge	Extract Content	## 3. Flatten ACF Structure & Translate with DeepL
Extract Content	Code	Flattens title/content/ACF strings + creates mapping paths	Needs Translations?	Split by Target Lang	## 3. Flatten ACF Structure & Translate with DeepL; ### ACF Keys\nOpen this node and update the "textKeys" array to match the actual ACF field names used on your WordPress site
Split by Target Lang	Item Lists	One item per language to translate	Extract Content	DeepL Translate	## 3. Flatten ACF Structure & Translate with DeepL
DeepL Translate	HTTP Request	Batch-translates strings with HTML handling	Split by Target Lang	RECONSTRUCT JSON	## 3. Flatten ACF Structure & Translate with DeepL; ### DEEPL Translate\nAdd your DeepL API Key in the Header Parameters.
RECONSTRUCT JSON	Code	Rebuilds translated post payload + ACF structure	DeepL Translate	Create Translated Post	## 4. Rebuild ACF Object & Push to WordPress
Create Translated Post	HTTP Request	Creates translated draft post linked to original	RECONSTRUCT JSON	—	## 4. Rebuild ACF Object & Push to WordPress; ### URL and WP APP Password\nUpdate the URL to your WP domain and select your WP App Password credentials.
Sticky Note	Sticky Note	Project info / prerequisites / setup notes	—	—	# Smart WP Translation (Gutenberg + ACF)\n\nThis workflow automatically translates new or updated WordPress posts into multiple languages using DeepL, while keeping your Advanced Custom Fields (ACF) JSON structure intact. It uses OpenAI to detect the source language and prevent duplicate translations.\n\n🛠️ Prerequisites:\n1. WordPress Application Password (for HTTP Basic Auth)\n2. OpenAI API Key\n3. DeepL API Key\n\n🚀 Setup Instructions:\n1. Add your credentials to the Webhook, Get Post Data, OpenAI, and Create Post nodes.\n2. In the "Smart Router & Targets" node, configure your supported languages.\n3. In the "Extract Content" node, update the array with your site's specific ACF keys.\n4. In the "DeepL Translate" node, add your DeepL API Key to the Header.\n\nP.S.\n\nYou can use the Deep Node, but in this template I use a generic node in case you get issues with the Deepl Node
Sticky Note1	Sticky Note	Section label	—	—
Sticky Note2	Sticky Note	Section label	—	—
Sticky Note3	Sticky Note	Section label	—	—
Sticky Note4	Sticky Note	Section label	—	—
Sticky Note5	Sticky Note	Reminder	—	—
Sticky Note6	Sticky Note	Reminder	—	—
Sticky Note7	Sticky Note	Reminder	—	—
Sticky Note8	Sticky Note	Reminder	—	—
Sticky Note9	Sticky Note	Reminder	—	—

4. Reproducing the Workflow from Scratch

Create a Webhook trigger
1. Add node: Webhook
2. Set:
  - HTTP Method: POST
  - Path: translate-post
3. (Recommended) Add authentication/secret validation (not included in template).
Fetch the WordPress post from REST API
1. Add node: HTTP Request named Get Post Data
2. Configure:
  - Authentication: HTTP Basic Auth (use WP Application Password credentials)
  - URL (expression):
    https://your-wordpress-site.com/wp-json/wp/v2/{{ $json.body.post.post_type === 'post' ? 'posts' : $json.body.post.post_type === 'solution' ? 'solutions' : $json.body.post.post_type }}/{{ $json.body.post.ID }}
3. Create credentials in n8n:
  - Credential type: HTTP Basic Auth
  - Username: WP username
  - Password: WordPress Application Password
4. Connect: Webhook → Get Post Data
Prepare a short text snippet
1. Add node: Function named Prepare Text Snippet
2. Paste logic that:
  - Builds textSnippet from title.rendered and stripped content.rendered (or acf.intro)
  - Truncates to 1000 chars
3. Connect: Get Post Data → Prepare Text Snippet
Detect language with OpenAI
1. Add node: OpenAI named OpenAI Language Detect
2. Resource: Chat
3. Model: gpt-4o-mini
4. Messages:
  - System: instruction to output only 2-letter code
  - User: include the snippet, e.g. {{ $json.textSnippet }}
5. Add OpenAI credentials (API key).
6. Connect: Prepare Text Snippet → OpenAI Language Detect
Compute routing (targets + fix flag)
1. Add node: Function named Smart Router & Targets
2. Implement:
  - allSupported array (edit to match your site, e.g. ['en','fr','de'])
  - detectedLang from OpenAI output
  - Remove already existing translations using originalPost.translations
  - Output: languagesNeeded, needsLangFix, postTypeUrl, originalId
3. Connect: OpenAI Language Detect → Smart Router & Targets
Optional: Fix original WP language flag
1. Add node: IF named Needs WP Lang Fix?
  - Condition: {{ $json.needsLangFix }} is true
2. Add node: HTTP Request named Fix WP Language Flag
  - URL: {{ $json.postTypeUrl }}
  - Method: POST (or adapt to PUT if your WP endpoint requires it)
  - Auth: same WP Basic Auth credential
  - Body: lang = {{ $json.detectedLang }}
3. Add node: Merge named Merge to rejoin branches
  - Connect:
    - IF true → Fix WP Language Flag → Merge (Input 1)
    - IF false → Merge (Input 2)
Stop if nothing to translate
1. Add node: IF named Needs Translations?
  - Condition: {{ $json.languagesNeeded.length > 0 }} is true
2. Connect: Merge → Needs Translations?
Extract/flatten content and ACF strings
1. Add node: Code named Extract Content
2. Configure:
  - Maintain textKeys list and customize it to your ACF field names.
  - If you have flexible/repeater fields, adapt the walk() entry points (e.g., acf.product_flexible).
3. Connect: Needs Translations? (true) → Extract Content
Split per target language
1. Add node: Item Lists named Split by Target Lang
2. Operation: split out items by field languagesNeeded
3. Include: “all other fields”
4. Connect: Extract Content → Split by Target Lang
Translate with DeepL
Add node: HTTP Request named DeepL Translate
Configure:
- URL: https://api-free.deepl.com/v2/translate (or paid endpoint)
- Method: POST
- Header: Authorization = DeepL-Auth-Key <YOUR_KEY>
- Body:
  - text = {{ $json.stringsToTranslate }}
  - source_lang = {{ $json.detectedLang.toUpperCase() }}
  - target_lang = {{ $json.languagesNeeded.toUpperCase() }}
  - tag_handling = html
Connect: Split by Target Lang → DeepL Translate
Reconstruct translated payload
Add node: Code named RECONSTRUCT JSON
Implement:
- Reapply DeepL translations to the cloned post using mapPaths
- Write back to title/content/ACF
- Clean empty strings in ACF (optional but recommended)
- Compute createUrl from the original self URL
Connect: DeepL Translate → RECONSTRUCT JSON
Create translated draft in WordPress
Add node: HTTP Request named Create Translated Post
Configure:
- URL: {{ $json.createUrl }}
- Method: POST
- Auth: WP Basic Auth credential
- Body:
  - title = {{ $json.newTitle }}
  - content = {{ $json.newContent }} (ensure the expression starts with a single =)
  - acf = {{ $json.newAcf }}
  - status = draft
  - lang = {{ $json.lang }}
  - translation_of = {{ $json.originalId }} (plugin-dependent)
Connect: RECONSTRUCT JSON → Create Translated Post

5. General Notes & Resources

Note Content	Context or Link
Smart WP Translation (Gutenberg + ACF): automatically translates posts into multiple languages using DeepL, preserves ACF JSON, uses OpenAI to detect source language and prevent duplicates.	Workflow sticky note (project overview)
Prerequisites: WordPress Application Password (HTTP Basic Auth), OpenAI API key, DeepL API key.	Workflow sticky note (prerequisites)
Setup: update WP domain URLs, configure supported languages in “Smart Router & Targets”, update ACF `textKeys` in “Extract Content”, add DeepL key in “DeepL Translate” header.	Workflow sticky note (setup instructions)
“You can use the DeepL node, but in this template I use a generic node in case you get issues with the Deepl Node.”	Workflow sticky note (implementation choice)

27 KiB Raw Blame History Unescape Escape

Translate WordPress posts and ACF fields using DeepL and OpenAI

1. Workflow Overview

1.1 Trigger & Fetch Raw WP Post Data

1.2 AI Language Detection & Smart Routing

1.3 Flatten ACF Structure & Translate with DeepL

1.4 Rebuild ACF Object & Push to WordPress

2. Block-by-Block Analysis

Block 1.1 — Trigger & Fetch Raw WP Post Data

Node: Webhook

Node: Get Post Data

Block 1.2 — AI Language Detection & Smart Routing

Node: Prepare Text Snippet

Node: OpenAI Language Detect

Node: Smart Router & Targets

Node: Needs WP Lang Fix?

Node: Fix WP Language Flag

Node: Merge

Node: Needs Translations?

Block 1.3 — Flatten ACF Structure & Translate with DeepL

Node: Extract Content

Node: Split by Target Lang

Node: DeepL Translate

Block 1.4 — Rebuild ACF Object & Push to WordPress

Node: RECONSTRUCT JSON

Node: Create Translated Post

3. Summary Table

4. Reproducing the Workflow from Scratch

5. General Notes & Resources

27 KiB

Raw Blame History