Description

n8n-nodes-ldxhub


> ✅ Free to try — 25,000 credits/month, no credit card required
> ✅ One key for everything — OpenAI, Anthropic, Google, AWS, Azure, xAI
> ✅ 30-second sign-up — GitHub, Google, or email; your API key is shown immediately

n8n community node for LDX hub — AI-powered document processing platform: structured data extraction (StructFlow), vision document analysis (AnalyzeDoc), XLIFF translation refinement (RefineLoop), layout-preserving OCR (RenderOCR), text-based PDF conversion (CastDoc), and plain-text/JSONL extraction (ExtractDoc).

Table of Contents

• Features
• Quick Start
• Installation
• Prerequisites
• Credentials Setup
• Usage

– StructFlow
– AnalyzeDoc
– RefineLoop
– RenderOCR
– CastDoc
– ExtractDoc

• AI Agent Integration
• Polling Settings
• Troubleshooting
• Support
• Changelog
• License

Quick Start

After installation:

1. Get a free API key at gw.portal.ldxhub.io (no credit card)
2. Import all-services-demo.json
3. Create an LDXhub API credential with both fields set to expressions:
– Base URL: ={{ $('On form submission').item.json.host }}
– API Key: ={{ $('On form submission').item.json.api_key }}
4. Select this credential in each LDXhub node
5. Run the workflow — enter your API key (and optionally a custom host) in the form

All 5 services (StructFlow / RefineLoop / RenderOCR / CastDoc / ExtractDoc) accessible from a single Form Trigger. The API key is consumed per execution, never stored in n8n.

For traditional single-service usage with a static credential, see Usage below.

Features

• StructFlow: Extract structured JSON from unstructured text using AI models (medical records, customer feedback, legal documents, and more)
• AnalyzeDoc: Extract structured JSON from PDF/JPEG/PNG documents using vision AI models, following your prompt and example output (the image counterpart to StructFlow)
• RefineLoop: Iteratively improve XLIFF translation quality using frontier AI models (Google Gemini, Anthropic Claude, OpenAI GPT, and more)
• RenderOCR: Convert PDFs and images to Word/Excel/PowerPoint with layout-preserving OCR (via industry-leading OCR engines)
• CastDoc: Convert text-based PDFs to Word/Excel/PowerPoint without OCR (high-fidelity layout preservation for digital-born documents)
• ExtractDoc: Extract plain text or JSONL from PDF/DOCX/XLSX/PPTX using the ki/extract engine (no AI, no OCR, free tier) — ideal as a preprocessing step before StructFlow
• One credential for all AI providers: OpenAI, Anthropic Claude, Google Gemini, AWS Nova, Azure OpenAI, xAI Grok — accessed through a single LDX hub API key
• HTTP long-polling architecture — compatible with n8n Cloud execution model
• Proven at scale: tested with 1.19M-character academic papers

Installation

In your n8n instance:

1. Go to Settings → Community Nodes
2. Click Install
3. Enter n8n-nodes-ldxhub and confirm

Prerequisites

• An LDX hub account
• n8n version supporting community nodes (v1.x or later)
• A paid or free tier subscription (free tier includes 25,000 credits/month, suitable for evaluation)

Credentials Setup

1. Sign up at the LDX hub DevPortal — free, no credit card. Use GitHub, Google, or email; your API key is shown immediately after sign-up.
2. Choose a subscription plan — start with Free (25,000 credits/month) to evaluate, or pick Starter / Standard / Pro for production. See pricing for details.
3. Click your account email in the top-right of the DevPortal, then select My Subscriptions
4. Under API Keys, copy the Current Key
5. In n8n, create a new LDXhub API credential:
– Base URL: https://gw.ldxhub.io (default; leave as-is for production)
– API Key: paste the key from step 4
6. Click Save

If API calls fail with 401 Unauthorized, verify:

• The API key is active (shown with a green dot in the DevPortal)
• The Base URL has no trailing slash
• Your network allows outbound HTTPS to gw.ldxhub.io

Dynamic API Key (Advanced)

For workflows where the API key should be provided dynamically per execution (e.g., from a Form Trigger, an upstream node, or an AI agent context), set the API Key field of your LDXhub API credential to an n8n expression instead of a static value:

1. Create a new LDXhub API credential
2. Switch the API Key field to Expression mode
3. Enter an expression that resolves at execution time, e.g., {{ $json.api_key }}
4. Save the credential

The Model and Engine dropdowns load from LDX hub’s public metadata endpoints without authentication, so the node UI works normally during workflow design even when the API Key is an unresolved expression.

Example: Dynamic credentials workflow

Usage

> 💡 Examples: All ready-to-use workflows are in examples/. Each individual demo requires a LDXhub API credential and a placeholder input file path to be updated.

1. Add LDXhub API credentials (see Credentials Setup)
2. Add the LDXhub node to your workflow
3. Select a resource and operation (see below)

StructFlow — extract structured data from text

1. Resource: StructFlow → Operation: Run Extraction Job
2. Configure Model, System Prompt, and Example Output
3. Choose Input Mode:
– Inline Inputs: Provide ID + Data pairs directly in the workflow (good for small batches, quick prototyping)
– Binary File: Provide a JSONL file as binary input (good for large batches, or as part of an ExtractDoc → StructFlow pipeline)

Examples: Inline mode · Binary mode · Dynamic API key

AnalyzeDoc — extract structured JSON from documents (vision AI)

1. Resource: AnalyzeDoc → Operation: Run Analysis Job
2. Provide a PDF, JPEG, or PNG file via binary input
3. Choose a vision model, write a System Prompt, provide an Example Output (JSON), and pick the output format

RefineLoop — XLIFF translation refinement

1. Resource: RefineLoop → Operation: Run Refinement Job
2. Provide an XLIFF file via binary input
3. Choose an AI model and set max revisions

Example: RefineLoop workflow

RenderOCR — PDF/image to Office

1. Resource: RenderOCR → Operation: Run Conversion Job
2. Provide a PDF or image file via binary input
3. Choose an OCR engine, target language, and output format (docx/xlsx/pptx)

Example: RenderOCR workflow

CastDoc — text-based PDF to Office (no OCR)

1. Resource: CastDoc → Operation: Run Conversion Job
2. Provide a PDF file via binary input
3. Choose an engine and output format (docx/xlsx/pptx)

Example: CastDoc workflow

ExtractDoc — plain text or JSONL extraction (no AI, no OCR)

1. Resource: ExtractDoc → Operation: Run Conversion Job
2. Provide a PDF / DOCX / XLSX / PPTX file via binary input
3. Choose the ki/extract engine and output format (text or jsonl)

Example: ExtractDoc workflow

AI Agent Integration

The LDXhub node is marked as usableAsTool: true, so it can be attached to an AI Agent node as a tool. This enables agentic workflows where an AI agent autonomously decides when to extract structured data, translate documents, or convert files using LDX hub.

Example use cases:

• Customer support agent that extracts structured complaint data from incoming emails (StructFlow)
• Document processing agent that automatically OCRs and translates uploaded PDFs (RenderOCR → RefineLoop)
• Knowledge base ingestion agent that converts and structures diverse document formats (CastDoc → StructFlow)

Polling Settings

For large documents, jobs may take several minutes. The node polls until completion:

| Setting | Default | Description |
|—|—|—|
| Max Polling Attempts | 180 | Maximum number of poll requests |
| Server Wait Seconds | 10 | Server-side long-poll wait per request |

Theoretical max wait = Max Polling Attempts × Server Wait Seconds seconds.
Defaults give 30 minutes. For longer documents, increase Max Polling Attempts
(e.g., 360 for 60 minutes).

> n8n Cloud users: your plan’s workflow execution timeout applies independently. Check your plan’s limits.

Troubleshooting

401 Unauthorized

• The API key is invalid, revoked, or expired
• Roll the key from My Subscriptions → API Keys → Roll API Key in the LDX hub DevPortal, then update the n8n credential

400 Bad Request — invalid file_id

• The binary input is missing or the binary field name is incorrect
• Check that the previous node outputs a binary property matching the Input Binary Field setting (default: data)

Job times out / polling exhausted

• Large documents may exceed the default 30-minute window
• Increase Max Polling Attempts in Polling Settings
• For n8n Cloud, also check your plan’s workflow execution timeout

StructFlow Inline mode — empty results

• Ensure Inputs collection has at least one record with non-empty ID and Data fields
• Verify Example Output is valid JSON

Credit limit exceeded

• Your subscription’s monthly credit allowance has been reached
• Check usage in the DevPortal’s My Subscriptions page
• Upgrade your plan or wait for the next billing period

Support

• Product: https://ldxlab.io/ldxhub
• Documentation: https://gw.portal.ldxhub.io/introduction
• Bug reports & feature requests: GitHub Issues

Changelog

See CHANGELOG.md for the full version history.

License

MIT

Copyright (c) 2026 Kawamura International Co., Ltd.