Docling in n8n

[patelc75]

Hi all, at one point I believe there was an n8n community node for Docling but can't seem to find it. Any recommended approach to use Docling in n8n? I am using this n8n template for reference. Thank you!

[dosubot[bot]]

Hi @patelc75! I'm Dosu and I’m helping the docling team.

There isn’t currently an n8n community node for Docling, but you can integrate Docling into n8n workflows using the HTTP Request node to call Docling Serve’s API endpoints. Here’s how you can set it up:

1. Run Docling Serve as an API service:
   Install with pip install "docling-serve[ui]" and start the server using docling-serve run --enable-ui. By default, the API will be available at http://127.0.0.1:5001 with docs at /docs and a UI at /ui
   source.

2. Configure n8n HTTP Request node:

   • For document conversion from a URL, POST to /v1/convert/source with a JSON body like:

     {
       "sources": [
         { "kind": "http", "url": "https://arxiv.org/pdf/2501.17887" }
       ],
       "options": {
         "to_formats": ["md", "json"],
         "do_ocr": true,
         "table_mode": "accurate"
       }
     }

   • For file uploads, use /v1/convert/file with multipart/form-data (n8n supports this).
   • If authentication is enabled, add the X-Api-Key header with your API key.
   • The response will include the converted document in your chosen formats (markdown, JSON, etc.)
     source.

3. Best practices for RAG workflows:

   • Use Docling’s chunking endpoints (/v1/chunk/hybrid/source or /v1/chunk/hierarchical/source) to split documents into semantically meaningful chunks for embedding and retrieval.
   • Docling’s JSON output preserves layout and metadata, which is useful for downstream semantic search or Qdrant ingestion.
   • For invoices or multi-column layouts, work with the JSON output directly to retain structure, as markdown export flattens columns
     source.

4. Error handling:

   • If you see a “field required” error, ensure you’re using the correct payload structure: use "sources" (not "http_sources") and include "kind": "http" for URLs
     source.

5. Advanced:

   • Async endpoints (/v1/convert/source/async) are available for large files or long-running jobs, with status polling and result retrieval.
   • Docling can run fully offline by downloading models in advance and setting DOCLING_SERVE_ARTIFACTS_PATH
     source.

Let me know if you want a sample n8n workflow JSON or have a specific use case!

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

   

[festusthomas]

Hello @dosu,
Thanks for the above helpful post.
I would appreciate if you can send me a sample n8n workflow JSON.
I have installed the docling docker in a coolify platform