Chunk Overlap Not Working with Text Splitters in Default Data Loader (n8n Workflow)

Questions
1

Hi n8n Support,
I’m encountering a critical issue in my document embedding workflow related to chunk overlap not functioning as expected when using text splitters in combination with the Default Data Loader.
:wrench: Workflow Overview:
I’m working on a use case where PDF documents are loaded, split into overlapping chunks, embedded using OpenAI, and then stored in Pinecone. Here’s how the flow is structured:
1. PDF documents are input via Default Data Loader (Type: JSON, Mode: Load All Input Data).
2. Text Splitting is set to Custom.
3. I have tested each of the following Text Splitters individually by connecting them to the Text Splitter input:
• Recursive Character Text Splitter
• Character Text Splitter
• Token Splitter
I ran each splitter in isolation — one at a time — and executed the full workflow after each configuration.
⸻
:gear: Configuration Attempted:
For each of the splitter types above, I configured:
• Chunk Size: 6000
• Chunk Overlap: 1000
• Tried various separators: " ", “”, “\n”, etc.
The output was then passed to the OpenAI Embeddings node and on to the Pinecone vector store.
⸻
:cross_mark: The Problem:
Despite setting chunk overlap correctly in every Text Splitter, no overlapping content is present between the resulting chunks. Each chunk is completely sequential, with no shared or duplicated text between adjacent chunks — even though the overlap is defined.
I checked the output from the Default Data Loader after each test run. The number of output chunks remains constant (e.g. 428), and no chunk contains any overlapping part of the previous one.
⸻
I’ve reviewed all available documentation and verified the connections and settings. Everything executes cleanly — but chunk overlap is simply never applied.
I’d really appreciate a clear clarification or workaround for this issue.
Best regards,
Dean Savkovic

What is the error message (if any)?

Please share your workflow

(Select the nodes on your canvas and use the keyboard shortcuts CMD+C/CTRL+C and CMD+V/CTRL+V to copy and paste the workflow.)

Share the output returned by the last node

Information on your n8n setup

  • n8n version:
  • Database (default: SQLite):
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app):
  • Operating system: