I simply don’t understand how the character text splitter works… it makes absolutely no sense to me… I tried to use Fiexed, Expression … I used “…” I used spaces in between… the node always inserts the text as one row into the Supabase Vector Store.
I want the seperator to trigger and not the chunk size. I set the chunk size to 10,100,1000 it always inserts it aas one block. If I use the recursive character text splitter in the same flow it perfectly seperates it.
Information on your n8n setup
n8n version: 1.10
Database (default: SQLite): SQLite
n8n EXECUTIONS_PROCESS setting (default: own, main): main
Running n8n via (Docker, npm, n8n cloud, desktop app): Docker
In fact I took it onto myself to gather the text from your previous post and re-use it. This is what I used:
Text separated with '...'
Q: What are your services?
A: We focus on improving your operations by analyzing processes and creating a data strategy.
This sets the foundation for larger projects, like integrating AI with your ERP or CRM systems,
allowing faster and more reliable interaction with your data. With a solid data structure in place,
you're better positioned to automate key parts of your processes.
Simply put, we cover consulting, digitization, process enhancement/automation, and custom software development to fit your needs.
...
Q: What about pricing?
A: We set pricing individually after a phone or video call to better understand your needs. For simpler tasks, we can usually give you a fixed price upfront.
But for bigger projects, like ERP/CRM integrations and data structuring, it’s harder to nail down costs from the start — much like large infrastructure projects.
That’s why we prefer to work in sprints, allowing us to adapt to new requirements as they arise and better plan the project as it evolves, avoiding the pitfalls of upfront cost estimates that may not capture everything.
...
Q: What is included in the maintenance?
A: The maintenance covers technology updates for ongoing operations and enhancements from gathering better interactions and the knowledge sheet, ensuring continuous improvement and knowledge enrichment based on client interactions.
...
Q: What do I need to provide for these chatbots?
A: It’s helpful if you already have some of your data digitized and structured, but if not, don’t worry—we can handle that for you.
The key is how your data is organized and how we can access and manage it, especially as it changes over time.
...
Q: How do you differentiate from other AI service providers?
A: This question often receives generic responses like "quality is our main focus," which doesn’t offer much insight.
But we aim to genuinely enhance your business by implementing logical solutions — meaning solutions that really make sense within your business, not something that sounds cool but does not provide any value.
Here is the workflow I created for embedding:
I chose the chunk size of 10 to enforce splitting pattern based on what was discussed in the previous topic and the delimiter of “…”. Here is what the execution looked like:
The problem seems to have been that I entered “…” manually into Google Docs and then manually into N8N… which wasnt ASCI so just 2 different types of dots. Funny how something that simple takes that long to figure out. Thank you!
