I have a TERRIBLY low success rate doing what should be a basic workflow - ingesting my existing documents (PDFs, DOCXs, RTFs, etc) into a Vector store, so I can later query and chat against this data source. Nothing seems to work, certainly not with any reliably to make continued use of this platform worthwhile.
I am asking as last resort if I am doing this wrong, or need some extra steps? Else I will move on from this project.
I have copied / tried / tested MANY existing workflows. There is no point uploading my own, as none of them work! Some published workflows use “Read/Write Docs” node, some don’t. Some use custom code to mess about with stuff. NONE are documented! I have a local Ollama.app running on my M1 MacBook Air, and typically use the nomic model to perform the data conversion / ingestion.
I have started from the Starter Kit Docker deployment, that includes the n8n container, as well as containers for Postgres and Qdrant.
Information on your n8n setup
**n8n version: 1.67.1 (calling the n8nio/n8n:latest image)
**Database (default: SQLite): I am actually trying to get ANY data into ANY vector data store! Qdrant, Postres, any thing I can run locally, I do not care.
**n8n EXECUTIONS_PROCESS setting (default: own, main): Any execution!
**Running n8n via (Docker, npm, n8n cloud, desktop app): Docker
Sorry to hear that you’re having issues with your workflow When you say ‘nothing seems to work’, could you describe what’s not working - what errors are you seeing? which workflow templates have you tried and which part of the workflow are you stuck on?
It would help us and the community a lot if you could share your workflow or what you’ve tried so far, even if it’s not working (most of the workflows people post in the forum aren’t!). You can learn how to do that here:
Tip for sharing your workflow in the forum
Pasting your n8n workflow
Ensure to copy your n8n workflow and paste it in the code block, that is in between the pairs of triple backticks, which also could be achieved by clicking </> (preformatted text) in the editor and paste in your workflow.
```
<your workflow>
```
Make sure that you’ve removed any sensitive information from your workflow and include dummy data or pinned data as much as you can!
Below is an example workflow I have tried. When a file is uploaded to a specific folder within the n8n container, it should read / extract / whatever other actions that might be needed. Then add it to a Vector Store (Qdrant in this case, as I am also struggling to get PGVector Store to work).
Viewing other workflows, people use various alternatives and/or tweaks to this, with custom JS code, more nodes of different types, etc.
I was hoping to have a working workflow that I can then expand upon, to include more data types & sources. MVP would be PDFs, DOCX, RTF, TXT and HTML (preferably retrieved from a HTTP Request node).
When I execute this by connecting the Extract from File and Qdrant nodes (I do not see an option to upload a RTF file), the Extract from Files node results in 8 items. Viewing it looks like it picks up 1 table within the RTF document, but nothing else. Nothing is inserted into the Qdrant Vector Store.
When you say ‘nothing seems to work’, could you describe what’s not working - what errors are you seeing? which workflow templates have you tried and which part of the workflow are you stuck on?