Noob question- creating chat against a set of preloaded docs

[rpeter]

Hi all,
Please be patient - I am new to this and learning by doing.

I installed GitHub - n8n-io/self-hosted-ai-starter-kit: The Self-hosted AI Starter Kit is an open-source template that quickly sets up a local AI environment. Curated by n8n, it provides essential tools for creating secure, self-hosted AI workflows. (amazing btw) and I can run a query, with a file, and it does a good job answering. My workflow is attached. However, I want to upload a whole bunch of documents (I dont need document upload) and all answers have to be strictly be limited to the contents of the files. I dont know how to do this I tried a few times but failed miserably.

you will see my attempt in the second workflow without attachment but it is picking up one collection in the vectore store. It is reading it but I get

<|python_tag|>tool call: Fundamental Objectiv<
I can see the vector store is getting some data and the embeddings, but the ollama chat against that is showing some generic responses that are not specific to the document even though i have set temp to 0. Help!

has anyone done something similar that they can share. I am on a windows 11 pc with Nvidia

[n8n]

It looks like your topic is missing some important information. Could you provide the following if applicable.

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

[rpeter]

• n8n version: Version: 1.72.1
• Ollama version ollama version is 0.5.4-0-g2ddc32d-dirty
• **Database (default: SQLite): qdrant Version: 1.12.5, build: 27260abd
• n8n EXECUTIONS_PROCESS setting (default: own, main): own?
• Running n8n via (Docker, npm, n8n cloud, desktop app): docker
• Operating system: windows 11

[Garry_Jakubiak]

Not a pro too but here what i did (All web based) and it’s working fine. Maybe it can help you.

On the top: load files into vector store (Supabase for meQuadrant for you)

• Many nodes but as a summary: it connects to a google drive folder and load every files into vector database
• It manages xlsx and text files. For xlsx, if you use clear and clean database format, the AI will find data easier than messy xlsx.
• I give an ID to every documents, so that it’s easier to filter, find them, … I extract the ID from the doc using “Set File ID” node then i use it while injecting into the vector database.

On the bottom: just an AI agent that read the vector database