Self Hosted AI Starter Kit

Questions
1

Can we please get instructions for setting this up without Docker?

I don’t use Docker and I won’t install Docker on my system as it adds one more potential weak link to be exploited. I would rather run the individual components (and then I also don’t have to mess around with workarounds to use Apple Silicon GPU in Docker).

I already have Ollama running locally, it would be useful to get configuration and installation steps for n8n, qdrant and postgres. Can these just be setup directly and then change the configuration for the APIs in n8n or is there extra configurations needed to get everything to talk to each other?

Many thanks.

2

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:
4

Hi @ThatPrivacyGuy

Thanks for posting here and welcome to the community! :cake:

Our self-hosted AI starter kit is just a convenient way to get started. It is not supposed to be the final setup for anybody, as there are hundreds different ways to host and we cannot support every edge case.

The kit itself is just a compose file. If you don’t want to use docker, you can install all the services listed in the compose file manually and configure them on their own.

You can check the official documentation of running qdrant and postgres locally.

Feel free to share your experience with the community! :slight_smile:

3 Likes
6

Thank you that is very helpful.

1 Like
7

I have managed to get everything installed and have n8n successfully integrating with Ollama and Postgres - the issue is Qdrant.

To avoid using Docker, I followed the Qdrant instructions for building a binary, which I have managed to complete and it is running - unfortunately the binary build does not come with /dashboard support (not sure why) so I installed another project called qdrant-dashboard which does successfully connect to qdrant and is able to run commands, view collections, create collections and has the qdrant console.

What it doesn’t do is provide any obvious way to generate an API key which means I cannot integrate Qdrant into my n8n workflow as it wants an API key.

Any idea how I can get around this issue?

8

Hi @ThatPrivacyGuy

Thanks for the feedback! That’s great to hear :raised_hands:

In regards to the Qdrant API key, I think you will need to check that with the Qdrant support

9

No it seems there is a bug with n8n integration as an API key is not needed for qdrant running locally via binary but n8n integration will not allow me to setup an integration without an API key.

I have spent the weekend in the qdrant discord trying to trouble shoot this and have confirmed that no API key is necessary.

10

Hi,
I have the similar issue. I’m not a programmer and happy to install everything with conda and pip. Is there a documentation around this topic?
Many thanks
Joerg

11

OK I have managed to fix this. It seems that n8n does not allow URLs to be localhost and instead you have to use http://127.0.0.1:6333 for n8n to be able to connect to it.

I also managed to get qdrant working properly by creating symbolic links to the config/ and static/ folders in the repo, to my homedir because qdrant binary runs from your homedir by default so if you don’t have those folders accessible in the homedir, qdrant cannot find them.

So now I have full access to the qdrant web ui as well.

Hope this helps others trying to avoid using docker.