Hillock

I don’t remember where I found it but this looks like an interesting project to explore for my RAG tests.

https://github.com/roandejager/Hillock

Installation is easy but with Claude Code… I think I can make it even easier!

Seems to have found a couple of issues during the install. Not sure if it’s him or me.

Here’s me being stuffed again by the GPU in my mini-PC…

Oh wow… Sometimes Opus really nails it with the investigations…

Running with qwen2.5:1.5b for now.

Let’s try with AnythingLLM…

We’re testing Hillock here so option 2

In the AnythingLLM workspace, click the LLM Provider selector…

Search Generic..

I used these settings…

Other fields like this:

Problemita…

Claude Code’s response…

Repeat the steps and all seems good.

Answered the question it ‘knows’ and not the one where it doesn’t have the data in the demo facts.

Let’s try with one of my own client’s documents…

Are we there yet?

Interesting.

And the results are in…

in AnythingLLM…

and…

Compared with the results from using Sonnet…

(from here: https://mcsomething.dev/2026/09/15/pdf-natural-language-search/)

So it seems at the moment (and possibly by design) that Hillock is a better fit for fact heavy material like people, dates etc. From the repository, the Goal for v1, production is “Goal: A stable, packaged, and highly reliable memory engine ready for researchers and privacy-first local AI applications.“

I’m following the project and will do another write up later.

Update: v0.8 is out so let’s take a look in the most lazy way possible…

Oh dear.. My unsupported GPU again…

More laziness but this is what Claude Code is good at, running repeated tests with different setups and giving opinions and results…

I’m not sure anything can be read into the results because my machine just doesn’t have the specs for this…