Semem

a new subproject seeded

The stuff below got messed up.

Recovered enough to continue, but to clear my head, a little timesaver -

hmm, wait - need a cat of my nginx sites-enabled...

There are #:todo bits I need to pull out of the past few day's entries here to go in their correct todo areas...

Today, something like :

  1. #:todo #:semem #:spike
  2. #:todo /home/danny/github-danny/hyperdata/workspaces/transmissions/todo/next-steps.md
  3. #:todo Verbo photo
  4. #:todo carving prep
  5. #:todo 1-page docs (Transmissions urgent)

I need to be keeping ClaudioB informed of what I'm working on, so around those #:transmissions next steps I need to bear that in mind. Also notes on AI tools I'm using/trying :

  • OpenAI API
  • Claude projects + API
  • Mistral API
  • cline
  • aider
  • []

etc.

Grr, lost a lot of time with node module stupidity.

#:todo add rule to #:um - failure is part of the process, but a goal is that things shouldn't fail more than once for the same reason

A thought crossed my mind on dogwalk, must-have base functions for processors :

message = expandLD(message, options); // enlarge the Logical Dimension/Linked Data
message = contractLD(message, options);
message = expandLM(message, options); // enlarge the Language Model dimension
message = contractLM(message, options);

Semem

#:semem

This morning Google's 'Discover' (on Chrome on Android) shortcircuited my morning reading in bed with a link to this from MarkTechPost : Meet Memoripy: A Python Library that Brings Real Memory Capabilities to AI Applications which led me to this on GitHub :

Memoripy is a Python library designed to manage and retrieve context-aware memory interactions using both short-term and long-term storage. It supports AI-driven applications requiring memory management, with compatibility for OpenAI and Ollama APIs. Features include contextual memory retrieval, memory decay and reinforcement, hierarchical clustering, and graph-based associations.

Long story short, it gives me a nice inroad into the practicals of the hybrid SPARQL+vector store bits high on my todo list.

First job, give this a place of its own. I've got lots of material related to the hybrid store scattered around, best I make an anchor. I've made a GitHub repo, now pop that in here :

cd /home/danny/github-danny/hyperdata/packages
git submodule add https://github.com/danja/semem.git semem

I'll put dev notes & docs as they emerge over on #:hyperdata.

First I need a plan, then port Memoripy from Python into Javascript.

Semem