- From: Danny Ayers <danny.ayers@gmail.com>
- Date: Sat, 16 Aug 2025 20:49:00 +0200
- To: Semantic Web <semantic-web@w3.org>
- Message-ID: <CAM=Pv=Q7iFGehWb7tf9umUzUBAisyFrH+JGdHEq42scGCddruQ@mail.gmail.com>
So many of the things I wanted to do with code were of a pipeline shape I ended up writing a thing. The workflow "transmission" is defined in a Turtle syntax list. Written in node.js. It is kinda DSL, and I have played fast and loose with namespaces. A Transmission is a series of Processors. A message - a JSON object gets passed between them. (Which sometimes contains an RDF graph, but of course). Event driven. So for my ridiculously over-engineered static site builder for my current blog, a filewalker triggers events for each markdown file it sees and passes them along, spawning a chain of processors for each. (I did start a graph-shaped GUI to wire things up, but this behaviour from event-driven made it kinda useless. They just spawn on their own). An observation I had along the way was that GPTs are good at small things, hopeless with big ones. a Processor in this system just takes a JSON object, does process(message) and passes it along. I'm using it for stuff, I doubt anyone else would want to. But I think the use of Turtle as a DSL is kinda interesting. It has been a lot of fun. The processor interface is so trivial, have to try things. Somewhere in there I have my own Lisp-like language, a reasoning chain based on EYE, etc. Cheers, Danny. https://github.com/danja/transmissions -- ---- https://danny.ayers.name <http://hyperdata.it/danja>
Received on Saturday, 16 August 2025 18:49:22 UTC