This directory contains some loosely related experiments that attempt to implement various ideas. These are small, personal single-author software experiments. They all pursue a common goal, using similar techniques. They are called "experimental" because the author (me, Linas) does not wish to commit to any specific software framework or architecture.
That is, these are experiments in creating software frameworks and architectures; these are path-finders, and may perhaps someday lead to a robust, usable system.
The Experiments, in chronological order:
The 01-pair-agent directory contains a prototype for an agent that could process text files and perform word-pair-counting on the text therein. It was intended to be a dynamic replacement for the static batch text-processing pipeline in the learn project.
Status: Created in Dec 2024; abandoned. The file-system crawler code was moved to (implemented in) the motor project.
Critique: It's not entirely "pure Atomese"; the Atomese remains glued together with scheme functions.
The 02-memory directory contains (will contain) an experiment that tries to combine an LLM API with the basic idea of Retrieval-augumented generation (RAG) of vector systems, with a sheaf-theoretic IDL (interface descrition langauge) of the pure Atomese algos.