It is now 2026-01-21. The date of my (Dirk Roorda's) retirement is 2026-03-01.
I consider my work on Text-Fabric as finished. I'm not saying Text-Fabric itself is finished, though.
Cody Kingham, who influenced the development of Text-Fabric, especially the capabilities of its query language, has found an optimization in the way the TF data is represented in the memory of the computer: by means of memory-mapped Numpy arrays. This is a huge efficiency gain in memory footprint, without compromising the time performance.
At one stroke, this widens the scenarios in which Text-Fabric can be deployed. It opens up an interface between Text-Fabric and AI.
Cody chose to take the TF core API and reimplement it with his new data representation.
The result is Context-Fabric, characterized as production-ready corpus analysis for the age of AI, which is born out by the fact that it comes with an optional connector to AI servers.
It was with great pleasure that I received word of this new twist from Cody, a few weeks before my retirement.
Although Context-Fabric does not incorporate the so-called advanced API of Text-Fabric, I am convinced that in time it will move miles beyond my Text-Fabric, because of the integration with AI.
The advanced API of Text-Fabric offers you easy corpus downloading, and sophisticated display of query results and corpus fragments. If you rely on that, Text-Fabric is still the way to go, but that might change, because Cody already showed that with a little prompting, an AI can also give you nice displays of snippets of the corpus.
The future as I see it, will have a by now classic Text-Fabric, that will not develop further. I will still respond to issues and fix bugs and change dependencies if the need arises.
But the cutting edge is Context-Fabric, and if the promises are made true, it will overshadow Text-Fabric sooner rather than later.
An obvious thing to do could be to also implement the advanced API in Context-Fabric, but that is up to the discretion of Cody.
Alternatively, someone could replace the current data representation in Text-Fabric by the much more efficient representation in Context-Fabric. But I am not the one who will do it. If someone else feels attracted to it, I will lend the support that is needed, but without diving deep in the code myself. Keep in mind though, that it might be a spurious effort. A core Text-Fabric + AI (= Context-Fabric) might progress better and faster than an optimized but classic TF!
I'll keep an eye on the two of you: Text-Fabric and Context-Fabric.