Knowlytix 1.0: The Geometric Memory System Is Now Stable

Today we’re cutting Knowlytix 1.0. The Geometric Memory System family — the deterministic layer that sits beneath an agent and makes its actions bounded, checkable and reproducible — is now a stable release on PyPI, with a settled API you can build on.

Since the beta went up, the family has been functional and installable but still moving: the surface was settling, and we told you to pin a version if you needed a reproducible build. That caveat is gone. 1.0 means the API is stable — the shape you install today is the shape you can depend on.

What’s in the release

Five packages ship at 1.0, each versioned on its own:

Package	Purpose
knowlytix-core	Geometric memory core (geometry + LLM-layer foundations)
knowlytix-harness	Policy-governance harness
knowlytix-knowledge	Geometric expert system with LLM-augmented learning
knowlytix-benchmark	Benchmarks and reference results
knowlytix-kal	Knowledge Adapter Layer for GMS

You don’t install those one at a time unless you want to. knowlytix is a meta-package — a single name that pulls the whole family in one command:

pip install knowlytix

That’s it. One line and you have all five in your environment.

Requires Python 3.12. That’s the version the family is built and tested against. On an older interpreter pip reports “No matching distribution found for knowlytix” — the version gate doing its job, not a missing release. Check with python --version before you install.

Run it on your own data, free

Once it’s installed, demo mode runs immediately on the built-in datasets — no key needed. To point it at your own data, claim a free 180-day developer key on the signup page. It’s generated instantly, right in the browser — no credit card, no sales call. Set it as an environment variable or drop it in ~/.knowlytix/license.key and you’re verifying your own agents against your own data.

What 1.0 changes for you

• A settled API. The interfaces are stable. Code you write against 1.0 keeps working — no chasing a moving target between releases.
• Still one command. The meta-package tracks the latest of each sibling, so pip install knowlytix always gives you the current family without us shipping a new meta release to bump pins.
• Production status, stated plainly. Every package now carries the Production/Stable classifier on PyPI. This is software we intend you to run in real workflows.

Where to go next

The two tools that sit on top of the Geometric Memory System — ForgeLoop to build governed agents and ProofLoop to prove them — install the same way, right alongside the family:

pip install knowlytix
pip install forgeloop
pip install proofloop

More on the approach, the benchmarks and the research is on the platform page. If you hit any trouble installing, or you want to tell us what you’re building, email us at hello@knowlytix.ai.

Thanks to everyone who ran the beta and sent feedback — it’s in this release. Now go build something you can defend.