@katy-sadowski @Gowtham_Rao — thank you both. This is exactly the informed perspective I was asking for, and rather than just say it landed, I’d like to show you what it changed.
First, though: to the people I tagged in my original post — that was poor form on my part. Calling individuals out by name in a public thread and pressing them for responses put people in an unfair position, and I apologize.
Now the substance. Your two replies, together with candid feedback I’ve received since posting, led to decisions I want to make publicly, here:
Parthenon will be free and open source — all of it, permanently, under Apache 2.0. The code is public at GitHub - Acumenus-Data-Sciences/Parthenon: Parthenon unifies the OHDSI ecosystem (Atlas, Achilles, Usagi, WhiteRabbit, and more) into a single AI-powered app — replacing two decades of language sprawl with one Docker Compose stack. · GitHub, and two concrete commitments both land by the end of this month. First, we are retiring the separate commercial Enterprise Edition and re-integrating it into the Community Edition — SSO/SAML, multi-tenancy, audit retention, and the Kubernetes operator will all live in the open repository, and the private EE repo will be archived. Second, we are relicensing the platform from AGPL-3.0 to Apache 2.0 — the same license Atlas, WebAPI, and the rest of the OHDSI stack carry — so no copyleft clause stands between Parthenon and any institution’s IT policy.
No paid tiers, and no “for those who can AFFORD it” edition — I wrote that line in my original post, and I retract it. Acumenus’s only commercial activity going forward will be optional services: implementation, deployment, and long-term support for institutions that want a partner rather than a product. Gowtham, that is the pattern you described as the one that actually succeeds in this community: services rather than proprietary platforms. The honest conclusion from your comment was to stop swimming against it.
Katy, your three points stand and will shape how we operate.
Modularity: Parthenon runs as roughly eighteen composable Docker services rather than one binary, and it maintains strict compatibility with the ecosystem: Atlas cohort definitions and concept sets import and export without modification, and a WebAPI-compatible layer supports legacy R package integration. We replaced runtimes, not semantics. But we’ll demonstrate modularity rather than assert it — we will carve out and publish individual components, starting with Darkstar, our rebuilt OHDSI R runtime (already a discrete service in the stack), as independently deployable pieces a team like yours could adopt à la carte without ever touching the rest.
Resourcing: We will not expect maintainers to donate scarce volunteer hours to our roadmap, and if contributors emerge from the user base, we’ll support them properly — there’s a CONTRIBUTING.md with our quality gates waiting for them, and with the move to Apache 2.0, contribution carries no copyleft strings.
The TAB pathway: the offer stands, and matters even more now. A structured intake and evaluation pathway would give projects like ours a legitimate front door and protect the community from the noise Gowtham describes. Acumenus will commit real engineering time to that effort, and Parthenon volunteers as a test case for whatever process the TAB designs.
What we’re asking for, and how to engage: serious beta testers across the world and across industries: AMCs, pharma, registries, public health networks, willing to run Parthenon against their own OMOP instance and tell us bluntly where it breaks, what’s redundant with your existing stack, and what’s missing. Structured, time-bounded evaluation, not volunteer development. A live demo, the full user manual, and the API reference are at parthenon.acumenus.net; a one-line bootstrap installs the latest tagged release. Issues and pull requests are open on the repo, and I’d welcome the hardest code review this community can offer.
For real-time help such as installs, troubleshooting, study questions; we run an open Discord at OHDSI (Acumenus.io), where we’ve been supporting users around the world; anything of community-wide interest, I’ll keep here on the forums. If any workgroup or community call would find a walkthrough useful, I’ll gladly present and take questions. Otherwise, reply here or message me directly.
On differentiation, I’ll keep offering artifacts rather than adjectives: twelve public releases with full CI, the IRSF natural-history conversion now supporting three pharma partners’ FDA-facing programs, the attached studies produced end-to-end on the platform, and code anyone here can read. We have successfully tested a complex Hypertension research protocol, and will be testing an Obesity protocol in the coming weeks. One design note relevant to the IT constraints Katy raised: Abby, our assistant, defaults to the openly available MedGemma running on a local Ollama stack, so the platform can run fully on-prem with no patient-level data leaving the institution.
Thank you both again. This thread changed our direction — and I believe for the better. I’ll report back here when the re-integration and relicensing are complete.
Sanjay M. Udoshi MD
Founder - https://www.acumenus.io
