Facilitator: @Heather Greer Klein

Note Taker: @Chris Awre

Attendees

• @Karen Cariani

• @Annamarie Klose

• @Chris Awre

• @Paul Walk

• @Emily Lynema (Deactivated)

• @Kirsten Leonard

• @Margaret Mellinger

• @Chrissy Rissmeyer

Agenda & Notes

1. Additions to the agenda?

   1. Nick: potential talk on Product Management

   2. Nick: Tech Lead / PO roles - could they paid?

These items were deferred to a subsequent call.

2. Open Repositories 2025 - open discussion time for any reflections, particularly relevant presentations, etc.

   1. Nick: What got people excited?

   2. Nick: Reports back from the N8 team

   3. Links to presentations of note:

      1. It doesn't have to be this way: Reimagining Institutional Repositories in-Transition

      2. Bridging the Silos of Institutional Data Repositories: Community Collaboration and Cross-Institutional Development

      3. Embedding Accessibility into ETD Workflows: A Case Study

      4. An Integrated Open Ecosystem: Whose Responsibility Is it?

      5. Managing Machine Access to Open Repositories in the Age of Generative AI

      6. IIIF at one end, OCFL at the other, Fedora in the middle

      7. Better Together: Growing Digital Repositories through Community-building and Collaboration (CSU system-level Hyrax management)

      8. Impacts on the Repository of COAR Notify, and tools to help you

      9. Princeton team: Beyond the Buzzwords: agile collaboration and rejecting a perfectionist mindset (how we did it, and you can too!)

      10. Michael Klein: We Need to Chat: A presentation about real-world AI use cases

      11. Paul Walk: The IRD: Improving knowledge about the state of the repository system landscape through automated curation processes

      12. Paul: Ben Zhao's closing keynote, Dealing with Generative AI, Harms and Mitigation Techniques (slides and downloadable MP4 recording)

      13. Duke team: Three Repositories Walk into a Library: recapping 30 years of repository development at Duke University Libraries (investigating replacements for Hyrax)

      14. OCFL: An application-independent file layout for repositories

The presentations listed are Heather’s initial review of what was of main interest to Samvera, but everyone is welcome to add to these to surface what they found useful/interesting.

Discission centered on the themes brought up in the closing keynote by Ben Zhao, which addressed misunderstandings around generative AI and LLMs and what the repository world needed to take note of. Ben views generative AI LLMs as being near the limit of what they can be from continued training. There are still many applications that might be built on the models and LLMs, but the underpinning LLMs may not evolve much further. The main area of potential in repository areas is viewed as retrieval-augmented generation (RAG) as a means of enhancing discovery.

A sepaarte AI-related question: does the spend by AI companies compared with theior income stack up? Or is it a sub prime mortgage scenario waiting to happen? The concern over time is that the price points of usage will become untenable, but with a need to prepare for this in case changes are needed in how AI provision is resourced?

There had been lots of discussion about COART Notify and FAIR across thr campus, the former particularly welcome given there wasn’t a specific presentation about it. Another theme across presentations was the connection with the preservation layer.

3. Discussion: Trends among small library IRs; JSTOR offering competitive service integrations; potential next steps to align Hyku based on these trends (Jill)

Background: JSTOR are heavily pushing their Digital Stewardship Services across the Sector, which now includes what appears to be a repository offering as part of the package. Roger Schonfeld is the Product Owner. The offer seems to have originated as the successor to what was ArtSTOR Shared Shelf, a service that helped institutions make digital archival materials more visible (which then linked back to your repository; it is is this last point that has now changed). As well as providing that presentation and discovery capability, the offer now has repository capability behind it (of some sort). A key selling point is the AI driven metadata generation that is also offered, as well as JSTOR being a trusted non-profit service provider. The service is being pushed through collections staff, not technical staff, and there has been a lot of interest (and some adoptions): there is a need to re-build connections with collections colelagues around this.

Some members of PALCI have expressed interest and at least one has switched from Hyku to JSTOR. It is acknowledged that budgets are tight, staffing even tighter, and added value offering like the JSTOR one is appealing. There is a need to consider how Samvera addresses this in terms of both understanding the offer better and laying out why it may serve a need but that Hyku serves needs as well and, in many ways, serves them better. UC looked at the offer, tested the AI generation and made the case to use that part but keep the repository local. El Colegio de Mexico has been looking at a Hyku instance for institutions that cannot maintain their own repository, and has been looking at metadata generation using AI as part of that. They view the JSTOR offer as a blob approach, the same for everyone, but wihtout the granularity of need that systenms like their own would need. It will be useful to consider how Samvera might more broadly integrate with such commercial offerings so sites can get the best of both.

Concerns around the JSTOR offer are that the repository offer is not transparent, and the AI prompts being used to generate the metadata are not being shared, so there is little awareness of what may be biasing the metadata, and there has been no known validation of the quality of the results. Some libraries are seeing the ability to more quickly surface hidden collections as oversoming these shortcomings. Open solutions are hard to push for with restrained budgets, but there remains a good opportunity for Samvera to move into the space and devlop/provide trusted solutions that others can make use of. That will be discussed further at the Board in the first instance. Where there is interest in setting up a AI WG (Nick offered), that could also help foster thinking about how Samvera engages in that space.

4. Potential new process for reporting vulnerabilities, sharing with Partners and the wider community

Briefly discussed, Nick S. proposes a Slack-based way of sharing information about vulnerabilities rather than the current email route. Part of the latter involves Partners receiving information sooner (a bonus of Partnership), but this hasn’t been re-visited recently. The Board will discuss further.

5. Latest Updates on Samvera Connect 2025 Mexico City Samvera Connect 2025 Online Schedule & Directory

   1. Two workshops on flexible metadata, plus a Samvera 101 session

   2. Call for Proposals extended to July 28th

   3. Registration will open ASAP (working out Stripe payment account to accept payments for Sched ticketing)

   4. Hotel block information will be posted ASAP for making reservations

Not discussed, but key information is listed in the bullets.

6. Other Updates: 

   1. Sign up to present or facilitate a topic (5 - 30 minutes) at a 2025 Partner call - see list of ideas on the page

      1. Next month: Paul Walk, COAR Notify work

   2. October Developer Training Camp - seats still available! Tuition is free in exchange for one 2-week Samvera Community sprint

   3. Reminder: Recruiting for a Hyrax Tech Lead for 2026

      1. Daniel and Randall are cycling off at the end of 2025

   4. Hyku 7 Upgrade Community Fundraising - several institutions looking into contributions

      1. Nick: Question about potential expansions of scope, or otherwise separate fundraising
         - Spotlight “integration”
         - Fedora6 support with Valkyrie
         - Automated testing

Nick S. briefly highlighted the ideas for Hyku 7 and welcomed thoughts on what is proposed. Key is consideration of the topics discussed above and how future developments can help provide the one stop shop institutions need (and want).

Three Repositories Walk into a Library: recapping 30 years of repository development at Duke University Libraries

Anything for the Samvera Board? (Standing item) 
Date of next call: August 14th, 2025 - COAR Notify in Hyrax overview and discussion Notetaker for next call: