2024-05-10 Partner Call

Samvera Partners Call

Friday, May 10th, 2024

11:30 am | Eastern Daylight Time (New York, GMT-05:00) | 1 hr

Join Zoom Meeting Link:

https://us06web.zoom.us/j/81973313501?pwd=dXVlcHZOODF2c2NiNGpoNEFad3NZQT09

Join by phone

Dial by your location
+1 646 931 3860 US
+1 646 558 8656 US (New York)
+1 253 215 8782 US (Tacoma)
+1 346 248 7799 US (Houston)
Meeting ID: 819 7331 3501
Passcode: 603236
Find your local number: https://us06web.zoom.us/u/kkbAJ584n

Code of Conduct

We want Samvera Community calls to be fun, informative, engaging events for all our partners and participants.

We encourage everyone to apply the Samvera community principles of openness, inquiry, and respect in their interactions at the event.
Please review the Code of Conduct and Anti-Harassment Policy.

If you have any questions or concerns, please feel free to reach out to community helpers.

Facilitator: @Heather Greer Klein

Note Taker: @Kevin Kochanski

Attendees

@Robin Lindley Ruggaber
@Karen Cariani
@Margaret Mellinger
@Collin Brittle
@Nicholas Mark Homenda
@Nora Zimmerman
@Kevin Kochanski
@Paul Walk
@Chris Awre
@Jill Morris
@Thomas Scherz
@Esmé Cowles
@Jon Dunn
@Carolyn Caizzi
@Emily Lynema
@david.schober
@Glen Horton
@Alberto Martinez

Agenda & Notes

Any other items for the agenda?
1. Partner: Welcome Paul from AntLeaf, new Partners
2. Board Election: Possibly two candidates; will do a vote when that is finalized
Samvera Virtual Connect reflections
1. 120 Registered, with 70-85 participating live at any given time
2. Zoom - may move from Webinar to standard Zoom to encourage participation
3. Videos should be up on YouTube by the end of today
4. Thoughts:
  1. Good mix of presentations
  2. Good duration/scope
5. Please remember to fill out the feedback survey that Heather sent - https://forms.gle/QEeDibu4wmSRhXZb9
Discussion: AI and LLM (@david.schober @Brendan Quinn )
1. Vector databases - turn letters into numbers in a three dimensional space. Requires a special kind of database to do this, elastic search can do this now. Semantic relationships represented as numbers. See James Griffin’s presentation at Virtual Connect for the details.
2. Transformer-Based Language Models (LMs) - allows you to take words, pictures, etc. and embed it in a language model
3. Vector Search - similar to Transformer-Based, we embed in a model and ask “How close are these to the question asked?”
4. Retrieval-Augmented Generation - Takes the question, puts it through a vector search, and packets it together - Answers and questions. The system is given a prompt, which is akin to a persona. It answers with a specific tone and providing specific augmented information.
5. Northwestern Built:
  1. Infrastructure - AWS Cloud
  2. Chatbot
  3. Sharing best practices
    1. Scrolling - from a design prospective, important for the user to know that text is being generated.
    2. Can toggle to public works only, limit to image audio or video
    3. Gives LLM response about what it doesn’t know, and what the user needs to look at and documents to start for finding the answers. It knows enough from metadata to know what documents might have the requested information. Indexing metadata, grouped into archival collections.
    4. Pulling from elastic search vector search.
    5. Multi-lingual, 150 languages. Even with English-only metadata, responded cleanly in Italian. Colmex finds the quality of English results is better than Spanish; NW hasn’t yet evaluated.
    6. Rough estimate for cost at launch is between $300-$500 a month with the current chatbot, limiting to logged in users to start. NW is currently evaluating the quality of response vs. cost to determine the right level (cost-benefit analysis).
    7. Team is 6 including David. 2 Senior Devs, 1 Lead and 2 front end devs
    8. NU applied for an IMLS grant for developing a tool for metadata description, transcribing handwritten text, element description.
Regional Meetings Updates and Planning
1. Midwest - Indiana University and PALNI (Private Academic Library Network of Indiana) are tentatively planning to host a Samvera Midwest Regional Meeting in Indianapolis, Indiana on September 25-26, 2024.
  In order to gauge interest and help ensure that we are structuring the meeting to best meet the needs of Samvera community members in the Midwest, if you’re interested in this meeting, we would appreciate it if you could complete the following brief survey by Friday, May 17: https://forms.gle/Nq6MFuurwnmAtfU77
2. Northeast - Princeton in October, save the date coming soon. Three small events in one week: Samvera, devops, and Blacklight. Samvera Regional event on 10/14-15, DevOps4Lib on 10/16-17, and a Blacklight on 10/18
3. West Coast - Chrissy at UCSB and Kevin at SoftServ met for an in initial viability conversation. Likely to host in San Diego or Portland, but will keep the community updated when we assess interest in hosting hopefully by the end of next week (May 17). This could happen as soon as August and hopefully before the end of October. Next steps are to contact Western US/Canada institutions to assess level of interest and gather broader input.
4. Europe - Chris is investigating
Other Updates:
1. Sign up to present or facilitate a topic (5 - 30 minutes) at a future Partner call. One requested topic without a facilitator is around hiring challenges and approaches.
2. Hyrax and Hyku development updates - see Virtual Connect presentations
3. Developer Congress May 13-17
4. Wishing Carolyn Caizzi well in new position at Harvard!

Anything for the Samvera Board? (Standing item)

Date of next call: June 14th

Notetaker: @Chris Awre