Attendance

Attendees:

• Rolf Krahl
• Andy Gotz
• Louise Davies
• Kevin Phipps
• Rodrigo Cabezas Quirós
• Santhosh Anandarama
• Patrick Austin
• Allan Pinto
• Alex de Maria
• Alan Kyffin
• Marjolaine Bodin

Agenda

F2F

Kevin to present info re: F2F and will send slides around after to those who couldn't attend.

Dates: 10-11th February 2026, The Cosener's House @ Abingdon

Went through the current attendee list and got some confirmations, some still to confirm

Went through proposed schedule - just initial thoughts, open to discussion/feedback

 

AG: cost for food?

KP: everything else is included/paid by us, you just need to pay for travel & hotel

 

RK: Might be possible to travel by train

PA: shouldn't take too much longer to travel from central london to Abingdon than from Heathrow

 

KP: ideas of things for the agenda? talks?

RK: Mojeeb to potentially present SEPIA. Would need invitation letter for visa purposes.

RK: Status of ICAT server. 6.2 soonish. February might see 7.0? Since most of these schema changes for 7.0 are coming from us I might be able to talk a little beit about how to upgrade your schema.

KP: Yes that would be nice.

 

AG: Will have a new co-worker by then. She will evaluate FAIRness. Thinking of including FAIR metrics for display. Wonder if that's interesting. Look at what's out there, if anyone has experiences?

RK: Measuring FAIRness is controversial...

AG: Not to say anything unilaterally. Common sense, see what's out there.

RK: nothing against presenting it, problem I have with tools that measure FAIRness, you need to define it before you measure it...

 

AG: PANfinder - replacement of PAN search API. ISIS is not connected?

LD: never connected to the old one as it did not support ISIS's volume of data.

AG: this is a new thing. Don't know if the ALC has anything. Are ALC doing anything with AI for data?

KP: There are a lot. Don't have visibility.

AG: Could Paul Quinn give a presentation on it?

 

AG: partial downloads - aria2 download tool. IDS only quite basic.

RK: yes, agree about the IDS, but IDS makes ZIPs on the fly.

AG: can show you logs, for us downloading is still an issue

 

AG: ideas document for F2F? KP: yes, will do

Site Updates

HZB

No updates

ESRF

MB: Everything ispyb is in ICAT & data portal now, will shut down old one in March. Everything working well so far.

MB: Still populating both systems currently as a redundancy, but IspyB will not be usable by users from March.

ISIS

Nothing major, automating certificate renewals. Still due to deploy 6.1.1

DLS

Nothing in particular of interest to the collaboration

ALBA

Long time since last update. Beginning of the year upgraded to ICAT 6, now on 6.1.1

Migrated from MySQL to MariaDB due to performance issues.

2 ICAT related projects.

K8s deployment / ICAT Helm chart: Rebuilt our ICAT infrastructure deployed in Kubernetes. Developed a Helm chart for easy deployment of all ICAT core components. Used to have a single container running all services. Now we have microservices like architecture, each component can be individually monitored, scaled or worked on without any downtime. Can present Helm chart at the F2F and see if there's interest.

ICAT data ingestion system: We're rebuilding our data ingestion system. Alex de Maria set us up with the ingester they are using a long time ago. Now we're rebuilding the system using different technologies (Python, changing message broker from ActiveMQ to RabbitMQ). A lots of processes that were independent from ICAT, have been now built into the new ingestion system. Can present at F2F.

4 beamlines working fully in the catalpogue, hopefully 3 more by end of the year.

KP: how many beamlines total?

RCQ: 12 or 13 total

KP: so by end of year large portion integrated!

LNLS/SIRIUS

Migrated to new data portal. Meeting 2 months ago decided to move to DRAC. Have both old and new one running, fully migrated to new one.

Integrating bluesky system, beamlines can write nexus files on our storage, hopefully by end of year allow them to ingest data into ICAT.

AG: users using interface or just in test?

AP: ready and running in our infrastructue, still need to set up e.g. URL, but it's integrated into our stack

For specific beamline, need specific features for sample tracking, moving work onto new infrastructure without changing things.

Other labs, integrating data management. Presented ICAT, trying to decide if ICAT fits other labs.

AG: new visualisation for plotting, separated from data portal as a separate service

AP: not running h5viewer anymore?

AG: yes, but as a separate service now

Other labs have same infrastructure? Hard to have data management system that fits all, and our team is quite small.

AG: same reason why we moved ispyb onto one system, only 2 people to maintain it.

AP: They changed parts of the system. ICAT is ready for integrating with ?MX2b?

Component Updates

icat.lucene 4.0.0

PA: Don't know if it was mentioned last time. Better support for datafile location field. Change default text parsing to handle file paths more intelligently.

icat.server 6.2.0

RK: Is this ready?

AK: can release it whenever

KP: there's already a snapshot right? if no one's found a problem them

AOB

LLMs

AG: Playing with LLM on top of metadata. Summarising metadata for datasets.

KP: Not yet, it has been mentioned. There's a related project round the corner. Wondered if there might be applications for ICAT.

RK: PANfinder is related

AG: What is the best sample? What was the diffraction? then gives a human readable response. Enthusiasm, but from a very small group of people (/person). Has potetial.

AP: Last year, did some small dev with a student here. Very very very small development on top of set of metadata. Very small metadata. Develop this kind of thing.

AG: need as much metadata as possible...

AP: Pull csv from website. Very interesting.

LD: Couple of years ago AGB wanted a project in collaboration with our ML group. Recommending related investigations. Didn't go anywhere and predicated on good metadata. Need to improve metadata.

AG: Hoping it will motivate people.

LD: In order have the dessert (AI) need to eat your meat/veg (improve metadata)

AG: Local LLM. gave a brief demo, Alex put it together quickly and hasn't find tuned.

PA: ISIS has open data, floated idea to DLS but no open data, having public metadata that you can then train on killed potential for DLS. Fine tuning, as more data is open, model has more data

AG: just hiring someone, would be good for them to come to F2F