BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.com//open-source-conference-luxembourg-2026//ta
 lk//3Q8XYD
BEGIN:VTIMEZONE
TZID:Europe/Luxembourg
BEGIN:DAYLIGHT
DTSTART:20251007T000000
TZNAME:CEST
TZOFFSETFROM:+0200
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251026T030000
RDATE:20261025T030000
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T030000
RDATE:20270328T030000
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:Sovereign RAG on Your Own Repository: Letting the Search Engine En
 force Permissions - Angel Borroy
DTSTART;TZID=Europe/Luxembourg:20261007T143000
DTEND;TZID=Europe/Luxembourg:20261007T150000
DTSTAMP:20260916T152615Z
UID:pretalx-open-source-conference-luxembourg-2026-3Q8XYD@pretalx.com
DESCRIPTION:Enterprise AI has a permission problem. Point a RAG assistant 
 at a document repository and it will happily quote a contract the user was
  never allowed to open. The usual fix is to ship everything to a hyperscal
 er and trust an application-layer filter. For a hospital\, a ministry or a
  court\, neither half of that is acceptable.\n\nFifteen years of open sour
 ce search inside a large content platform\, and where it landed: a stack t
 hat runs on your own infrastructure\, where the search engine itself decid
 es who may read what.\n\nThree eras\, three permission models. Alfresco st
 arted on a patched Apache Solr fork with custom Lucene scorer classes and 
 a separate ACL document type joined by hand at query time\, so every Solr 
 upgrade meant editing search-engine code. Moving to Elasticsearch and then
  OpenSearch replaced all of it with a standard client and a denormalised r
 eader/denied filter. Today OpenSearch Document-Level Security pushes enfor
 cement below the application entirely: one role\, a terms lookup query\, a
 nd no query the application can write will leak a document.\n\nWe then app
 ly that to working retrieval-augmented generation over existing repositori
 es. Live demo: two users\, one question\, two different answers\, embeddin
 gs and language model running locally\, nothing leaving the building.\n\nY
 ou will leave with three patterns that transfer to any content source: enf
 orce authorization in the storage layer\, keep retrieval hybrid rather tha
 n vector-only\, and make ingestion idempotent so batch and live sync can r
 un together.
LOCATION:Artificial Intelligence
URL:https://pretalx.com/open-source-conference-luxembourg-2026/talk/3Q8XYD
 /
END:VEVENT
END:VCALENDAR