Search connected collections, then review the results with ARCHAI
Start with a material, place, period, technology, or theme. Search every connected source or focus on one institution. The public demo prioritises working, rights-aware images; the staff app can also interrogate metadata-only records.
Talk to the connected collection
Ask a question across the connected records. Choose a staff lens to shape how ARCHAI reads the material, then type, dictate, or listen back to the response through the browser audio interface. Responses are grounded in source metadata and do not represent the views of the contributing institutions.
Tap or scan an object
QR · NFC · hyperlink · map · spatial trigger
It answers from its own record
Grounded in Obtext — no invented facts. Source, attribution & legal status stay visible.
Read, ask, listen & share
and follow related objects across the collection.
AUXIO turns a collection record into a mobile visitor encounter
Open an object through QR, NFC, hyperlink, map, or spatial trigger. Visitors can read, ask, listen, zoom into IIIF source images where available, share, and follow related objects while the source record, attribution, and legal status remain visible.
ARCHAI is an open-source, sovereign semantic interface for cultural heritage — an active research toolkit designed to sit above existing CMS and DAMS infrastructure. Institutions can run semantic search, grounded collection interpretation, spatial tools, accessibility support, and visitor engagement on their own network and under their own governance.
The current research corpus contains 3,147 records across museum, heritage, public-art, library, archive, and cultural-data sources. Its public-art layer gives the demo a place for street art, murals, civic commissions, precinct works, and social-practice records alongside museum and archive material. The wider source network includes Library of Congress, Wikimedia Commons / Wikidata, Trove / National Library of Australia, Digital Public Library of America, and Internet Archive, bringing library, archive, civic-memory, open-media, born-digital, and linked-data context into the ARCHAI ecology. Staff can search the complete metadata layer; public interfaces use a separate rights and media-health gate so a connected record is not automatically treated as reusable visitor media.
AUXIO offers optional factual, third-person, and era-informed interpretation modes through NFC, QR, hyperlink, map, or location-aware exhibition triggers. The working audio interface supports microphone questions and spoken replies through browser speech services, opening the prototype to detected browser languages on supported devices, with further accessibility and multilingual testing in development.
The data format powering each conversation is Obtext (Object Context) — a structured record containing identity, physical properties, temporal context, significance, and crucially, constraints — what the object explicitly does not know.
ARCHAI remains an active research prototype. This website is a public-facing demo layer rather than the full operational app. FAMTEC is open to funded research partnerships, institutional pilots, accessibility testing, collection-data collaborations, development support, and evaluation with museum, gallery, archive, and university teams, and looks forward to collaborating, growing, learning, researching, and testing with partners across the wider GLAM sector.
Developed as part of doctoral research at RMIT University, School of Design. The research paper was accepted for the 6th Summit on New Media Art Archiving (ISEA2026 Dubai).
Sources & Rights
ARCHAI is presented as a 24-source research network spanning indexed live sources and onboarding connectors. The staff layer currently contains 3,147 indexed records; the public AUXIO layer publishes 1,402 image-backed pages that passed the current availability and rights gates. Every object retains its source, attribution, and legal status. Institutional or platform inclusion does not imply partnership or endorsement.
- Qdrant vector database (on-premises)
- Ollama local inference (Qwen 2.5 stack)
- nomic-embed-text embeddings (768-dim)
- AUXIO universal access interface
- Working TTS/STT audio interface · detecting browser languages
- Maps and spatial interpretation layer
- Obtext structured data format
- Five-layer hallucination prevention
- AI-moderated visitor comments
- Conversational collection search
- 24-source network · 3,147 indexed records
- 1,402 rights-gated AUXIO visitor pages
- Designed QR poster, access sticker & visitor postcard outputs on 440 cleared records
- SQLite persistent storage
- Zero cloud dependencies
- ~$3,500 USD semantic layer (Mac Studio)
Full staff test build with curatorial search, AUXIO management, vocabulary tools, and source-pipeline controls. Draft WIP for institutional testing.
Open full ARCHAI app WIP