Public sector
AI for public administration, on the authority's own infrastructure
Resolutions, decisions, circulars, regulations and records searchable in plain language, on the authority's servers — without entrusting citizens' data to external vendors.
For a public body the question is not only where the data ends up: it is also being able to explain, under audit, what the software does and which documents it answered from. A local, open-source system answers both.
The authority's memory is written down, but not searchable
An administration produces and keeps more records than any person can hold in mind: resolutions, decisions, regulations, circulars, opinions, specifications, minutes. It is all archived and all formally accessible — but finding a specific answer inside it means already knowing where to look.
The cost falls on front-desk and case-handling staff, and ultimately on response times to citizens. It is the case where document retrieval with cited sources shows a measurable benefit within the first days.
Why self-hosting is the shortest route for a public body
- Data stays on the authority's infrastructure: the question of transfer to non-EU providers never opens.
- The Community engine is open-source under AGPL-3.0: the code can be verified by anyone, which carries weight in public assessments.
- No per-user fee, so no cost that grows with internal adoption after award.
- The system works without a network after the first run, making it usable on separated segments too.
- No data lock-in: documents, index and configuration all stay on the authority's machines.
Regulation (EU) 2024/1689 on artificial intelligence has applied progressively since February 2025, on top of existing GDPR obligations. Which classification applies to a specific deployment has to be assessed case by case: we can support that analysis, but it is not a box a product can tick on your behalf.
Concrete applications
- Search across the body of resolutions and decisions, referencing the act and the passage.
- Support for case handling: finding the administrative precedent on a comparable matter.
- Regulations and internal circulars, with answers citing the version in force.
- Tenders and contracts: quickly checking requirements, deadlines and criteria across bulky documentation.
- Digitised historical archives: scanned records made searchable through OCR.
- Staff training and operational continuity when the person handling a procedure changes.
Frequently asked questions
- Can a public administration use this software?
- The Community edition is released under AGPL-3.0 and usable in production without a commercial agreement. For bodies that need a commercial licence, a support contract with defined response times and deployment assistance, there is the Pro edition. The administrative requirements specific to your procurement have to be checked against your own specification: we are happy to supply the technical documentation needed.
- Where is the data processed?
- On the authority's infrastructure. Documents, vector index, application database and language model all run on the same machine, and after the first run the system makes no outbound calls.
- How does this relate to AI Act obligations?
- Regulation (EU) 2024/1689 has applied progressively since February 2025, and classification depends on how you use the system rather than on the software itself: the same tool used to consult internal regulations or to affect a procedure concerning a person does not fall in the same category. The analysis has to be done on the intended use, and we can take part in it.
- Does adopting it require a tender?
- That depends on the amounts and on your administration's rules, not on us. The Community edition carries no licence fee, which makes a pilot straightforward; adopting Pro with support follows your ordinary procurement procedures.
Ready to run RAG on your own infrastructure?
Start with the open-source Community edition, or talk to us about Pro with structured extraction, SSO, audit log and SLA.