Open-source · AGPL-3.0
AI on your documents.
On your server. Nowhere else.
Upload your documents, ask questions in plain language, get answers grounded in what those documents actually say — with sources cited. One binary on your own server. No Docker, no API keys, nothing leaves your network.
What is actually in the request path
Sovereignty is about where the data goes, not where the model was born.
Three open-source components stand between your question and its answer, and all three run on your own hardware. The model weights are files on your disk — you can read the licence, swap the model, or run the whole thing with the network cable pulled out.
No US cloud, no API key, no vendor in the middle
Why I3K RAG Enterprise
Built for organizations that can't outsource their data.
Most RAG platforms assume you're happy sending your documents to a US cloud. We assume you're not — and we engineer accordingly.
EU-sovereign by design
Servers, data, models and the engineering team are all in the European Union. No transatlantic data transfers, no Schrems II exposure.
Backups that restore consistently
A scheduled daily archive holds the application database and a vector-store snapshot taken together, so the pair is never out of step. Restore over a running install from the admin panel. Archives are local files — nothing is uploaded anywhere.
Runs on the hardware you have
An NVIDIA GPU is detected and used automatically when present; without one it runs on CPU. Same binary either way — no separate GPU build to download, no vendor lock-in.
Multilingual out of the box
BAAI/bge-m3 embeddings cover 100+ languages, so an Italian question finds the answer in an English contract. No per-language fine-tuning, no separate index.
One binary, no orchestration
Extract the tarball and run it. It starts the vector database and the inference engine itself — no Docker, no Compose, no Java. After the first run it needs no network at all.
GDPR & EU AI Act ready
JWT auth, RBAC with 3 roles, audit log and retention policies. AGPL-3.0 source — auditable end to end.
How it works
The RAG pipeline in 4 steps.
Ingest, embed, retrieve, generate. Every stage runs locally. No data ever leaves your infrastructure.
- 01
Ingest
Upload from the web UI or the REST API. PDF, DOCX, XLSX, HTML, TXT, Markdown and CSV. Scanned pages are detected automatically and passed through OCR in Italian and English.
Tika · OCR
- 02
Embed & store
Text is split into overlapping chunks and embedded with BAAI/bge-m3 (1024 dimensions, 100+ languages), running in-process. Vectors go into Qdrant with the metadata used for role-based filtering.
bge-m3 · Qdrant
- 03
Retrieve
The question is embedded the same way and answered from the closest passages. Relevance threshold and top-K are configurable, and role filtering is applied here — a query never reaches the model with passages the caller may not see.
I3K orchestrator
- 04
Generate
The passages go to the local language model, which streams an answer token by token with each source shown. The model runs on your hardware. Zero external calls.
EuLLM · Mistral 7B
100+ Languages understood
1 Binary to deploy
0 Calls to external APIs
100% Local
Compare
I3K RAG Enterprise vs the alternatives.
We've tried to be fair. If we've got something wrong about a competitor, tell us.
| Capability | I3K RAG Enterprise | Onyx | Glean | Cohere North |
|---|---|---|---|---|
| Open-source core | Yes | Yes | No | No |
| EU-sovereign LLM engine included (EuLLM) | Yes | No | No | No |
| 100% self-hosted | Yes | Yes | Hybrid | No |
| Air-gapped deployment | Yes | Partial | No | No |
| EU-resident development & data | Yes | No | No | No |
| Consistent scheduled backup + restore | Yes | No | N/A | N/A |
| Runs without Docker or Kubernetes | Yes | Limited | N/A | N/A |
| Multilingual (100+ languages, bge-m3) | Yes | EN-first | EN-first | Yes |
| GDPR + EU AI Act tooling | Yes | DIY | Add-on | Add-on |
| No vendor lock-in | Yes | Yes | No | No |
Comparison based on publicly available documentation as of 2026. Vendor capabilities evolve — verify against the latest releases.
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.