What is EuLLM and who builds it?

An open-source platform from I3K Technologies Srl in Italy for creating, distributing, and running sovereign, EU-compliant language models; it is tested on EuroHPC supercomputers and carries a Zenodo DOI and the website eullm.eu.

How does EuLLM relate to Jev?

It tags itself jev, jev-ai, jev-api, and jev-model and calls its RAG gate "Jev-Style 2B", so the connection is the typed-decision class: a probability for every option and no generated text. The excerpt shows no TypeSafe endpoint or dependency, so treat it as a resemblance.

How fast are the decisions and what do they return?

The README's headline example: 64 questions about one page in 0.66 seconds, about 10 ms each, on one RTX 5070 Ti; the model gives the probability of every offered option instead of writing an answer, and the same question returns the same answer bit for bit.

What is the RAG gate good for?

On Italian law it judges whether retrieved text can answer a question: AUROC 0.90 versus 0.79 for embeddings, with 2.5x fewer good answers blocked, according to the README's Jev-Style 2B comparison against a plain embedding approach.

What does it cost and under what licence?

The code is AGPL-3.0-or-later with copyright held by I3K Technologies Srl, which offers EuLLM under a separate commercial licence; contributions require a CLA. The digest shows no pricing for the commercial licence or for the Forge and Hub services.

Does my data leave my hardware?

The README says no telemetry and no external API by default, and highlights an audit trail designed for the EU AI Act. It runs locally — the headline measurement uses one RTX 5070 Ti — and the project says it was tested on EuroHPC supercomputers.

What was missing from the README excerpt?

The digest truncates mid-sentence in the "Decisions you can audit" bullet, so install steps, model download details, the Forge and Hub workflow, and licence terms beyond the AGPL headline are unavailable. Metadata: 59 stars, 5 forks, Python, created 2026-03-20, last pushed 2026-10-17.

Jev directory

CommunityJev Tools & CLIs

EuLLM

An open-source platform for creating, distributing, and running sovereign EU-compliant LLMs: one Rust binary with OpenAI- and Ollama-compatible APIs, decisions answered as probabilities in milliseconds, and an audit trail designed for the EU AI Act.

See highlights Open github.com

Stars: 59
Language: Python
License: AGPL-3.0
Reading time: 3 min
Published by: Community
Author: eullm
Last verified: Oct 2026

eullm/eullm

Organization repository on GitHub

View on GitHub

Open-source platform for creating, distributing and running sovereign EU-compliant LLMs. Verticalize any model for your domain, language and brand. AI Act ready.

Forks: 5
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

Tagscommunity github local-llm sovereign rust

Works forJev for Decision gating Jev for Verification Jev for Scoring

Used inJev for Software & DevTools Jev for Legal & Compliance

Languages: Python Rust
Platforms: API
Capabilities: Open sourceSelf-hostedRuns locally

Highlights

README: local AI deciding in milliseconds — one Rust binary, OpenAI- and Ollama-compatible APIs, no telemetry, and an audit trail designed for the EU AI Act.
At a glance: 64 questions about a page in 0.66 s — about 10 ms each — on one RTX 5070 Ti, giving the probability of every option and generating nothing.
A RAG gate for Italian law reaches AUROC 0.90 against 0.79 for embeddings, blocking 2.5x fewer good answers, attributed to a Jev-Style 2B model.
The same question returns the same answer bit for bit, checked on both CPU and GPU; badges show engine v0.7.70 and Forge + Hub in early development.
AGPL-3.0-or-later, copyright I3K Technologies Srl, which also offers a separate commercial licence; contributions require the CLA.

Watch out

README reviewed 2026-10-17 and truncated mid-sentence inside the audit-trail bullet, so install commands, the Forge and Hub workflow, and pricing are not in the digest. The repo tags itself jev, jev-ai, jev-api, and jev-model and the README calls the RAG gate "Jev-Style 2B", but the excerpt shows no TypeSafe endpoint or dependency, so treat the link as a class resemblance. All timings and AUROC figures are the project's own.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

Jev at home: voice pipelines, Assist routing, and sentinels

Oct 16, 2026

How Home Assistant conversation agents, macOS voice pipelines, and an advisory sentinel put a fast typed decision in the loop — one call per utterance, thresholds in code, and a daily token budget.

Jev Tools & CLIs#home-assistant#voice#automation

Read article

Search visibility, audited by decisions

Oct 16, 2026

Jev-based SEO tools split the crawl from the call: deterministic crawlers count and parse, while typed judgments rank fixes, place internal links, score search results, and track AI-answer mentions — with costs and caveats attached.

Jev Tools & CLIs#seo#search-visibility#audits

Read article

Jev on the Gateways: Vercel, Workers AI, and OpenRouter

Oct 12, 2026

The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.

Jev Tools & CLIs#gateway#vercel#cloudflare

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai

EuLLM

Highlights

Questions about this entry

What is EuLLM and who builds it?

How does EuLLM relate to Jev?

How fast are the decisions and what do they return?

What is the RAG gate good for?

What does it cost and under what licence?

Does my data leave my hardware?

What was missing from the README excerpt?

More Jev resources like this

From the Jev guides

From the Jev AI model community