eu/jev

Product Introduction

1. Definition: eu/jev is a specialized inference API (Application Programming Interface) that provides access to a System One-class large language model (LLM). It is a technical infrastructure product designed for developers and businesses requiring structured AI decision-making.
2. Core Value Proposition: It exists to offer a fully GDPR-compliant, EU-hosted alternative to proprietary System One models like Typesafe's Jev. Its primary value is enabling the use of advanced, structured-output AI for European builders who face data sovereignty restrictions, legal compliance requirements like GDPR, and the need for low-latency inference within European borders.

Main Features

1. EU-Based Inference Infrastructure: The API endpoint and underlying GPU compute are physically hosted on Hetzner servers in Germany. This ensures all data processing—prompt ingestion and answer generation—occurs within the EU. There is no data proxying through third-party CDNs or external model APIs, minimizing latency and jurisdictional risk.
2. GDPR-Compliant Data Processing: The service is engineered for strict privacy compliance. Prompts and answers are not stored or logged post-processing. The provider, Bevelites GmbH, offers a readily accessible Data Processing Agreement (DPA) and a transparent subprocessor list, which is critical for enterprise data governance and legal due diligence.
3. Unified Structured Decision Endpoint: The API exposes a single endpoint (POST /v1/systemone) that can return three types of structured outputs: Choice (selecting from defined options), Noul (generating probabilities for options), and Score (assigning a numerical rating). This consolidates multiple AI tasks into one consistent, application-ready interface, simplifying backend integration.

Problems Solved

1. Pain Point: European software developers and companies cannot use powerful US-hosted AI models like Jev for production applications due to GDPR data transfer rules, data residency requirements, and concerns over foreign surveillance laws (e.g., Cloud Act).
2. Target Audience: The primary personas are EU-based SaaS founders, backend engineers, and product teams building applications that require AI-powered classification, routing, moderation, or decision logic. Compliance officers and legal teams in tech companies are secondary stakeholders who benefit from the clear data governance.
3. Use Cases: Essential scenarios include automated customer support ticket routing (e.g., to "billing" or "technical" teams), real-time content moderation for user-generated content platforms, sentiment analysis of customer feedback, and structured data extraction from unstructured text for EU-centric applications.

Unique Advantages

1. Differentiation: Unlike using a generic LLM API or a US-hosted System One model, eu/jev provides a direct, feature-compatible alternative with a legally sound EU operational base. It differentiates from other EU-hosted AI services by specifically replicating the powerful and unique "System One" structured output paradigm, not just offering a standard chat completion model.
2. Key Innovation: The core innovation is the combination of the open-source Winnow model (the top-ranked System One model on JevBench) with a sovereign, minimalist EU infrastructure stack. This "best-of-breed open model + compliant infra" approach delivers high-performance AI capabilities without compromising on the legal and data sovereignty requirements critical for the European market.

Frequently Asked Questions (FAQ)

1. Is eu/jev fully compatible with the Typesafe Jev API? Yes, eu/jev maintains high API compatibility. Developers can switch the base URL to https://eu-jev.com and use the model identifier jeff-latest to make requests identical to those for Jev, making migration straightforward for existing Jev users.
2. How does eu/jev guarantee GDPR compliance and data privacy? Compliance is ensured through infrastructure choice (Hetzner Germany), a no-logging policy for prompts/responses, and the provision of standard contractual clauses via a DPA. As a German GmbH, Bevel is directly subject to EU data protection law enforcement.
3. What is the pricing model for the eu/jev API? The pricing mirrors Typesafe's Jev at €37 per billion input tokens, with output tokens provided free of charge. This transparent, predictable cost structure is advantageous for high-volume applications compared to per-output-token pricing models.
4. What are the technical limits or quotas for the eu/jev API? While the landing page highlights a 12,000-character context window in the playground, developers should consult the official documentation (/docs#compatibility) for detailed rate limits, maximum context length (token count), and concurrency limits for production API usage.
5. Can eu/jev be used for commercial applications and enterprise contracts? Yes, it is built for commercial and production use. The offering of a DPA and clear subprocessor information is specifically aimed at facilitating enterprise adoption where formal compliance documentation is a prerequisite for procurement.