RAG as a Service
on Sovereign Cloud

Fully managed RAG-as-a-Service for developers.

API & SDKs

For developers. Browse our REST API and SDK documentation. Integrate RAG in your own services.

Indexing & Retrieval

State-of-the-art RAG engine, which leverages structured chunking to preserve semantic meaning, multi-layered indexing, and LLM-aware optimizations.

100% European Cloud

GDPR-compliant services remain beyond the reach of extraterritorial laws, ensuring the full sovereignty of your data and that of your clients.

Private & Secure LLM

We use private & opensource LLMs for our AI prompts. Your data does not leave our private environment.

RAG engine

Data extraction

Data sources.: We support all types of data sources: ODF, DOC, PPT, CSV, XLS... files. HTML public webpages.
Advanced parsing.: We are able to parse deeply and consistently, with understanding of tables, graphs, images content.
Extraction.: Maintain content structure and entities. Allows consistent grouping of relevant data. Metadata context.

Data chunking and indexing

Advanced chunking.: Content is systematically divided into logically grouped chunks through type-specific methodologies, ensuring the preservation of context and facilitating improved recall.
Multi-layered indexing.: Enables semantic search, keyword retrieval, and high-level recall by building vector, keyword, and summary indexes.

Data retrieval

Multi index retrieval.: Query multiple index types (vector, keyword, and summary) and gather a richer and more diverse set of contextual information.
Data and structure aware.: Full traceability. Always get context and metadata related to retrieved content.
Ranked results.: Get insights on retrieved content relevancy. Use ranking to filter quality and size of results.

European Sovereign Cloud

We are hosted on the Scaleway cloud:

Hosting in France.: All you data and prompts remain private and inaccessible to third parties, and physically located in French datacenters.
GDPR compliant.: We are dedicated to ensuring compliance with the GDPR to safeguard your personal information.
Cloud & Patriot Act immune.: Non-European authorities cannot have any access to the data in our cloud, since they have no jurisdiction over Scaleway.

Open source Models

We only open source LLM models, privately deployed, for our internal LLM usage.

Mistral AI.: 7B v 0.3. Mistral AI's 7B excells in reasoning, mathematics, and code generation. It's user for code generation, NLP applications, and creative writing.
Google Gemma.: Gemma 3:4b Gemma excels at tasks like text generation, summarization, question answering, and even coding. It has multilingual support for over 140 languages, image understanding, and function calling.