New: Open Intelligence’s Custom RAG, built on your own data. Learn more →

New

An in-house RAG, built for your data

We design, deploy and evolve a retrieval-augmented search engine specific to your company: your sources, your vocabulary, your access rights.

Sourced answer

Le préavis de résiliation est de 90 jours pour les contrats cadres 1 et de 30 jours pour les avenants signés après 2024 2.

1Contrat-cadre_2023.pdf · p. 14
2Avenant_n3.docx · §4.2
3Politique_achats.md

They trusted us

Our method

From your documents to reliable answers

01 Scope

Sources, users, reference questions and success criteria.

02 Prepare

Cleaning, chunking adapted to each document type, metadata and rights.

03 Index

Hybrid search, reranking and settings measured against your questions.

04 Measure

Continuous evaluation, user feedback and regular improvement.

A RAG is not an off-the-shelf product: its accuracy depends on the care given to your data.

Quality

Accuracy is built upstream

Structural chunking
Headings, tables and clauses are kept with their context.
Useful metadata
Date, version, author and scope guide the search.
Evaluation set
Your real questions serve as a reference at each evolution.
01Vos documents
02Indexation
03Recherche
04Réponse citée

Control

Your infrastructure, your rules

Hosting of your choice
European cloud, private cloud or on-premises.
Model of your choice
Open, proprietary, or hosted on your premises.
Reversibility
Your indexes and your data belong to you.

Weekly summary

1 240Questions traitées
94 %Réponses sourcées
−38 %Temps de recherche

Frequently asked questions

What is a RAG?

Retrieval-augmented generation: before answering, the model retrieves the relevant passages in your documents, then writes based on them.

Why a custom RAG rather than a generic tool?

Because quality depends on the chunking, the metadata and the rules specific to your content. Generic tuning plateaus quickly.

How long does a project last?

A first scope is usually in production within a few weeks; the expansion then happens through iterations.

What happens when the agent does not know?

It says so, indicates what it searched and suggests who to turn to.

Go from the question to the decision

The fastest way to understand Open Intelligence is to see it answer on your own documents. Describe your need: we’ll come back to you with a tailored solution and a costed quote.