New

Yumi: Your documents become an AI assistant, ready to use

Yumi is the portal version of our AI: upload your documents, create topic-based channels and get answers with their sources. No hardware to buy or install: we run the infrastructure, you use the assistant.

Answers that cite their sources
No dedicated hardware
Monthly plan, nothing to install
How it works

Three steps, no setup

Organization, channels, conversations: the same structure for a small team and for a whole company.

1

Upload your documents

Public web pages and, from the Lite plan, PDF, Word, ODT and text files. Yumi reads and indexes them; from the Lite plan it re-checks web pages every day to keep them up to date.

2

Create channels by topic

Each channel is a chat environment with its own documents, instructions and assistant name. The same document can serve several channels.

3

Ask, and check the sources

Answers arrive in real time with references to the documents used. When the information is not in your documents, Yumi says so.

Features

What you get

Everything needed to bring an assistant to your team, your website or your applications.

Verifiable answers

Every answer can show the passages it is based on, with page and source. References can be turned off per channel.

Portal and website widget

Publish a channel on a public page or embed it in your website with a widget, limited to the domains you authorize.

API for your applications

From the Pro plan, connect a channel to your software with revocable tokens, each limited to a single channel.

Team and roles

Administrators, curators and readers, with permissions per channel. Private channels are visible only to the people you assign.

Secure access

Two-step login, authenticator app and recovery codes, protection against repeated attempts.

Usage under control

AI credit balance and usage at a glance, message limits per channel and per conversation for public access.

Use cases

Inside and outside your organization

Every channel is public or private. Use Yumi for your team, for your customers, or for both at the same time.

Private channels

For internal use, from the Lite plan: only the people you assign can sign in and ask questions.

  • Access with personal login, per-channel permissions
  • Each person keeps their own conversation history
  • Internal documents never reach the public pages

Public channels

For customers and visitors: no account needed to ask a question.

  • Public page with its own link, or a widget on your website
  • Message limits per day and per conversation
  • Can be suspended at any time with one switch
Private

Rules and internal procedures

Regulations, HR policies, procedures and forms in one place: staff ask instead of searching through folders.

“How do I request leave, and how much notice do I need to give?”

Private

Technical knowledge base

Manuals, technical sheets and operating procedures available to technicians and support staff, with the source of each answer.

“What is the reset procedure for model X?”

Private

Onboarding new colleagues

A channel with everything a new hire needs to know, available from day one without taking up colleagues’ time.

“Who do I contact for access to the accounting system?”

Public

Customer support on your website

FAQs, product sheets, terms and conditions: the widget answers your website visitors at any time.

“What is the warranty period, and how do I return a product?”

Public

Information for the public

Opening hours, events, services and regulations published on a public page that is easy to share.

“Is the museum open on public holidays?”

Public + private

Same documents, two audiences

A document can serve several channels: a public channel answers customers, a private one supports the staff with internal procedures too.

Public: product FAQ · Private: FAQ + internal price lists and procedures

Your data

Where your data goes

Clear answers, no fine print.

Documents and index on our infrastructure

Files, text passages and the semantic index are processed and stored on infrastructure managed by digitalCode, with the same engine as Akiko.

Each organization is separate

Separate files, separate index, separate permissions: one organization's content never enters another's answers.

Answers from a cloud language model

To write the answer, Yumi sends a cloud language model the question, the relevant passages and the latest turns of the conversation, never your whole library.

If your data cannot leave your perimeter, the answer is Akiko: the same engine, on a dedicated server.

Discover Akiko
Two versions, one engine

Yumi or Akiko?

Both are born from Project Nicky, our AI laboratory. The difference is where they run and who they are for.

YumiAkiko
FormatReady-to-use portalDedicated server
Where it runsdigitalCode infrastructureOn-premise or on digitalCloud
Answer generationCloud language modelFully internal models
Getting startedSign up and choose a planTailored analysis and setup
PricingMonthly plans: Start, Lite, Pro, MaxCustom quote
Ideal forTeams, SMEs, professional firmsPublic administration, confidential data, large organizations

Try Yumi right on this page

The chat button in the bottom-right corner opens Yumi: our own assistant, answering from the content of this website. When you want yours, create an account on the platform or talk to us: we will help you choose the plan.