This page is machine-translated from French. Read the French original.

WivenLLM Documentation

WivenLLM is a private artificial intelligence platform: it installs on your workstation, your servers, or in a Swiss data center, and works on your documents and business tools. In a fully local configuration, your data never leaves your network.

This documentation covers the entire product: understanding what it does, installing it, using it daily, administering it and integrating it into your information system.

Documented version: 1.17.2


Where to begin

You are… Start by
Curious — you discover WivenLLM What is WivenLLM Then Use Cases
User — you have just been given access First steps Then Workspaces
Administrator — you deploy for your organization Facility Then Users and roles
Developer — you integrate WivenLLM into your applications API developer

Large capacities

Discuss with your documents

Upload PDFs, Word documents, spreadsheets, emails, web pages, or audio recordings. WivenLLM will parse them, index them, and answer your questions, citing its sources. → Documents

Choose where the AI runs

Open models run locally, your organization's inference server, or a cloud provider if you choose. The choice is made per workspace and can change at any time. → Choosing your models

Automate with agents

An agent performs a series of actions—web searches, file reading, SQL queries, emails, calendar entries—to complete a task, either independently or under your supervision. These skills are activated one by one: upon installation, only those that remain on the machine are active. Sending, replying to, and deleting an email requires confirmation each time. → Agents Verification Status

Capture a meeting

Record a meeting from the application: it is transcribed locally using the default engine, then summarized or transformed into a structured report upon request, and remains searchable afterwards. → Meetings

Remember what matters

Memories (referred to as «Memories» in the interface, which can be activated) store useful facts about you. The Second Brain (referred to as «Second Brain» in the interface) is a personal knowledge base that spans all your workspaces. → Memoirs · Second Brain

Governing usage

Roles, organizations, single sign-on, content control, event log, audit log: the administration interface is designed for business use. Multi-user mode is covered by the WivenLLM Business license. Several protections, including content control and the privacy shield, are disabled by default: they only apply once activated and configured. → Administration


New since version 1.15

  • Packages for Red Hat 9 / Rocky 9 and for NVIDIA DGX Spark.
  • My connections: Each member of an organization connects their own Outlook or Gmail account and their Google Calendar; the agent only uses those of the person making the request.
  • Entra Groups: The directory groups decide who connects, with what role, and where.
  • Guest access: daily limit of messages per guest (20 by default).
  • Wiven console: New tabs for Installation Health and Outgoing Email; Response Control now explains each failure.
  • Meetings: transcription on the machine by NVIDIA Parakeet, and direct summary offered in the input bar after recording.

Conventions in this documentation

  • The paths in the interface are labeled Settings → Section → Page. The settings menu is labeled «Instance Settings» on the screen.
  • Wiven Console refers to the separate administration screen, opened by the «Wiven Console» button located next to the model selector. Its tabs are labeled Wiven Console → Tab.
  • Default values are shown where they exist; almost all of them are modifiable.
  • Functions marked Administrator are only visible to accounts with that role.
  • Functions marked Multi-user Mode assume that the instance has been switched to multi-user, which requires the WivenLLM Business license.