This guide will help you get started with Lehmus AI. You will learn how to sign in to the service, select the right project, and start using language models in Chat.
What you need before you start
You will need:
- Access to Lehmus AI
- A connection to the University of Oulu network or a VPN connection
- Your organization’s user account
If you do not yet have access, see the guide Lehmus AI: A secure generative AI platform – ict.oulu.fi.
What is Lehmus AI?
Lehmus AI is the University of Oulu’s generative AI platform. It allows you to use language models in Chat, use RAG knowledge sources based on your own materials, use AI agents and tools, and connect language models to other applications via OpenAI-compatible APIs.
Lehmus AI offers several ways to use generative AI:
- Chat: interact directly with language models
- RAG: let the language model retrieve information from your own materials
- AI agents and tools: use preconfigured AI agents and their associated RAG and MCP tools. Project administrators can also create their own agents for the project.
- APIs: use Lehmus AI’s language models in your own applications or development tools, for example
Sign in to Lehmus AI
- Connect your device to the University of Oulu network or establish a VPN connection.
- Open the Lehmus AI portal at https://portal.lehmus-ai.oulu.fi/.
- Sign in with your organization’s user account
If you cannot sign in, first check your network connection and access rights.
Select the right project
In Lehmus AI, project workspaces are called tenants. A tenant keeps the project’s conversations, materials, RAG configurations and other resources separate from the contents of other projects.
If you have access to more than one project:
- Check which project is currently active in the portal
- Switch to the right project if necessary
- Confirm your selection before you start working
Important: if you belong to more than one project, always check out the active project before adding project materials to the service or creating new resources.
Get started in Chat
In Lehmus AI’s Chat, you can interact with language models and preconfigured AI agents.
- Open Chat from the portal’s left-hand sidebar
- Open the model menu
- Select a language model or AI agent suited to the task (if configured)
- Enter your request in the message field
- Send the message
- Check the response before using it
The model menu in Chat shows the chat language models and agents you have access to.
You can find tips on writing effective prompts in the guide Tips for using AI effectively – ict.oulu.fi.
Note: always check the language model’s response before using it. A language model may produce incorrect or outdated information. Where necessary, check sources, calculations, code, and information security implications, for example.
Use your own materials with RAG
RAG (Retrieval-Augmented Generation) allows language model responses to be supplemented with information retrieved from your own documents. A RAG endpoint allows you to manage the documents attached to it and use them for information retrieval.
You can see the RAGs available to you in Chat by selecting the Knowledge bases button at the bottom of Chat. From there, you can activate RAGs for use in Chat, so that the language model will try to formulate its response based on your own materials.
In Lehmus AI, you can create a RAG endpoint, add your own materials to it and use RAG as a knowledge source for a language model, for example through Chat.
You can find step-by-step instructions for creating a RAG in the guide Lehmus AI: setting up RAG to use your own material – ict.oulu.fi
Use a preset agent
Lehmus AI allows you to use preconfigured Preset Agents. An agent can be configured with a language model, instructions and, where applicable, tools.
An agent you have access to appears in Chat’s model menu as Preset agent: [AGENT NAME]. Select the agent from Chat’s model menu and send it a message as you would in a normal Chat conversation. Using an agent in Chat does not require an API key.
If the agent you expect to see does not appear in Chat, check that you have selected the right project and have been granted access to the agent. You can check your access to the agent with your project administrator.
Project administrators can find instructions for creating Preset Agents in the guide Lehmus AI: Creating and managing a Preset Agent – ict.oulu.fi.
Use tools in Chat
Tools can also be connected to Lehmus AI’s language models and agents using MCP (Model Context Protocol). Tools from an external MCP server can be added to Lehmus AI’s shared selection of tools.
You can see the available MCP tools in the Tools menu at the bottom of the Chat. From there, you can activate available MCP tools for your Chat conversation, allowing the language model to use MCP tools to answer questions.
If a tool requires you to sign in with your own account, it appears Not Connected in Chat’s Tools menu. Select Connect, sign in to the service in the browser tab that opens and grants the required access. The tool’s functions will then become visible.
The tool accesses the external service using the user’s own permissions.
Using Lehmus AI in other applications
Lehmus AI provides a supported set of OpenAI-compatible APIs. Model Gateway supports Chat Completions, Responses and Embeddings APIs, as well as model listing, among other features.
To use Lehmus AI’s language models in external applications, you will need:
- Access to Lehmus AI
- An API key with the language models you want to use linked to it
- Lehmus AI’s API URL
- The identifier of the model you want to use, i.e. its Public Model ID
The API URL is shown under the API URL at the bottom of the Lehmus AI portal’s sidebar. You can find model identifiers on the Models tab in the Lehmus AI portal.
You can find more detailed instructions in the following guides:
You can find instructions for connecting Lehmus AI’s language models to other applications:
For project administrators
Project administrators can find instructions on managing users and access rights in the guide to Lehmus AI: Managing project users and permissions – ict.oulu.fi.
For software developers
Lehmus AI supports OpenAI-compatible Chat Completions and Responses APIs. The Responses API also supports conversation history and tools, for example.
You can read more in the guide Lehmus AI: Using the APIs – ict.oulu.fi
Support and further information
Lehmus AI is based on Confidential Minds technology. You can find more advanced settings and in-depth technical information in the Confidential Mind documentation.
If you encounter problems using the service, you can contact us by email at ict@oulu.fi.