Using Chats

You can chat about a wide range of topics with Alan, whether it's answering specific questions, creating and optimizing text, developing ideas, or extracting and structuring information.

Each chat with Alan is a separate conversation, and Alan remembers the previous contents of a conversation via the context. Therefore, it is a good idea to start a new chat for each new question or topic.

Starting a Chat

To start a chat with Alan, click on the "New Chat" button in the sidebar.

You can now enter your prompt (e.g. a question or instruction) directly in the chat input field at the bottom of the screen and submit it.

Alternatively, you can first select an expert, more information about this advanced feature can be found in the section Experts.

Interacting with Alan

To interact with Alan in a chat, follow these steps:

1. Formulate questions: Enter your question or request in natural language into the chat interface. Precise formulations improve the response quality.
2. Start a conversation: If the initial results do not exactly match your expectations, do not hesitate to ask follow-up questions. Alan is designed to meet your needs through conversation.
3. Check: Before using the answer in your work, it is important to verify its accuracy. Make sure the information meets your requirements and verify critical details if necessary.

For more tips on chatting with Alan, consult our Prompting Guide.

Message Actions

At the bottom of a submitted or completed message, you see various actions that you can execute by clicking on them:

• Copy: Copy the content of the message as Markdown, preserving formatting.
• Regenerate: You can ask Alan to generate an additional answer. This is helpful if an answer does not yet exactly meet your requirements or if you want to collect multiple answer ideas. You can navigate between the different answers by using the arrow buttons that appear after generation.
• Edit: You can edit your submitted inputs and ask Alan to generate a response to the edited prompt. This is helpful if you want to make your original prompt more specific. You can also navigate between the different prompt options and Alan's respective responses by using the arrow buttons that appear after generation.

When you regenerate an answer or edit a prompt, all previous variants are preserved. You can move between these versions and their branches using the arrow controls of the switch bar ("Variant created").

While an answer is being generated, you can click on the "Abort?" button at the top of the message to abort the generation.

Understanding Alan's Response Process

For more complex requests, Alan works through several intermediate steps before formulating the final answer. Alan summarizes these intermediate steps in a collapsible response process block above the respective answer.

The block aggregates two kinds of intermediate steps:

• Reasoning: Alan's considerations while planning the next steps. This is only available if it is supported by the model in use. More details can be found in the glossary entry on reasoning.
• Tool calls: the calls of abilities that Alan uses to answer your question.

While Alan is working, the current step is shown live. As soon as the step is complete, the header of the block shows its completion state (e.g. "Done"). Using the arrow icon, you can expand and collapse the block to inspect the individual intermediate steps in detail. This lets you transparently follow how Alan arrived at the answer.

Chat with Files

If you want to use files as the basis for your request, you can upload them directly to the chat using the Chatting with Files feature. This allows Alan to consider the entire content of your files and provide more precise answers. This is especially helpful when:

• Background knowledge is needed to understand your question.
• You want to summarize a file.
• You want a direct comparison between multiple files.
• You want to include more extensive content without having to manually copy and paste.

Chat Configuration

Before or after Alan generates a response, you can adjust the chat configuration to ensure your chat with Alan is tailored to your specific needs.

You can set the following options:

• Knowledge Databases: Select one or more knowledge databases for Alan to use when answering your questions. This is useful for accessing specific information. More details can be found in the Knowledge Databases section.
• Abilities: Select one or more abilities which may support Alan in answering your questions. More details can be found in the Ability section.
• Models: Choose the large language model (LLM) Alan should use for generating answers. More details are available in the Model Selection section.

For this, you will find three separate buttons below the input bar: the model button (shows the currently selected model), the "Knowledge databases" button, and the "Abilities" button. Clicking one of these buttons opens a quick-select popover, in which you see pinned, suggested, or currently selected resources and can activate them directly.

Via the "Show more..." row at the bottom of a quick-select popover, you reach the full "Configure chat" dialog. There, you will find the tabs "Knowledge databases", "Abilities", and "Models" on the left, along with an overview and further details of all available resources, which you can select as needed for your current chat.

Settings made here affect the entire chat. Previously generated answers are not changed, but all future answers will use the new settings. This applies regardless of the position in the message history and therefore also affects regenerating older answers in the chat.

Model Selection

Depending on your organization's settings, you can select which large language model (LLM) Alan should use for generating answers. Different models can affect the style, accuracy, and response speed of answers.

Below the chat input bar, you can see which model is currently selected.

Information

A language model is pre-selected for you in each chat by default. You can start chatting immediately without selecting a model, but you can adjust the model selection as needed.

To change the model, click the desired model in the model quick-select or in the "Models" tab of the chat configuration.

You can adjust the model selection at any time during the chat. For example, you can use the "Regenerate" action to generate answers with different models and compare or combine them for optimal results.

In the "Models" tab of the chat configuration, a detail panel for the currently selected model is shown to the right of the model list. There you will find the model's Description, Protection status, Status, Context window size, Full name, and API name.

Model Status

For each model, you see the current status in the chat configuration. Models with the status available are ready for use.

Status	Description
Available	The model is available and can be used for chatting.
Updating	The model is being updated and will most likely be available for chatting again once the update is complete.
Inactive	The model is currently not available, but it can be started by sending a chat message. After about 15 minutes, it will be ready.