LLM Chat Interface – User Documentation#

Purpose#

The LLM Chat Interface is a web-based application for using Large Language Models (LLMs). It enables you to engage in direct dialogue with AI models for tasks such as text analysis, translation, brainstorming or scientific research.

The basic principle: you enter questions or tasks, the system sends them to a locally operated language model and returns the answers to you in real time. You retain full control over the conversation history, the style of the answers and the prompts used.

Range of functions#

The system offers the following core functions:

• Chat with streaming responses: Receive responses from the model in real time as they are generated. You can interrupt the generation at any time.

• System prompt configuration: Choose between predefined role profiles (standard HU, tutor, inspirational) or define your own instructions for the model.

• Response modes: Control the style of responses using four modes – from precise and factual to creative and associative.

• Prompt templates: Use predefined templates for common tasks such as text analysis, summaries or translations. One click copies the prompt into the input field.

• Session management: Conduct multiple independent conversations in parallel. The history shows you all active sessions with automatically generated titles.

• Multiline input: Write longer texts directly in the input field. The system supports multi-line input with automatic resizing.

• Dark/Light Mode: Adjust the interface display to your preference or lighting conditions.

Operation#

Basic usage#

1. Enter message: Type your question or task into the text field at the bottom of the screen.

2. Send: Click on ‘📤 Send’ or press Enter (Shift+Enter for line break).

3. Receive reply: The reply appears in real time in the chat area. You can cancel during generation with ‘⏹️ Stop’.

4. Continue conversation: The system remembers the conversation history. Each new message refers to the previous context.

Advanced features#

Change system prompt:

• Open the accordion ‘⚙️ System prompt’
• Select a template from the drop-down menu or
• Edit the prompt directly in the text field

Select response style:

• Select one of the four modes under ‘Style’:
• Balanced: Default setting, balanced between creativity and precision
• Creative: For associative and unconventional responses
• Precise: For factual and scientifically sound responses
• Short: For compact and focused responses

Use template:

• Open ‘📝 Quick Prompts’
• Click on a button (e.g. ‘Just explain: Explain the topic to me in an understandable way’)
• The complete prompt appears in the input field
• Add your specific topic at the end

Start a new session:

• Click on ‘🗑️ New chat’
• A new, empty conversation will open
• The previous session will remain in the history

View history:

• Click on ‘☰’ in the top left corner for the sidebar
• Select a session from the list
• The entire conversation will load

Important notes#

• The system supports Markdown formatting and LaTeX formulas in responses.
• Session titles are automatically generated after the first message.

Example of use#

Initial situation: You want to check a scientific abstract for comprehensibility and simplify it.

1. Select template: Open ‘📝 Quick Prompts’ and click on ‘Explain simply: Explain the topic to me in an understandable way’.

2. Insert text: The prompt appears in the input field. Add your abstract at the end:

  Please explain the following topic to me in a simple and understandable way. ..
  
  Topic: [Your abstract text]

3. Adjust mode: Under ‘Style’, select the “Precise” mode to ensure that the simplification remains technically correct.

4. Send: Click on ‘📤 Send’.

5. Result: The system provides a simplified version with everyday examples and a clear structure.

6. Follow-up questions: You can ask follow-up questions directly: ‘Explain term X in more detail’ – the system refers to the previous context.

7. New task: For a completely different topic, start a new session with ‘🗑️ New chat’.

Recommendations for efficient use#

• Use templates: The predefined prompts are structured and deliver better results than free formulations.

• Precise questions: The more specific your task, the more targeted the answer. Avoid overly open questions.

• Build context step by step: Start with a basic question and refine it by asking follow-up questions. The model remembers the course of the conversation.

• System prompt for roles: Use ‘Tutor’ for learning scenarios, “Inspirational” for creative processes and ‘Standard HU’ for scientific questions.

• Choose modes consciously: ‘Precise’ for facts and definitions, ‘Creative’ for brainstorming and changing perspectives.

• Use the stop function: If an answer is not going in the desired direction, stop early and rephrase the question.

• Organise sessions: Use separate sessions for different topics – this makes it easier to find them again later.

System limitations#

The tool CANNOT:

• Access external sources: The model has no internet access and is not updated beyond January 2025.

• Process files: You cannot upload or analyse documents.

• Use multiple models in parallel: One LLM model is active per installation. Changing models requires technical adjustments.

• Guaranteed accuracy: The model may generate incorrect information. Please check critical facts.

• Saving personalised data: Sessions are only stored temporarily in the browser. Closing the browser deletes the history.

Summary#

The LLM Chat Interface is a tool for text-based AI interaction. It simplifies access to language models through an intuitive interface with predefined prompts and flexible settings.

You control the dialogue by selecting the system prompt, the response mode and by asking specific questions. The system provides suggestions but does not complete your task independently – it is a dialogue partner, not an autonomous agent. Use it as a support for your scientific work, not as a substitute for critical thinking and source verification.