💬 Basic Chat: Mastering Simple LLM Conversations

Welcome to the Basic Chat tutorial! This guide will help you understand how to have effective conversations with Large Language Models (LLMs) using CellMage.

🎯 What You’ll Learn

In this tutorial, you’ll discover:

  • How to structure effective prompts

  • The basics of conversation flow

  • How to refine and improve LLM responses

  • Best practices for clear communication with LLMs

🧙‍♂️ Getting Started with Basic Chat

Before diving in, make sure you have:

  • CellMage installed (pip install cellmage)

  • An API key configured (see the Quickstart guide if needed)

  • The extension loaded in your notebook:

%load_ext cellmage

🗣️ Step 1: Your First Conversation

Let’s start with a simple conversation:

%%llm
Explain what makes a good prompt for an LLM. Keep it concise.

You’ll notice the LLM responds with a concise explanation. But what if we want to follow up on this topic?

🔄 Step 2: Follow-Up Questions

CellMage maintains conversation context automatically, allowing for natural follow-ups:

%%llm
Can you give me 3 specific examples of what you just explained?

The model remembers the previous discussion about good prompts and provides examples based on that context.

🧩 Step 3: Structuring More Complex Prompts

For more detailed responses, structure your prompt with clear sections:

%%llm
I need information about Python's list comprehensions.

Please structure your response as follows:
1. A simple definition
2. Basic syntax with an example
3. Three advanced examples with explanations
4. Common pitfalls to avoid

Notice how providing a structure helps the LLM organize its response in a more useful way.

🔮 Step 4: Different Response Styles

You can request different styles or formats for responses:

%%llm
Explain the concept of "technical debt" in software development.
First, explain it as if I'm a senior developer.
Then, explain the same concept as if I'm a non-technical project manager.

This approach is useful when you need explanations tailored to different audiences.

🌡️ Step 5: Adjusting Temperature

The temperature parameter controls the creativity and randomness of responses:

# More deterministic, focused response
%%llm --temperature 0.1
List 5 best practices for writing clean code.

# More creative, varied response
%%llm --temperature 0.8
List 5 best practices for writing clean code.

Compare the two responses. The lower temperature produces more predictable, conventional advice, while the higher temperature might introduce more unique or creative suggestions.

✏️ Step 6: Refining Responses

If a response isn’t quite what you wanted, you can ask for refinements:

%%llm
Your previous response about clean code was helpful, but could you focus more specifically on Python best practices and include examples for each?

This iterative refinement is one of the most powerful aspects of conversational LLMs.

🎮 Step 7: Controlling Response Length

You can guide the LLM to produce shorter or longer responses:

%%llm
Explain quantum computing in 2-3 sentences only.

Or for a more detailed explanation:

%%llm
I'd like an in-depth explanation of quantum computing, including:
- Key principles and concepts
- How it differs from classical computing
- Current limitations and challenges
- Potential future applications

Please be thorough and provide examples where helpful.

👔 Step 8: Setting the Tone

You can request a specific tone for the response:

%%llm
Write a brief explanation of blockchain technology with an enthusiastic tone.

Now, explain the same concept with a more formal, academic tone.

🚫 Step 9: Handling Unhelpful Responses

If you receive a response that isn’t helpful, you can:

%%llm
That explanation wasn't quite what I was looking for. Let me clarify: I'm trying to understand blockchain specifically in terms of its application for supply chain tracking. Can you focus on that aspect?

🧠 Best Practices for Basic Chat

  1. Be Clear and Specific: State exactly what you need

  2. Provide Context: Give background information when needed

  3. Structure Complex Requests: Use numbered lists or sections

  4. Iterate: Refine your questions based on previous responses

  5. Experiment: Try different prompts, temperatures, and approaches

🔍 Examining Your Chat History

To review your conversation:

# Show your full conversation history
%llm_config --show-history

🧹 Starting Fresh

When you want to start a new conversation without previous context:

# Clear conversation history
%llm_config --clear-history

🎓 What’s Next?

Now that you’ve mastered the basics of chatting with LLMs through CellMage, explore these tutorials:

Happy chatting with your magical AI assistant! ✨