What is an LLM?
LLM stands for Large Language Model — a powerful AI system trained on a massive amount of text data. The most well-known models are ChatGPT, Claude, GigaChat, YandexGPT, DeepSeek, and Qwen. Some of them are open-source, and you can even run them on your own server.
For a user, an LLM is a black box:
You give it text — and get a meaningful output: a translation, a summary, or even code.
The key concept here is the prompt — that’s what you feed into the model.
For example:
“Translate this text into Spanish" or “Write a Telegram bot in Python.”
It’s important to understand:
- An LLM doesn’t know facts.
- It doesn’t search Wikipedia.
- It just continues text based on probability.
That’s why sometimes it hallucinates — confidently making things up.
Where Are LLMs Useful?
Here are some common and powerful use cases:
- Customer support
- Internal assistants
- Feedback processing
- Report and documentation generation
- Content translation and adaptation
The most important part:
→ Formulate your prompt clearly and thoughtfully.
Hallucinations: Why Do They Happen?
Let’s take an example. You ask:
“What books did Shakespeare write in the 20th century?”
The model might reply:
“In 1923, Shakespeare published The Queen’s Tragedy…”
Sounds convincing — but it’s a complete fabrication. Shakespeare died in the 17th century.
These confident but false answers are called hallucinations.
To solve this problem, we use the RAG approach — Retrieval-Augmented Generation — where we feed the model real, factual data.
More on that in Lesson 2.
Your First Practical Project
Let’s build a simple app on Directual: An assistant that recommends movies.
The user enters a description or preference, and the model suggests 3 suitable movies We’ll use the ChatGPT API, with an API key from your OpenAI account.
We’ll set up:
- A database table for requests
- A UI form and a list of responses
- Two backend scenarios:
- One to save the user request
- Another to send it to the LLM and save the model’s reply
The response is saved to the database and instantly shown in the interface.
→ In the end, you’ll have a working AI-powered service that responds to real users.
You can download the app snapshot from this lesson.
⚠️ After importing, you’ll need to manually install all the required plugins!
What’s Next?
Congratulations — you’ve taken your first step!
You now know:
- What an LLM is
- How to work with it
- And how to integrate it into your own web app — without writing code
But things are about to get even more interesting.
In Lesson 2, we’ll learn how to build a smarter AI — one that replies based on your actual data, not just guesses.
We’ll explore:
- Vector databases
- Embeddings
- Semantic search
… and build all of that with your own hands.