Learning Path: From Chat to Agent

A step-by-step plan from first model run to a working multi-agent team.

🇷🇺 Russian version: learning-path.ru.md

← FAQ · Install on Windows →

Route overview (2–3 hours)

This plan takes you from installing your first program to building your own AI agent. Each step takes 10–30 minutes. Follow them in order.

Step 1  LM Studio — just try it
Step 2  Install Ollama and run a model
Step 3  Understand model types and differences
Step 4  Learn about context and quantization
Step 5  Choose a model for your task
Step 6  Set up a local coding assistant
Step 7  Try Aider — terminal coding agent
Step 8  Create your first AI agent
Step 9  Set up Open WebUI — web chat interface
Step 10 Build an agent team for your project

Step 1. LM Studio — just try it

Time: 5 minutes Result: you talk to a local AI model for the first time.

1. Download LM Studio — GUI app, no terminal needed
2. Install and open
3. Search for “Qwen 3.5 4B”, click Download
4. After download click Load Chat
5. Type: “Hello! What can you do?”

Done. You just ran an AI model on your computer. Free, offline, no registration.

Step 2. Ollama — install and run your first model

Time: 10 minutes
Result: you install Ollama and run models via the terminal.

Choose your platform:

• Mac: ../local-models/getting-started.md
• Windows: setup-windows.md
• Linux: setup-linux.md

ollama run qwen3.5:4b

Chat with the model. Try /bye to exit.

Step 3. Understand model types

Time: 15 minutes
Result: you understand how models differ and which to choose.

Read:

• ../local-models/running-models.md
• what-is-ai.md

Run two different models and compare:

ollama run phi4-mini        # small (2.5 GB)
ollama run llama3.3:8b      # larger (4.9 GB)

Step 4. Context and quantization

Time: 15 minutes
Result: you understand why models are slow and how to fix it.

Read:

• ../local-models/memory-and-context.md
• ../local-models/quantization.md

Try:

OLLAMA_CONTEXT_LENGTH=16384 ollama run qwen3.5:4b

Step 5. Choose a model for your task

Time: 10 minutes
Result: you know which model to use for coding, chat, translation, RAG.

Read ../local-models/models.md. Download a coding model:

ollama pull qwen2.5-coder:7b

Step 6. Set up local coding assistant

Time: 20 minutes
Result: AI helps you write code right in your editor.

Follow ../use-cases/coding.md:

• Install VS Code + Continue.dev
• Connect to Ollama
• Start coding with AI

What to try:

• Cmd+I — open AI chat in your editor
• Select code and ask “explain this function” or “find bugs”
• Ask it to write a function: “write a function that sorts a list in descending order”

Step 7. Aider — terminal coding agent

Time: 15 minutes
Result: you can give AI coding tasks and it modifies code on its own.

# Install
pip install aider-chat

# Run with a local model
export OLLAMA_API_BASE=http://127.0.0.1:11434
aider --model ollama_chat/qwen2.5-coder:7b

What to try:

1. Open any Python file
2. Type: “add docstrings to all functions in this file”
3. Aider reads the file, makes changes, and shows you the diff

More: ../use-cases/coding.md

Step 8. Create your first AI agent

Time: 30 minutes
Result: you wrote a program where the model decides when to call a tool on its own.

Follow the tutorial: ../agents/tutorials/01-first-agent.md

You’ll build an agent that:

• Receives a task
• Decides whether it needs to call a tool (calculator, search)
• Executes and returns the result
• Repeats until the task is solved

Step 9. Open WebUI — web chat interface

Time: 15 minutes
Result: you get a web interface for models, like ChatGPT, but local.

Via Docker:

docker run -d -p 3000:8080 \
  -v open-webui:/app/backend/data \
  ghcr.io/open-webui/open-webui:main

Or without Docker, via pip:

pip install open-webui
open-webui serve

Open http://localhost:3000

Step 10. Multi-agent team

Time: 30–60 minutes
Result: you have several AI agents working together — one plans, another codes, a third tests.

Follow tutorials:

1. ../agents/tutorials/02-agent-team.md — agent team
2. ../agents/multi-agent.md — multi-agent systems

After 10 steps

Congratulations! You now have:

• A working local AI on your computer
• AI assistant in your code editor
• Terminal coding agent (Aider)
• Your first AI agent
• Understanding of how it all works

Where to grow further:

Shortcut route (1 hour)

If you’re really short on time:

Step 1 → LM Studio (5 min)
Step 2 → Ollama + model (10 min)
Step 6 → Continue.dev + VS Code (20 min)
Step 8 → First agent (30 min)

What’s next

In section: what-is-ai · how-models-work · cloud-vs-local · hardware-guide · glossary · faq · learning-path · setup-windows · setup-linux
Related sections: Local Models · AI Agents · Use Cases
Navigation: ← Zero Level · ↑ Back to main · 🇷🇺 Русский