Troubleshooting

Start with the setup check from the code folder. It tells you which part is failing:

python -m agentlab.check

Setup

Could not reach http://localhost:11434/v1. Is Ollama running?

Ollama isn’t running. Open the Ollama app (macOS, Windows), or run ollama serve in a separate terminal (Linux).

404 from model endpoint: model "..." not found

The model in .env isn’t downloaded. Run ollama list to see what you have, then either ollama pull qwen3:4b-instruct or change LLM_MODEL to match a model in the list.

ModuleNotFoundError: No module named 'agentlab'

Your virtual environment isn’t active, or you skipped the install. From the code folder:

source .venv/bin/activate      # Windows: .venv\Scripts\activate
pip install -e .

python: command not found or the wrong Python version

On macOS and Linux, try python3. Check the version with python3 --version. You need 3.10 or newer.

Windows: running scripts is disabled on this system

PowerShell blocks the activate script. Run this once, then try again:

Set-ExecutionPolicy -Scope CurrentUser RemoteSigned

.env changes are ignored

Make sure the file is called exactly .env (not .env.txt) and lives in the code folder. Windows Explorer hides extensions by default, so check in your editor.

Running the modules

Every reply takes a long time

The first call loads the model into memory, which can take a while. After that it should be faster. If it’s still slow:

• Close other heavy apps, especially browsers with many tabs
• Switch to a smaller model, such as qwen2.5:1.5b
• Use a cloud provider instead. See Use any model provider

Read timed out

The model took longer than the timeout. Add LLM_TIMEOUT=300 to .env.

The model answers without using tools

Some models don’t support tool calling, or support it poorly. Use a model from the list on Use any model provider. Also check that your tool descriptions are clear. Vague descriptions make models skip tools.

The model wanted to call another tool in Module 2

That’s expected. Module 2 allows only one round of tool calls. Module 3 fixes it.

NotImplementedError: TODO ...

You’re running a starter file with a TODO still to do. Fill it in, or run the matching solution file.

Results are different every time

That’s how language models work, and it’s the reason Module 4 exists. Use --runs 3 with the eval runner to see how consistent your agent is.

Still stuck?

Ask a helper, or if you think you’ve found a bug, please open a GitHub issue.