Set up a local AI
Install a local model runner, download a model and verify that local inference works.
Ollama is a local model runner that can download and run supported AI models on your computer.
What this step is for.
A local model runner manages model downloads and exposes a way to run them on your computer. Ollama is one common option, but the workflow matters more than the brand: install a runner, obtain a model that fits your hardware, start it and verify a response.
Keep the scope small enough that you can inspect the result yourself. AI can accelerate the work, but it should not erase the distinction between a suggestion, a changed file, a successful build and a verified product.
Take one concrete action.
Install a trusted local model runner from its official source. Download one model appropriate for your hardware. Run a simple prompt, then restart the runner and confirm you can run the model again.
Install, run and remove a first local model on Windows
1. Check the course baseline. Open Task Manager → Performance → Memory. For this starter exercise, use llama3.2:3b only when Windows shows at least 8 GB installed RAM and you have at least 8 GB free disk space. This is a conservative course rule, not a universal Ollama requirement. If you are below either number, use the hosted path from Step 03.
2. Install Ollama. Open ollama.com/download/windows, download the Windows installer, open it, approve the normal Windows prompt if shown and let installation finish. Close old PowerShell windows and open a new one.
3. Verify the command. Run ollama --version. If you see a version, continue. If the command is not recognized, restart Windows once. If it still fails, open Settings → Apps → Installed apps and confirm Ollama exists; if it does not, rerun the official installer. Do not edit PATH manually for this exercise.
4. Run the model. Run ollama run llama3.2:3b. Wait for the first download. At the prompt type Reply with exactly: local ai works. That exact response is your first success evidence.
5. Verify persistence. Type /bye, reopen PowerShell, run ollama list and confirm llama3.2:3b appears. Run it again and repeat the exact prompt.
6. Storage and rollback. On a normal Windows installation, Ollama stores model data under %USERPROFILE%\.ollama\models. You may inspect that folder in File Explorer, but do not delete model files there manually; use Ollama commands so its state stays consistent. Remove only the practice model with ollama rm llama3.2:3b, then confirm it is gone with ollama list. Remove Ollama itself, if desired, through Settings → Apps → Installed apps → Ollama → Uninstall.
Success: your machine passed the course baseline, the model runs after a restart, and you can remove it safely.
Turn the idea into evidence.
Record the RAM/storage check, Ollama version, model name and the exact response you received. Then perform the removal step so you know how to reclaim the practice model storage.
Do not move on until this is true.
You can produce a local response after a restart, list installed models with ollama list, and remove the practice model with ollama rm llama3.2:3b.
Why we use this principle.
Nyfir Studios treats generated output and verified output as different states. Development work on local AI, Android software and bookkeeping workflows has repeatedly shown that recoverable state, explicit tests and clear product status are more useful than simply producing more output.