How to set up local Ollama

What it is, and when to pick it

Ollama runs open models (Llama, Qwen, Mistral) as a small service on your own computer. There is no key and no account: this site’s page talks to localhost on your machine, so your resume never leaves it. Pick Ollama if you have 8 GB+ of RAM and want zero cost and total privacy; the trade-off is quality and speed — a local model on a laptop trails the hosted frontier models.

Step by step

1. 1

   Download and install Ollama for your operating system. ollama.com/download

2. 2

   Pull a model — in a terminal, run ollama pull llama3.2 (about 2 GB), or pick any model from the library. ollama.com/library

3. 3

   Allow browser access: Ollama only answers requests from origins listed in OLLAMA_ORIGINS. Set it to https://practicalwebtools.com for your OS below, then restart Ollama.

4. 4

   Windows — search Start for “Edit the system environment variables”, add a user variable OLLAMA_ORIGINS = https://practicalwebtools.com, then quit Ollama from the system tray and start it again.

5. 5

   macOS — in Terminal, run launchctl setenv OLLAMA_ORIGINS https://practicalwebtools.com, then quit Ollama from the menu bar and reopen it.

6. 6

   Linux (systemd) — run sudo systemctl edit ollama, add Environment="OLLAMA_ORIGINS=https://practicalwebtools.com" under [Service], then run sudo systemctl daemon-reload and sudo systemctl restart ollama.

What the key looks like

Ollama uses no API key — the key box stays empty for this provider.

What it actually costs

• Free — models run on your hardware, so the only cost is electricity.
• llama3.2 (3B) fits in 8 GB of RAM; bigger models need proportionally more.
• Runs can be slower than hosted providers on older laptops.

Paste it into Practical Web Tools

In the AI resume tailor, choose Ollama (local) — there is no key to paste. The tool reads whatever models your machine has and talks to http://localhost:11434 directly. Open the AI resume tailor.

If something goes wrong

Could not reach the provider

Ollama is not running or OLLAMA_ORIGINS is not set. Open http://localhost:11434 in a tab — you should see “Ollama is running” — and re-check the environment-variable step for your OS.

Model not found

Pull it first: ollama pull llama3.2. The tailor can only offer models your machine already has.

Request blocked right after setup

The OLLAMA_ORIGINS change needs a full restart of Ollama — on Windows quit it from the system tray, on macOS from the menu bar, on Linux via systemctl restart.