How-to · Ollama

How to install Ollama and run a local model

A repeatable Ollama quickstart that tests the local runtime, model license, hardware fit, and API before connecting other applications.

Setup commands, limits, and security terms change. Follow the linked official documentation before running anything.

Keep/Cut Weekly

Before you start

Use a test profile, synthetic data, and no production credentials. This guide is strategy and evaluation guidance, not managed implementation or support.

  1. Install Ollama from the official source

    Use the official download or installation instructions for your operating system. Avoid third-party installers that obscure the binary or model location.

  2. Choose a model deliberately

    Read the model card and license before pulling it. Note parameter size, quantization, context length, capabilities, and hardware requirements.

  3. Run one interactive test

    Pull and run the model, then send one fixed prompt from your real workflow. Record cold start, response quality, and whether the machine remains usable.

  4. Call the local API

    Ollama documents a local API at localhost:11434 by default. Send the same prompt through the API and save response metadata without saving sensitive content.

  5. Measure the real cost

    Record RAM/VRAM use, tokens per second, disk usage, energy/heat, context behavior, and model download time. Compare with one smaller model before buying hardware.

  6. Connect downstream tools carefully

    After the local API is understood, connect Open WebUI, an editor, or an agent. Keep the API bound and authenticated according to current official guidance.

Completion checklist

  • Model license checked
  • Hardware requirement recorded
  • Cold start measured
  • Local API tested
  • Context behavior tested
  • Downstream connection documented

Official sources

Review tool →Create a decision brief