Ollama Cheat Sheet

A local-LLM quick reference for QA engineers who want privacy-friendly AI workflows and on-device experimentation.

AI Modules 1 and 3 Local LLMs Privacy-focused

Install and Start

Ollama runs models locally and exposes a simple API.

ollama serve

Download the model you want to use locally.

ollama pull mistral
ollama pull llama3
ollama pull deepseek-coder

Quick CLI interaction is enough for many experiments.

ollama run mistral
ollama run llama3

Pick models based on capability, speed, and system limits.

General chat: llama3
Code help: deepseek-coder
Smaller device: lighter models

Ollama also works well as a local backend for tools and scripts.

POST http://localhost:11434/api/generate

Local models are useful when data cannot leave the machine easily.

- private log analysis
- internal requirement drafts
- code explanation
- local prompt experiments

Local models bring privacy and control, but usually weaker quality than top hosted models.

Pros:
- privacy
- offline access
- lower marginal cost

Cons:
- hardware limits
- slower setup
- model quality varies

Ollama often works best when paired with LM Studio, editors, or local RAG components.

Good pairings:
- Ollama + VS Code
- Ollama + LangFlow
- Ollama + local vector store