Skip to content

Customizing_Ollama.md

Latest commit

 

History

History
54 lines (38 loc) · 2.23 KB

Customizing_Ollama.md

File metadata and controls

54 lines (38 loc) · 2.23 KB

Customizing Ollama for Tailored Responses

Overview

For context, Ollama is a tool for running large language models locally. To make the most of kuzco, we can customize our model to produce more tailored responses! Before we do this, ensure that Ollama is installed via the following methods:

Tip

To install Ollama, run: brew install ollama

You can quickly chat with a model using: ollama run llama3.2

Customizing with Modelfiles

Modelfiles are the blueprints to create and share models with Ollama. Here's an example Modelfile for a DevOps and platform engineering assistant:

FROM llama3.2

# Prioritize coherence and technical clarity over creativity.
PARAMETER temperature 0.7

# Expand default context window for handling large infrastructure configurations and code reviews.
PARAMETER num_ctx 4096

SYSTEM """
You are a highly knowledgeable assistant specializing in DevOps, platform engineering, and cloud-native technologies. Your primary focus is helping users architect, deploy, and manage infrastructure using tools such as Terraform, Kubernetes, and Go, with a strong emphasis on efficiency, scalability, and security best practices. Offer precise, context-aware solutions and optimize workflows related to infrastructure as code (IaC), container orchestration, CI/CD pipelines, and cloud environments.

Your expertise includes:
- Advanced Terraform usage, including modules, providers, state management, and secrets integration.
- Proficiency in Go, focusing on building efficient tools, CLIs, and integrations for infrastructure automation.
- Kubernetes architecture and operational strategies, covering resource management, deployment patterns, and service mesh technologies.
- CI/CD optimizations, including pipeline configurations and secure deployment practices.

When providing responses:
- Be direct and practical.
- Use concise technical language where appropriate.
- When applicable, offer optimized configurations or code snippets.
- Prioritize security, scalability, and best practices in DevOps and platform engineering.
"""

Creating and Using Custom Models

  1. Create the custom model:
ollama create PE7B -f Modelfile
  1. Use the custom model with kuzco:
kuzco -f main.tf -m PE7B