Ollama

Ollama is a local LLM backend that serves open-source language models for inference. It provides the model runtime for Open WebUI and supports a growing library of models including llama3.2, Mistral, Gemma, and nomic-embed-text for embeddings.

Key Features

• Local model serving: Run open-source LLMs locally without external API dependencies.
• Model library: Download and serve models from a curated library (llama3.2, Mistral, Gemma, Phi, and more).
• REST API: Full API for chat completions, embeddings, and model management.
• GPU acceleration: Supports NVIDIA GPU acceleration via CUDA for faster inference.
• Lightweight: Minimal resource footprint for CPU-only deployments.

Integration with openDesk Edu

Ollama is part of the Collab Services suite and deploys via its upstream Helm chart (ollama.github.io). It is deployed first in the Helmfile dependency chain (stage 010-infra) as the LLM backend that Open WebUI depends on. It runs as an internal service not directly exposed to users.

Learn More

• Official Documentation — Ollama docs and resources
• Model Library — Available open-source models