Ollama
on your own machine
Ollama runs open large language models such as Llama, Gemma and Qwen on the CPU, behind an HTTP API. Downloads the chosen models at start; its API has no login, so share the address only with apps you trust.
$12 a month per machine, with as many servers as it can run. Already on Raptor? It's in your team's Eggs, as software/ollama.
What Raptor does for it
- Backups
- Incremental and encrypted on your machine, every day by default. Restore one file or the whole server.
- Runs on
- x86 and ARM machines, on Raptor's Debian image.
- Every server
- A live console, a file manager with a trash, schedules, crash restarts, alerts, and SFTP. It keeps running if Raptor goes down.
Settings you can change
From the server's Startup tab, without touching a file.
- Version
- The Ollama version to install, such as 0.40.16, or latest. Reinstall the server to change it.Default
latest - Models
- Models from ollama.com/library to download when the server starts, separated by spaces, such as llama3.2:1b qwen3:4b. Models you already have aren't downloaded again.Default
llama3.2:1b - Keep loaded
- How long a model stays in memory after its last request, such as 5m or 1h; -1 keeps it loaded.Default
5m - Parallel requests
- How many requests each model answers at once. More needs more memory.
- Context length
- The context window in tokens, unless a request sets its own. Empty lets Ollama choose from the memory it has.
- Allowed origins
- Extra web origins allowed to call the API from a browser, separated by commas, such as your chat app's address.
- Cloud features
- Allow Ollama's cloud models and web search, which send prompts to ollama.com.
From a bare machine
Link your machine
One command on any Debian or Ubuntu box, at home or rented.
Pick Ollama
Raptor installs it and starts it, on images it keeps updated.
Open it
At an address your machine comes with, or your own domain.
Made by Raptor and MIT-licensed. Written for Raptor from Ollama's documentation (https://github.com/ollama/ollama, MIT) Icon: the Ollama logo, from https://github.com/ollama (Ollama, Inc.) The egg's source.