Ollama is different from every other provider in SynaPost: it’s not a hosted API, it’s software you run yourself, so there’s no API key at all – just a local address SynaPost talks to.
1. Install and run Ollama
Download Ollama from ollama.com for your OS and install it, then pull at least one model from the command line, for example ollama pull llama3.1.
Leave Ollama running – it serves an API on your machine in the background (by default at http://localhost:11434) the whole time SynaPost needs to reach it.
2. Point SynaPost at it
- Go to Settings → AI & Images and click Configure on the Ollama card.
- Enter the Base URL where Ollama is reachable. If SynaPost and Ollama are on the same machine, the default
http://localhost:11434is normally correct as-is; if Ollama runs on another machine or server, use that machine’s real address instead. - Click Test connection. SynaPost checks that the URL is actually reachable before it will let you save it.
- Once the test succeeds, save – only then does the Ollama card show as Connected.
This verify-then-save order matters: just typing a URL into the field, without a successful test, never marks Ollama as connected. That’s deliberate, so the dashboard never shows a false “Connected” for a model server that isn’t actually reachable.
3. Choose a model
Enter the exact model name as Ollama itself knows it (matching what you pulled, e.g. llama3.1 or mistral) as part of the same setup step. You can pull additional models with Ollama at any time and switch by reconfiguring – no need to disconnect first.
Why choose Ollama
Everything runs on hardware you control, nothing leaves your machine or server, and there’s no per-token bill from a hosted API – the trade-off is that output quality and speed depend entirely on the model you choose and the hardware running it, and you’re responsible for keeping Ollama itself running.
Ollama is a paid-plan provider – see Choosing an AI Provider for what’s available on every plan.