Ollama (Local)

Ollama can run downloaded model files on your machine. In Sythoria v0.4.4 the preset connects to Ollama on loopback, so ordinary model prompts do not require a hosted model provider.

Prerequisites#

Install Ollama from the official Ollama download page.
Confirm the service is running.
Download a model that fits your available RAM or VRAM.

Review installation scripts before running them; this guide does not recommend piping a remote script directly into a shell.

Install the preset model#

Sythoria v0.4.4 defaults to deepseek-r1:7b:

ollama pull deepseek-r1:7b
ollama list

Configure Sythoria#

Setting	Value
API base	`http://localhost:11434/v1/chat/completions`
Default model	`deepseek-r1:7b`
API format	OpenAI-compatible chat completions
API key	Not required for the default local service

Open Settings and add a model.
Select Ollama (Local).
Confirm the API base and exact installed model tag.
Save and run the connection check.

Sythoria's native backend connects to loopback directly. Do not configure a wildcard browser origin for normal desktop use.

Remote Ollama hosts#

A remote or LAN Ollama service is no longer an offline boundary. It can expose prompts and models to other devices on that network.

Bind the service only to the interface that needs it.
Restrict access with a firewall and private network.
Put TLS and authentication in front of the service before crossing an untrusted network.
Enable Sythoria local-network access only for a host you control and trust.
Never expose an unauthenticated Ollama port directly to the public internet.

Troubleshooting#

ollama list
ollama ps

If Sythoria cannot connect, confirm that the service is listening on localhost:11434, the model tag matches exactly, and local-network policy has not been unnecessarily broadened. A model that is too large can load slowly or exhaust memory; choose a smaller installed tag rather than changing network security settings.