Providers
Another server
A model server you run yourself, or a service with an OpenAI-compatible address, as a provider of its own.
Add one
- Open Settings → Providers and choose Another server.
- Type its address:
localhost:8080,gpu-box:8000or anhttps://address. Conch looks at it as you type and says what it found, like “llama.cpp, with 3 models”. - Give it a name, and a key if it asks for one. Press Add server.
Its models join the picker under the name you gave it. A server already running on this computer, on its usual port, is offered under Found on this computer: one press adds it.
Starting points
The page offers the ones people add most: llama.cpp, vLLM, Jan and LiteLLM on your computer, and Together AI, Fireworks, Hugging Face, NVIDIA and Venice online. Choose one and its address is filled in.
Good to know
- Plain http stays at home. It works on this computer and your own network, including your tailnet. Anywhere else, the address must be
https, so your key and your chats never cross the internet in the clear. - A key goes to that server only, and is kept like every key in Conch.
- On this computer means offline too. A server at
localhostanswers with no internet, and Conch can hand a chat to it when the internet drops. - Ollama and LM Studio on this computer have their own cards. If you type their address, Conch sends you there: they can do more through their own cards. One on another machine is added here like any server.
- Hugging Face and Venice need their key first. They list their models to anyone, so Conch asks for the key before calling them ready.
- A backup names its servers. Restoring a backup that adds a server shows it first, so an old or borrowed backup can’t quietly send your chats somewhere.