The desktop app starts the local server for you. For headless machines, scripts,
or remote dev boxes, the conifer CLI runs the same OpenAI-compatible server
directly.
conifer serve --model /path/to/model.gguf
# conifer serve: model "…" on http://127.0.0.1:8080/v1 — press Ctrl-C to stop| Flag | Default | Meaning |
|---|---|---|
--model <path.gguf> |
(required) | The GGUF weights to load and serve. |
--port <N> |
8080 |
TCP port to listen on. |
--host <iface> |
127.0.0.1 |
Interface to bind. Loopback by default. |
--backend <B> |
auto | Execution backend (e.g. metal, vulkan, cpu) where applicable to the platform. |
The server has no authentication. Binding loopback (127.0.0.1) keeps it
reachable only from your own machine. Set --host 0.0.0.0 only when you
intend to expose it on your network and have your own access controls in front
of it — otherwise you are publishing an unauthenticated inference endpoint.
Identical to the desktop server — see README.md and openai-compat.md. In short:
# OpenAI
curl http://127.0.0.1:8080/v1/chat/completions -d '{"model":"conifer","messages":[{"role":"user","content":"hi"}]}'
curl http://127.0.0.1:8080/v1/models
# Ollama-native (for editor auto-discovery)
curl http://127.0.0.1:8080/api/tags# Terminal 1 — start the server
conifer serve --model ~/models/qwen2.5-coder-7b-q4_k.gguf --port 8080
# Terminal 2 — point a tool at it
export OPENAI_BASE_URL=http://127.0.0.1:8080/v1
export OPENAI_API_KEY=local
# …run your agent / editor extension…