LM STUDIO / OLLAMA is not setup to allow for any use case

[cdvankammen]

Description

There should be the ability to use ip address's not just local host. This can be on a vps, lxc conatiner or small computer that uses a main omputer for offline llms

Expected Behavior

Allow for all forms and easy setup (open claw has this feature now as well could be used for reference).

Init logs are not setting up themselves to run this after reboot automatically either

Steps to Reproduce

init >>>> lm studio local only

OpenFang Version

any

Operating System

Linux (x86_64)

Logs / Screenshots

No response

[jaberjaber23]

Fixed locally in commits 6bed6c0 and b00af5e. Will ship in the next release.

You can now point Ollama or LM Studio at any host (VPS, LXC, LAN, another box) without editing config.toml. Three options:

1. Env vars (new, recommended):

   • Ollama: OLLAMA_HOST=192.168.1.50:11434 or OLLAMA_BASE_URL=http://192.168.1.50:11434/v1
   • LM Studio: LMSTUDIO_HOST=10.0.0.5:1234 or LMSTUDIO_BASE_URL=...
   • vLLM: VLLM_HOST=... / VLLM_BASE_URL=...
   • Lemonade: LEMONADE_HOST=... / LEMONADE_BASE_URL=...
     Scheme defaults to http:// and /v1 is appended if missing, so bare host:port works.

2. Dashboard: Settings -> Providers -> set custom base URL (already worked, persisted to [provider_urls]).

3. ~/.openfang/config.toml:

   [provider_urls]
   ollama = "http://192.168.1.50:11434/v1"
   

Precedence: explicit [provider_urls] > env vars > built-in localhost default.

Re: the "Init logs are not setting up themselves to run this after reboot" part, that is a separate concern (systemd unit / autostart) and is not covered by this fix. Open a separate issue if you want autostart-on-reboot tracked.

Code refs:

• crates/openfang-runtime/src/drivers/mod.rs:38-101 (local_provider_url_from_env)
• crates/openfang-runtime/src/drivers/mod.rs:540-549 (driver fallback chain)
• crates/openfang-runtime/src/model_catalog.rs:342-361 (apply_local_env_overrides)
• crates/openfang-kernel/src/kernel.rs:823-825 (boot wiring)