Connect an OpenAI-compatible provider
Ta treść nie jest jeszcze dostępna w Twoim języku.
Connect an OpenAI-compatible provider
Section titled “Connect an OpenAI-compatible provider”Use a Custom OpenAI Endpoint when the model server already exists outside Model Manager and exposes an OpenAI-compatible API.
This is the normal path for self-hosted servers, compatible gateways, and other services that LM Nexus should use without trying to start or stop them.
1. Open AI Providers settings
Section titled “1. Open AI Providers settings”Open Settings → AI Providers.
Under Provider settings, choose:
Custom OpenAI Endpoint
This selects which provider configuration you are editing. It does not automatically change the active provider in Chat or another module.
2. Add the provider
Section titled “2. Add the provider”Choose Add Provider.
Enter:
- Name — a recognizable local name, such as
vLLMorLab server; - Endpoint URL — the OpenAI-compatible API root, for example
http://127.0.0.1:8000/v1; - API Key (optional) — only when the endpoint requires one.
Choose Add Provider to save it.
3. Test the connection
Section titled “3. Test the connection”Use Test on the provider row, or Test All for all configured custom providers.
The provider table shows the saved endpoint and current status.
If the test fails, check:
- the external server is actually running;
- the URL includes the correct API root/path;
- the host/port are reachable from the LM Nexus backend environment;
- the API key is correct when required;
- the target really implements the OpenAI-compatible endpoints LM Nexus needs.
4. Select it where you want to use it
Section titled “4. Select it where you want to use it”Provider configuration and provider selection are separate.
After the endpoint is configured, choose that provider from the consumer that should use it, for example:
- Chat’s provider/model selection;
- an Agent profile’s Default or Fallback provider;
- another LM Nexus module that consumes AI Providers.
Custom provider versus managed local model
Section titled “Custom provider versus managed local model”Use Model Manager when LM Nexus should own the local GGUF/profile and
llama-server process.
Use a Custom OpenAI Endpoint when another service already owns the runtime.
This keeps LM Nexus from trying to start or stop a server it did not create.
Local network safety
Section titled “Local network safety”A working connection does not make an endpoint safe for public exposure.
If the service has no authentication or network access controls, keep it on a trusted local/private interface rather than exposing it directly to the internet.