How to Configure an OpenAI-Compatible Provider in Navin
Navin can call OpenAI-compatible model providers by setting a base URL, optional API key, and a model configuration that references that provider - from Settings.
What you will build
- a custom provider entry in Settings
- a model configuration pointing at that provider
- one successful chat reply
When to use this
Use this for local or hosted services that expose OpenAI-compatible endpoints, including internal gateways, local model servers, and provider proxies that are not already named in Navin.
Configure in Settings
- Open Navin → Settings → Providers.
- Add a custom (or named custom) provider.
- Paste the API key if required, and set the base URL (include
/v1when the service expects it). - Open Settings → Models, add a configuration that uses that provider and the model ID the service serves.
- Set the configuration Active.
- Send a short test message in chat.
Advanced JSON shape (optional)
{
"providers": {
"custom": {
"apiKey": "${CUSTOM_API_KEY}",
"apiBase": "https://api.example.com/v1"
}
},
"modelPresets": {
"primary": {
"label": "Custom",
"provider": "custom",
"model": "provider-model-name",
"maxTokens": 4096,
"contextWindowTokens": 65536,
"temperature": 0.1
}
},
"agents": {
"defaults": {
"modelPreset": "primary"
}
}
}
Prefer Settings for everyday setup. Confirm the provider's own dashboard or docs list the model ID before debugging Navin.
Production notes
- Include the version path in the base URL when the service expects
/v1. - Use separate provider names for separate endpoints.
- Use a placeholder key such as
EMPTYonly when the endpoint requires a non-empty key but does not validate it. - Leave OpenAI
apiTypeunset for ordinary OpenAI-compatible custom endpoints.
Security notes
- Keep provider keys in Settings or OS environment variables.
- Treat internal model gateways as sensitive network services.
- Do not point Navin at untrusted proxy endpoints for private projects.
Troubleshooting
- Endpoint unreachable: fix the provider service or base URL before changing Navin.
- Model unknown: check the model ID expected by the provider.
- Auth fails: confirm Bearer auth requirements and that the key is saved in Settings for this install.