Custom
Use this when the real endpoint is OpenAI-compatible but does not map cleanly to one of Loong’s built-in provider families.At A Glance
| Field | Value |
|---|---|
| Built-in kind | custom |
| Provider group | Gateway And Compatibility Providers |
| Protocol family | openai_chat_completions |
| Feature family | openai_compatible |
| Auth scheme | bearer |
| Credential envs | CUSTOM_PROVIDER_API_KEY |
| Aliases | openai_custom, custom_openai |
| Default base URL | https://<openai-compatible-host>/v1 |
| Request endpoint | https://<openai-compatible-host>/v1/chat/completions |
| Models endpoint | https://<openai-compatible-host>/v1/models |
Minimal Config
Verify It
list-models is unreliable for this account or region, pin an explicit provider.model or add preferred_models instead of leaving recovery implicit.
Auth And Routing Contract
| Contract | Value |
|---|---|
| Auth optional | no |
| Model probe auth optional | no |
| Default API key env | CUSTOM_PROVIDER_API_KEY |
| OAuth env | none |
| Primary request route | https://<openai-compatible-host>/v1/chat/completions |
| Primary model-catalog route | https://<openai-compatible-host>/v1/models |
Operator Notes
- custom requires tenant-scoped base_url configuration: replace
<openai-compatible-host>inprovider.base_urlwith your real OpenAI-compatible endpoint root such ashttps://api.example.com/v1; current template:https://<openai-compatible-host>/v1. - Alternative auth path: add
Authorization/X-API-Keyinprovider.headers. - Use
provider.headersonly when the upstream auth contract does not fit the default bearer-token path.
Related Docs
- Continue to Providers And Models when you still need the broader chooser page.
- Continue to Provider Guides for the full built-in provider matrix.
- Continue to Provider Recipes for representative rollout recipes.
- Continue to Configuration Patterns for the shared provider-profile shape.