Custom providers and models
Use a custom provider when you have a compatible endpoint or a model missing from the built-in catalog.
Provider definition
Section titled “Provider definition”A custom provider requires:
- a lowercase slug ID, 1–64 characters;
- display name and endpoint URL;
- a supported pi-ai API transport;
- optional headers and compatibility settings;
- provider authentication where required.
The UI supports the transports Nerve deliberately registers with pi-ai. “OpenAI-compatible” is not a guarantee that every server variant implements streaming, tools, reasoning, images, or usage identically.
Model definition
Section titled “Model definition”Choose Add model, select an already configured or authenticated provider, and paste one pi-compatible model object into the JSON editor. Copy the object inside the provider’s models array—not the outer providers configuration—from pi.dev/models, or edit the template Nerve provides.
{ "id": "my-model", "name": "My Model", "reasoning": false, "input": ["text"], "contextWindow": 128000, "maxTokens": 16384, "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 }}Nerve accepts pi model metadata such as thinkingLevelMap, compat, model headers, sampling parameters, and tiered costs. Invalid JSON, unknown fields, and full provider configurations are rejected instead of being silently discarded.
Provider URLs, API types, headers shared by every model, and API keys remain in provider settings. The runtime inherits those connection fields from the selected custom provider or built-in provider template. Test a new definition with read-only permission before granting mutation.
Removal
Section titled “Removal”Deleting a custom provider cascades its manually defined models and provider secret. Export or record configuration you need before deletion.