CodeBuddy with Nbility
CodeBuddy releases that support custom models can register OpenAI-compatible models in user-level models.json. Confirm that your IDE or CLI release documents this schema before applying it.
Add the model configuration
Edit ~/.codebuddy/models.json, or .codebuddy/models.json in your Windows user directory. Merge with the existing models array:
{
"models": [
{
"id": "YOUR_MODEL_ID",
"name": "Nbility model",
"vendor": "Nbility",
"apiKey": "${NBILITY_API_KEY}",
"url": "https://api.nbility.ai/v1/chat/completions",
"maxInputTokens": 8192,
"maxOutputTokens": 1024,
"supportsToolCall": true,
"supportsImages": false
}
]
}
Replace the ID and adjust limits and capabilities to the real model. The numbers are illustrative initial limits; enable supportsToolCall only for a model that actually supports tools.
Supply credentials
Set export NBILITY_API_KEY="YOUR_API_KEY" and restart CodeBuddy from that terminal. A desktop shortcut may not inherit shell variables; use the client’s custom-model credential UI in that case.
A project .codebuddy/models.json can override user settings. Save valid UTF-8 JSON and keep real keys out of project repositories.
Verify and distinguish URL fields
Select the Nbility model, send a short message in a new conversation, and check Usage logs.
The file’s url requires the full /v1/chat/completions endpoint. A graphical Base URL field that appends the endpoint instead needs https://api.nbility.ai/v1. Read the field definition before transferring settings between interfaces.
If your version does not load models.json, use its documented custom-model UI or choose Cline.