@zzjcool/pi-custom-endpoint0.1.0
Pi provider for any compatible endpoint: log in with a base URL, API key, and protocol, and refresh its models automatically
Untested on omp
This is a pi package. No one has reported whether it works on omp yet. If you try it, add your result below.
Reports 0
No reports yet.
Readme
@zzjcool/pi-custom-endpoint
A Pi package that connects to any compatible endpoint. Log in once with a base URL, an API key, and a protocol. Pi refreshes the model list when it starts and whenever you open /model, and keeps the last list for offline use.
Install
pi install npm:@zzjcool/pi-custom-endpoint
Use
Run /login and choose Custom endpoint (URL + key):
- Base URL, for example
https://example.com/v1 - API key
- Protocol:
| Protocol | Typical endpoints |
|---|---|
| OpenAI Completions | one-api, new-api, vLLM, Ollama, LM Studio, LiteLLM |
| OpenAI Responses | OpenAI Responses API and compatible gateways |
| Anthropic Messages | Claude API and Anthropic-compatible proxies |
| Google Generative AI | Gemini API |
Models then appear under the custom-endpoint provider in /model.
Logging in again replaces the endpoint. Only one endpoint is active at a time.
Model metadata
The list comes from GET {baseUrl}/models. Fields are used when the endpoint provides them and fall back otherwise:
| Field | Fallback |
|---|---|
| Context window | 128000 |
| Max output tokens | 8192 |
| Reasoning | disabled |
| Image input | text only |
| Display name | model id |
Costs are reported as zero. Models whose id contains embedding are skipped.
Authentication follows the protocol: Authorization: Bearer for both OpenAI formats, x-api-key for Anthropic, and ?key= for Google.