OpenRouter, NIM & Groq

Configuring alternative cloud inference providers in NIKI.

OpenRouter, NVIDIA NIM & Groq

NIKI supports any OpenAI-compatible provider through custom base_url settings.

OpenRouter routes to hundreds of open and proprietary models under a single API key:

[providers.openrouter]
base_url      = "https://openrouter.ai/api/v1"
api_key       = "sk-or-v1-..."
default_model = "deepseek/deepseek-r1"

Run enterprise accelerated open models with NVIDIA NIM:

[providers.nvidia]
base_url      = "https://integrate.api.nvidia.com/v1"
api_key       = "nvapi-..."
default_model = "meta/llama-3.3-70b-instruct"

Ultra-low latency inference for high-speed coding passes:

[providers.groq]
base_url      = "https://api.groq.com/openai/v1"
api_key       = "gsk_..."
default_model = "llama-3.3-70b-versatile"