NVIDIA NIM Chat Model
AI-generatedSummary
Instantiate an OpenAI‑compatible chat model from NVIDIA NIM using the supplied parameters and return the configured model.
Inputs
- baseUrl (required) — The base URL for the NVIDIA NIM service. Change this if using a self‑hosted instance.
- model (required) — The model to use. Choices are fetched dynamically from the NIM API.
- temperature — Controls randomness: higher values mean more creative but less predictable responses.
- maxTokens — The maximum number of tokens to generate.
- additionalParameters — Additional sampling parameters (e.g., topP, frequencyPenalty, presencePenalty, timeout).
Output shape
a model instance returned as { response: <model> }
Only the model is returned; no further data is provided; errors are propagated by the node.
Examples
Example 1: Generate a creative response using the `dolly-v2-3b` model with temperature 1.0 and maxTokens 768
baseUrl: https://integrate.api.nvidia.com/v1, model: dolly-v2-3b, temperature: 1.0, maxTokens: 768, additionalParameters: {topP:0.9}
Example 2: Classify text with low randomness using the `nemotron-4-340b` model, temperature 0.2, maxTokens 256
baseUrl: https://integrate.api.nvidia.com/v1, model: nemotron-4-340b, temperature: 0.2, maxTokens: 256, additionalParameters: {frequencyPenalty:0.5, presencePenalty:0.5}