Google Vertex Advanced Chat Model
AI-generatedSummary
Configure a Google Vertex AI chat model and expose it as an AI language model for an n8n chain or agent. The node uses the selected Google Cloud project and credential settings, and can attach billing labels and generation options to model requests.
Inputs
- Project ID (required) — Select a project from the list or enter its ID. The list is fetched using the Google API credential.
- Model Name — Model used to generate completions; defaults to gemini-2.5-flash. The implementation reports unsupported models unless their names start with “gemini”.
- Labels — Optional key/value string labels sent with requests for billing cost tracking; entries without a key are skipped.
- Maximum Number of Tokens — Maximum generated tokens; defaults to 2048.
- Safety Settings — Optional harmful-content categories and blocking thresholds to pass to Vertex AI.
- Sampling Temperature — Controls generation randomness; defaults to 0.4, with lower values producing less random output.
- Thinking Budget — Reasoning-token budget for thinking models; defaults to -1 for dynamic thinking, and 0 disables automatic thinking.
- Top K — Limits selection to higher-probability responses; the UI default is 32, while the implementation fallback is 40.
- Top P — Nucleus sampling setting for response diversity; defaults to 1 in the UI, while the implementation fallback is 0.9.
- Google API credential (required) — Provides Google account credentials and region used to configure Vertex AI.
Output shape
An AI language model connection, returned as a model object for an AI chain or agent to use.
This is a model provider node with no regular input or item output. It must be connected to an AI chain or agent. Generation results and tracing data are handled through the AI connection; project authorization, billing, and model errors may be reported during use.