Prewave Vertex Chat Model icon

Prewave Vertex Chat Model

Gemini or Claude on Vertex AI as a chat model for the AI Agent, authenticated by a stored Google service account

Prewave Vertex Chat Model

AI-generated

Summary

Execute a chat model on Google Vertex AI, either Gemini or Claude, returning a language model response.

Inputs

  • provider — Select model provider (Google Gemini or Anthropic Claude)
  • model (required) — Model identifier; choose from list or enter ID directly
  • effort — Reasoning effort tier for Claude adaptive thinking (high, xhigh, max)
  • thinkingEnabled — Enable Claude adaptive thinking; when on, Vertex ignores temperature/top_p/top_k
  • maxOutputTokens — Maximum number of tokens to generate
  • temperature — Sampling temperature (0 to 2)
  • maxReasoningTokens — Max tokens reserved for internal reasoning (0 = model default, -1 = dynamic)
  • topK — Top K value for sampling
  • topP — Top P value for sampling

Output shape

a single AI language model response record

Contains generated text, token usage, and optionally tool calls; may be empty on error.

Examples

Example 1: Use default Gemini model with no extra options

provider: google; model: gemini-1.5-pro; options: {}

Example 2: Enable Claude thinking with high effort

provider: anthropic; model: claude-opus-4-8; options: {effort: high, thinkingEnabled: true}

Example 3: Custom model ID with temperature and topP

provider: google; model: gemini-2.5-flash; options: {temperature: 0.8, topP: 0.9}

Links

Discussion