Vertex AI icon

Vertex AI

Google Vertex AI Gemini API

Actions3

Multimodal

AI-generated

Summary

Process text and images together using a Gemini model via Google Vertex AI

Inputs

  • model (required) — Gemini model to use
  • text — The text to send with the image
  • imageSource (required) — Source of the image: binary, url, or base64
  • binaryProperty — Name of the binary property containing the image (shown when imageSource is binary)
  • imageUrl — Image URL (shown when imageSource is url)
  • base64Image — Base64 encoded image data without data: prefix (shown when imageSource is base64)
  • options — Generation config: maxOutputTokens, temperature, topP, topK, systemInstruction, thinkingLevel, timeout, enableDebug
  • responseFormat — Response format: text/plain, application/json, or text/x.enum
  • schemaMode — Schema input mode: simple UI or advanced JSON Schema
  • enumValues — Comma-separated enum values (shown when responseFormat is text/x.enum and schemaMode is simple)
  • schemaProperties — JSON response properties (shown when responseFormat is application/json and schemaMode is simple)
  • responseSchema — Raw JSON schema (shown when responseFormat is application/json or text/x.enum and schemaMode is advanced)

Output shape

object

Each returned item contains: text (generated response), optional json (parsed if responseFormat is application/json and parsing succeeds), model, operation, usage metadata, safety ratings, finish reason, and optional _debug info when enableDebug is true.

Examples

Example 1: Describe an image using a Gemini model

Set model to gemini-2.5-flash, imageSource to url, provide an imageUrl, leave text empty or provide a caption prompt.

Links

Discussion