Gemini Embeddings

Generate vector embeddings using Google Gemini embedding models.

Overview

Gemini Embeddings uses the embedContent API to generate dense vector representations of text via Google AI. Supports the Gemini embedding models with configurable task types (RETRIEVAL_QUERY, RETRIEVAL_DOCUMENT, SEMANTIC_SIMILARITY, CLASSIFICATION, CLUSTERING). Optionally accepts a title for RETRIEVAL_DOCUMENT tasks. Each input item produces a float embedding vector placed in a configurable output field.

Category: AI
Tool Name: gemini_embeddings
Version: 1

Appearance: Icon: gemini | Color: #ffffff

Node Type

Action — processes input items and produces output

Input / Output

Direction	Port(s)
Input	Input
Output	Output, Error

Credentials

This tool requires Google AI credentials. See the Credentials Guide for setup instructions.

Parameters

Parameter	Type	Required	Default	Description
Model	options	No	(current default)	The Gemini embedding model to use.
				Options: the Gemini embedding models available to your workspace — pick one from the dropdown.
Text	string	Yes	—	The text to generate an embedding for. If empty, falls back to item.json.text. Supports expressions like {{ $json.chunk }}.
Task Type	options	No	RETRIEVAL_DOCUMENT	The downstream task type to optimize the embedding for.
				Options: RETRIEVAL_QUERY (optimize for search query matching), RETRIEVAL_DOCUMENT (optimize for document indexing), SEMANTIC_SIMILARITY (optimize for comparing text similarity), CLASSIFICATION (optimize for text classification tasks), CLUSTERING (optimize for grouping similar texts)
Title	string	No	—	Optional title for the text. Only used when Task Type is RETRIEVAL_DOCUMENT. Supports expressions.
Options	collection	No	{}	Optional output settings.
— Response Field Name	string	No	embedding	Field name in the output JSON where the embedding vector will be placed.
Include Input	boolean	No	false	Whether to include the original input item fields in the output alongside the embedding.
Max Concurrency	number	No	10	Maximum number of items to process concurrently.

Output Data

One embedding per input item, and one output item per input item. The vector lands on the field named by Response Field Name (embedding by default), with model beside it. The rest of the input item JSON is dropped unless Include Input is on; binary data on the input item is forwarded unchanged.

{
  "embedding": [0.0123, -0.0456, 0.0789],
  "model": "the embedding model that was used"
}

• The vector is a plain array of floats at the model’s native length. Nothing else about the call is returned — there is no usage block on this node.
• Turn Include Input on when you are writing vectors to a store and need the source text or an ID to travel with them.
• Index and query with matching task types: embed your corpus with RETRIEVAL_DOCUMENT and the incoming question with RETRIEVAL_QUERY. Vectors written under mismatched task types still compare, but retrieval quality drops.

Reference the result downstream by expression, e.g. {{ $json.embedding }}.

Usage Examples

• Generate embeddings for document chunks using Gemini
• Embed search queries with RETRIEVAL_QUERY task type
• Create document embeddings with title metadata
• Vectorize text for semantic similarity comparison
• Embed content for classification downstream tasks

Example Configuration

Embed the text field of each incoming item for indexing:

{
  "type": "gemini_embeddings",
  "parameters": {
    "text": "{{ $json.text }}",
    "taskType": "RETRIEVAL_DOCUMENT"
  }
}

Index document chunks with a title, keeping the source fields for the vector store:

{
  "type": "gemini_embeddings",
  "parameters": {
    "text": "{{ $json.chunkText }}",
    "taskType": "RETRIEVAL_DOCUMENT",
    "title": "{{ $json.documentTitle }}",
    "includeInput": true,
    "maxConcurrency": 10
  }
}

Embed an incoming question before a similarity lookup, under a field name the store expects:

{
  "type": "gemini_embeddings",
  "parameters": {
    "text": "{{ $json.userQuestion }}",
    "taskType": "RETRIEVAL_QUERY",
    "options": {
      "responseFieldName": "queryVector"
    }
  }
}

Error Handling

Mode	Behavior
stop	Halts workflow on first error
continue	Skips failed items, passes successful ones through
errorPort	Routes failed items to Error output port

Tips

Gemini Embeddings converts text into dense vector representations using a Google AI Gemini embedding model via the embedContent API. Use this tool when your workflow requires semantic search, document retrieval, similarity ranking, classification, or clustering and needs text transformed into a numerically comparable format. Each input item produces a float embedding vector placed in a configurable output field, with processing failures routed to the error output.