What is the token limit for input text?

The gemini-embedding-001 model has an input limit of 2,048 tokens per text snippet.

What are the rate limits for the Gemini API free tier?

The free tier typically allows for 100 Requests Per Minute (RPM) and 30,000 Tokens Per Minute (TPM).

Can I use this with Cloudflare Workers?

Yes, this skill includes specific fetch-based patterns designed for edge environments where the full Node.js SDK may not be available.

What is the default dimension for Gemini embeddings?

The default is 3072 dimensions, but the model supports flexible output dimensionality (Matryoshka Representation Learning) from 128 to 3072.

When should I use the RETRIEVAL_QUERY task type?

You should use RETRIEVAL_QUERY specifically for user search strings to optimize the embedding for finding relevant content in your vector database.

Google Gemini Embeddings

Name: Google Gemini Embeddings
Author: jezweb

byjezweb

•

Data Science & ML

Generates high-performance text embeddings for RAG systems, semantic search, and document clustering using the Gemini API.

About

This skill provides a comprehensive implementation framework for integrating Google's gemini-embedding-001 model into your Claude Code workflows. It enables developers to build production-ready Retrieval-Augmented Generation (RAG) systems and semantic search features by leveraging Matryoshka Representation Learning for flexible vector dimensions ranging from 768 to 3072. The skill includes specialized guidance on task type optimization, batch processing to navigate rate limits, and seamless integration patterns for vector databases like Cloudflare Vectorize.

Key Features

Flexible dimensionality (768-3072) for optimized storage and performance
Comprehensive error handling for rate limits and token truncation
Specialized task types for queries, documents, and semantic similarity
Batch processing patterns for efficient large-scale document indexing
91 GitHub stars
Cloudflare Workers and Vectorize integration examples

Use Cases

Building Retrieval-Augmented Generation (RAG) systems for custom knowledge bases
Automating document clustering and topic modeling for large datasets
Implementing semantic search and duplicate detection in applications

About

Key Features

Flexible dimensionality (768-3072) for optimized storage and performance
Comprehensive error handling for rate limits and token truncation
Specialized task types for queries, documents, and semantic similarity
Batch processing patterns for efficient large-scale document indexing
91 GitHub stars
Cloudflare Workers and Vectorize integration examples

Use Cases

Building Retrieval-Augmented Generation (RAG) systems for custom knowledge bases
Automating document clustering and topic modeling for large datasets
Implementing semantic search and duplicate detection in applications