Do I need to normalize Gemini embeddings?

Yes, if you use dimensions other than the default 3072, you must normalize the vectors manually before computing similarity scores to ensure accurate search results.

What vector dimensions does Gemini support?

The gemini-embedding-001 model supports flexible dimensions from 128 to 3072, with 768, 1536, and 3072 being the recommended standards for balancing performance and accuracy.

Can I process multiple texts in a single request?

Yes, the skill provides patterns for batch processing. However, it is recommended to keep batches under 100 texts to avoid known ordering bugs in the current API version.

What are the rate limits for Gemini embeddings?

The free tier supports 100 requests per minute (RPM), while paid tiers scale from 3,000 to 10,000 RPM depending on your account's billing history and usage tier.

Is this skill compatible with Cloudflare Workers?

Yes, it includes specific REST API fetch patterns designed for edge environments like Cloudflare Workers where the full Node.js SDK might not be available.

Google Gemini Embeddings

Name: Google Gemini Embeddings
Author: ma1orek

byma1orek

0•

Data Science & ML

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

This skill provides a production-ready implementation guide for the gemini-embedding-001 model, enabling developers to build sophisticated Retrieval-Augmented Generation (RAG) systems and semantic search engines. It covers advanced features like Matryoshka Representation Learning for flexible vector dimensions (768 to 3072), batch processing optimizations, and specific task-type configurations. Crucially, it includes documented workarounds for known API bugs, normalization requirements, and integration patterns for platforms like Cloudflare Vectorize to ensure reliable, scalable vector search performance.

Key Features

01Flexible dimensionality (768-3072) using Matryoshka Representation Learning

02Cloudflare Vectorize integration for production-grade RAG implementations

03Documented fixes for common SDK issues and batch ordering bugs

040 GitHub stars

05Batch processing patterns with built-in rate limit handling and error prevention

06Support for 8 specialized task types including retrieval, classification, and clustering

Use Cases

01Building semantic search engines with optimized retrieval performance

02Clustering large text datasets for topic modeling and automated categorization

03Implementing RAG systems for private documentation or knowledge bases

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add ma1orek/replay google-gemini-embeddings

For use in Claude.ai and ChatGPT

Download Skill