google-gemini-embeddings

Semantic search uses embeddings to find conceptually similar content by converting text into vectors. This Skill generates Gemini embeddings optimized for retrieval tasks, enabling you to index documents and match user queries based on meaning rather than keyword matching.

Embeddings convert text into numerical vectors that capture semantic meaning. This Skill supports 128-3072 dimensions using Matryoshka learning—smaller dimensions reduce storage and latency for edge deployments like Cloudflare Vectorize, while larger dimensions improve retrieval accuracy for complex queries.

Yes. This Skill provides production-ready RAG workflows combining document ingestion, vector search, and context-aware generation. It handles task-type optimization, batch processing, and rate limiting to retrieve relevant documents that augment language model responses.

This Skill offers specialized task types including RETRIEVAL_QUERY, RETRIEVAL_DOCUMENT, SEMANTIC_SIMILARITY, and CLUSTERING. RETRIEVAL_QUERY optimizes for matching user queries, RETRIEVAL_DOCUMENT for indexing content, SEMANTIC_SIMILARITY for comparing text pairs, and CLUSTERING for grouping similar documents.

Yes. This Skill integrates with Cloudflare Vectorize to deploy embeddings at the edge, enabling low-latency semantic search across distributed datasets. It handles dimension matching and batch processing patterns for edge environments.

This Skill enforces a 2048-token input limit per text and includes batch processing with rate limiting. It provides robust error handling and guidance on matching output dimensions to your vector database, supporting both SDK and REST integration patterns.