Why do I need to create metadata indexes before inserting vectors?

In Vectorize V2, metadata properties are only filterable if the metadata index exists at the time of insertion. Vectors added before an index is created must be re-upserted to be searchable.

What metadata filters are supported?

Vectorize V2 supports equality, inequality, inclusion ($in/$nin), and range queries for both numbers and strings, which are all covered by this skill.

Does this skill support OpenAI embedding models?

Yes, it includes specific configurations for text-embedding-3-small and large, ensuring the Vectorize index dimensions (1536 or 3072) match the model output exactly.

What are the main differences between Vectorize V1 and V2?

V2 offers significantly higher capacity (up to 5M vectors), lower latency (31ms median), and introduces asynchronous mutations and a new string-based returnMetadata enum.

How do I handle asynchronous mutations in V2?

V2 returns a mutationId for every write operation. You can track the status of your data by using the describe() method to check the processedUpToMutation timestamp.

Cloudflare Vectorize Implementation

Name: Cloudflare Vectorize Implementation
Author: jezweb

byjezweb

•

データベース管理

Implements high-performance semantic search and RAG workflows using Cloudflare Vectorize V2.

概要

This skill provides a comprehensive framework for building AI-powered applications on Cloudflare’s globally distributed vector database. It streamlines the transition from Vectorize V1 to V2, addressing critical breaking changes such as asynchronous mutations and mandatory metadata indexing. Developers can leverage this skill to implement efficient document chunking, embedding generation via Workers AI or OpenAI, and advanced metadata filtering for production-ready Retrieval Augmented Generation (RAG) systems while avoiding common pitfalls like dimension mismatches and timing issues.

主な機能

Native integration patterns for Cloudflare Workers AI and OpenAI embedding models.
Comprehensive migration guides for transitioning legacy V1 indexes to high-performance V2.
Full support for Vectorize V2 breaking changes including async mutations and returnMetadata enums.
91 GitHub stars
Advanced metadata filtering implementations using range queries and nested dot notation.
Pre-built templates for document ingestion, RAG chatbots, and semantic search pipelines.

ユースケース

Building context-aware RAG chatbots with Cloudflare Workers and Vectorize.
Implementing lightning-fast semantic search for large-scale documentation or product catalogs.
Migrating existing vector workflows to Cloudflare's 5-million-vector-per-index V2 architecture.

概要

主な機能

Native integration patterns for Cloudflare Workers AI and OpenAI embedding models.
Comprehensive migration guides for transitioning legacy V1 indexes to high-performance V2.
Full support for Vectorize V2 breaking changes including async mutations and returnMetadata enums.
91 GitHub stars
Advanced metadata filtering implementations using range queries and nested dot notation.
Pre-built templates for document ingestion, RAG chatbots, and semantic search pipelines.

ユースケース

Building context-aware RAG chatbots with Cloudflare Workers and Vectorize.
Implementing lightning-fast semantic search for large-scale documentation or product catalogs.
Migrating existing vector workflows to Cloudflare's 5-million-vector-per-index V2 architecture.