What are the main differences between Vectorize V1 and V2?

V2 offers 18x faster median latency (31ms), supports up to 5M vectors per index, introduces asynchronous mutations with mutationId, and requires metadata indexes to be created before data insertion.

Why isn't my metadata filtering working in V2?

In Vectorize V2, metadata indexes must exist before vectors are inserted. If vectors were added first, you must create the metadata index and then re-upsert the vectors for them to be filterable.

How do I check if an async mutation is complete?

You can use the env.VECTORIZE_INDEX.describe() method to retrieve the index info and check the 'processedUpToMutation' timestamp against the mutationId returned by your insert or delete operation.

How do I handle the dimension mismatch error in Vectorize?

Ensure your index dimensions match your embedding model exactly: 768 for Workers AI (bge-base-en-v1.5), 1536 for OpenAI text-embedding-3-small, or 3072 for OpenAI text-embedding-3-large.

Cloudflare Vectorize V2

Name: Cloudflare Vectorize V2
Author: jezweb

byjezweb

•

117

•

数据库管理

Implement high-performance semantic search and RAG workflows using Cloudflare Vectorize V2 with optimized metadata filtering and async mutation handling.

This skill provides a comprehensive implementation guide and set of patterns for Cloudflare Vectorize V2, the globally distributed vector database. It streamlines the creation of AI-powered applications by handling the critical breaking changes introduced in the GA release, such as asynchronous mutation tracking, advanced metadata indexing, and optimized query performance. It is essential for developers building Retrieval-Augmented Generation (RAG) systems, semantic search engines, or migrating legacy V1 indexes to the more scalable and faster V2 infrastructure while ensuring compatibility with Workers AI and OpenAI embeddings.

主要功能

01Advanced metadata filtering with range queries and nested property support

02Full support for Vectorize V2 GA features including async mutations and 5M vector limits

03Pre-built templates for RAG chatbots and document ingestion pipelines

04Deep integration with Workers AI (@cf/baai/bge-base-en-v1.5) and OpenAI models

05117 GitHub stars

06Automated metadata index configuration patterns to prevent filtering errors

使用场景

01Implementing a RAG (Retrieval-Augmented Generation) pipeline for AI chatbots using Cloudflare Workers

02Building a production-ready semantic search engine for high-scale documentation or product catalogs

03Migrating legacy Vectorize V1 indexes to the high-performance V2 infrastructure before deprecation

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jezweb/claude-skills cloudflare-vectorize

For use in Claude.ai and ChatGPT

Download Skill