Building an E-commerce AI Agent with AWS Serverless: Building Product Search and RAG (3/7)

Building a RAG Pipeline for Semantic Code Search: A Developer Diary and Field Notes

Part 2: Storage, Search, and Agent Tooling In this blog series, we aim to share some of the lessons we learned over the past six months as we built and rolled out Air Context, a production-grade RA

Learn Vector Databases for Beginners – Embeddings, Similarity Search & ChromaDB

A vector database stores data as lists of numbers called embeddings, which lets it search by meaning instead of matching exact keywords. It's the piece behind a lot of recommendation systems, RAG apps

Announcing Web Answer on Amazon Bedrock AgentCore

Discover more about what's new at AWS with Announcing Web Answer on Amazon Bedrock AgentCore

Vercel Schedules is now in public beta

Vercel Schedules is now in public beta. Run tasks on a cron cadence or at a single point in time, dispatching to a Vercel Function or a Vercel Queue topic, in any Vercel environment. Schedules can also be used through dynamic schedules in Eve.

RAG vs. Vectorless RAG, clearly explained!

...explained visually (with tradeoffs in production).

Publish APIs, Set Usage Limits And Build A Developer Portal With Amazon API Gateway

Your company has data or a service that partners, customers or your own mobile app need to reach. You want a front door for it: a single address, with a key ...

10 Best AI GitHub Repositories to Supercharge Your Agent

GitHub là kho lớn nhất chứa công cụ AI miễn phí nhưng phần lớn lập trình viên chưa khai thác hiệu quả. Các agent như Claude Code thường được cài đặt mà không tận dụng được nguồn tài nguyên phong phú trên GitHub. Các kho lưu trữ AI hàng đầu như LangChain, LlamaIndex hay AutoGPT cung cấp framework mạnh để xây dựng agent thông minh hơn. Việc kết hợp các công cụ này có thể tăng đáng kể khả năng xử lý và tự động hóa của ứng dụng AI.

Bài viết này giúp bạn khám phá 10 kho GitHub AI mạnh nhất để tối ưu hóa khả năng của trợ lý lập trình.

Benchmarking Billion-Scale Vector Indexes

Benchmarking a vector index at billion-scale means tracking P99/P99.9 latency, Recall@K, QPS at a fixed recall target, index build time, RSS memory, and CPU usage, since averages hide failures. The team used three datasets for different jobs: SIFT-1B (128-d) to test raw scale, LAION-100M (768-d CLIP floats) to test real float behavior, and a self-built BookCorpus-10M (4096-d, via Llama 2) to approximate LLM-scale embeddings. Building BookCorpus-10M took 16 hours on five GPUs at about 173 vectors/sec, producing 164 GB that restores via cbbackupmgr in 2-3 minutes. Partition elimination, tagging vectors with a key like category_id, let a 10-partition layout skip 90% of the work per query, tightening P99 tail latency and raising throughput per core.