skills.rest
masermediagroup-stack
masermediagroup-stackCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

using-vector-databases

Implement vector databases, embeddings, and hybrid search for RAG systems.

1|Updated Feb 24, 2026
Search Tags:#semantic search#embeddings#rag#vector database#hybrid search#qdrant#chunking
One-click install
npx skills add https://github.com/masermediagroup-stack/maser-media --skill using-vector-databases-masermediagroup-stack
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: using-vector-databases
Source: https://github.com/masermediagroup-stack/maser-media/tree/main/.cursor/skills/community/ai-design-components/skills/using-vector-databases
Command: npx skills add https://github.com/masermediagroup-stack/maser-media --skill using-vector-databases-masermediagroup-stack

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires qdrant-client, openai, numpy, ragas, datasets, tqdm, psycopg2-binary, pgvector, and includes scripts (resource) and references (resource) components.

What problem does it solve? Building semantic search and RAG pipelines requires choosing among many vector databases, embedding models, and chunking strategies, and mistakes lead to poor retrieval quality, high costs, and slow queries. ## Core Features & Use Cases - Decision Frameworks: Compare Qdrant, Pinecone, Milvus, pgvector, and Chroma, plus embedding providers like OpenAI, Voyage, Cohere, and self-hosted models with cost and quality trade-offs. - Production Patterns: Covers chunking strategies, hybrid search with Reciprocal Rank Fusion, metadata filtering, re-ranking, and RAGAS-based evaluation. - Executable Scripts: Includes scripts for batch embedding generation, similarity benchmarking, and automated RAG evaluation. - Use Case: Build a documentation chatbot by chunking docs at 512 tokens, embedding with OpenAI, storing in Qdrant, and retrieving with hybrid vector plus BM25 search. ## Quick Start Ask the AI to set up a Qdrant-based RAG pipeline with hybrid search for your documentation using this skill.

Frequently Asked Questions about using-vector-databases

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I choose a vector database for a RAG application?▼

Choose based on scale and infrastructure: Qdrant for under 100M vectors with complex metadata filtering, pgvector if you already use PostgreSQL, Pinecone for fully managed zero-ops, and Milvus for over 100M vectors with GPU acceleration.

What is hybrid search and why use it in RAG?▼

Hybrid search combines vector similarity with BM25 keyword matching, then merges results using Reciprocal Rank Fusion. It improves retrieval quality by 15-30% over vector-only search because semantic search catches paraphrases while keyword search ensures exact technical term matches.

OpenAI vs Voyage vs Cohere embeddings: which should I use?▼

Voyage AI voyage-3 scores highest on MTEB benchmarks at about $0.12 per million tokens. OpenAI text-embedding-3-large offers enterprise reliability, text-embedding-3-small is 6x cheaper for budget constraints, and Cohere embed-v3 supports over 100 languages.

What chunk size should I use for document embedding?▼

Use 512 tokens with 50 tokens of overlap as the default for most RAG systems. Smaller chunks fragment concepts while larger chunks dilute relevance; code files should instead be split at function or class boundaries using AST-based chunking.

How do I evaluate RAG pipeline quality?▼

Use RAGAS metrics: faithfulness above 0.90 to minimize hallucination, answer relevancy above 0.85, context recall above 0.80, and context precision above 0.75. The included evaluate_rag.py script runs these metrics against a test dataset automatically.

Why is my vector search returning irrelevant results?▼

Poor retrieval usually stems from wrong chunk sizes, overly restrictive metadata filters, or vector-only search. Switch to hybrid search, add a cross-encoder re-ranking stage, and verify your chunking strategy matches your content type.

Related Skills

Explore More Software Engineering Skills→
HosungYouHosungYou
5

i3

Download PDFs, extract text, chunk documents, and create a local ChromaDB vector database.

Community
Advanced
JhabbigJhabbig

rag-implementation

Build retrieval-augmented generation systems that ground LLM answers in external knowledge.

Community
Advanced
masermediagroup-stackmasermediagroup-stack
1

embedding-optimization

Optimize vector embeddings for RAG systems through model selection, chunking, caching, and monitoring.

Community
Advanced
nilecuinilecui
27

hybrid-search-implementation

Implement hybrid search pipelines combining vector and keyword results with RRF fusion.

Community
Advanced