opensearch-skills

OpenSearch Skills provides a modular toolkit for constructing search applications with advanced retrieval strategies including vector and semantic search, plus observability features for log analysis and trace investigation. Deploy locally or to AWS, process documents at scale, and route queries intelligently across search, observability, and cloud infrastructure tasks.

OpenSearch Skills helps you build search applications with vector and semantic search, plus log analytics and AWS deployment capabilities.

AI-generated summary based on this skill's SKILL.md

★ 37 ⁮ 30 Apache-2.0 updated 2026-07-22 by opensearch-project

Install

opensearch-project/opensearch-agent-skills/opensearch-skills · repository language: Python

Manual (git clone) CLI (skillfed)

git clone https://github.com/opensearch-project/opensearch-agent-skills
cp -r opensearch-agent-skills/skills/opensearch-skills ~/.claude/skills/opensearch-skills

npx skillfed install opensearch-project/opensearch-agent-skills/opensearch-skills

Frequently asked questions

AI-generated answers based on this skill's SKILL.md and metadata

How do I build a search application with OpenSearch?

OpenSearch Skills helps you construct search applications by providing modular components for vector search, semantic search, and hybrid retrieval. Start by setting up an OpenSearch cluster (locally via Docker or on AWS), define your index schema with appropriate field mappings, ingest your documents, and configure search queries using BM25, dense vectors, or hybrid approaches that combine multiple ranking strategies for optimal relevance.

What are OpenSearch vector search and semantic search setup steps?

OpenSearch Skills supports both vector and semantic search through k-NN similarity search and dense vector embeddings. To set up: create an index with a vector field using the appropriate dimension and similarity metric (cosine, L2, or inner product), generate embeddings for your documents using an embedding model, ingest the documents with their vector representations, and query using the k-NN search API or combine vectors with BM25 for hybrid search.

How can I ingest and query logs for observability with OpenSearch?

OpenSearch Skills enables log analytics through PPL (Piped Processing Language) queries and trace analytics for debugging. Ingest logs using Fluent Bit or Logstash, structure them as JSON documents in OpenSearch indices, then query error patterns, trace events, and system health using PPL syntax or standard queries. The observability stack helps you correlate logs and traces to investigate issues across your infrastructure.

How do I set up OpenSearch for RAG with document embeddings?

OpenSearch Skills supports RAG (Retrieval-Augmented Generation) workflows by processing documents at scale and indexing them with embeddings. Convert documents (PDFs, text files) to JSONL format, generate embeddings for each document chunk, create an index with vector fields, ingest the documents, and query using semantic search to retrieve relevant context for your LLM applications.

What deployment options does OpenSearch Skills support?

OpenSearch Skills covers both local and cloud deployments. Run OpenSearch locally using Docker for development and testing, or deploy to AWS using managed services or EC2 instances. Configure index lifecycle policies for data retention, set up ingestion pipelines via Fluent Bit or Logstash, and manage cluster health and observability across your chosen infrastructure.

How do I implement hybrid search combining BM25 and dense vectors?

OpenSearch Skills enables hybrid search by combining BM25 keyword matching with dense vector similarity. Create an index with both text fields and vector fields, ingest documents with both representations, then use a query that scores both BM25 and k-NN results and combines them (via normalization and weighting) to balance lexical and semantic relevance for improved search quality.

SKILL.md

rendered from the published skill — quoted content, verbatim