Amazon Opensearch Service

AWSOfficialPublished by AWS

Skill
Workflow
Cloud and DevOps
Apache-2.0
Low risk

From the author

Guides migration, provisioning, search, log-analytics, trace-analytics, and Agentic AI Assistant workflows for Amazon OpenSearch Service and Serverless across six capabilities — migration (Solr/ES/self-managed into AOS/AOSS, schema/query translation, sizing, cutover); provisioning (domain + AOSS lifecycle, upgrades, FGAC, monitoring); search (vector / semantic / hybrid / RAG with Bedrock); log-analytics (PPL, OSI, anomaly detection, Dashboards); trace-analytics (OTel spans, service maps, Data Prepper); ai-assistant (natural language data exploration, incident investigation, root cause analysis). Triggers on OpenSearch, AOS, AOSS, Elasticsearch, Solr, vector/k-NN/semantic/hybrid search, RAG, log analytics, PPL, trace analytics, ISM, FAISS, HNSW, Migration Assistant, UltraWarm, OR1, query my data, analyze logs, investigate errors, root cause analysis.

— the description by AWS

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

This skill provides a unified framework for interacting with Amazon OpenSearch Service (AOS) and OpenSearch Serverless (AOSS). It covers the full lifecycle of OpenSearch operations, including migration from legacy systems like Solr or Elasticsearch, infrastructure provisioning, advanced search implementation (including vector and RAG), and observability through log and trace analytics. It also features an Agentic AI Assistant for natural language data exploration and incident investigation, ensuring users can navigate complex cluster management and data analysis tasks without manual query crafting.

What it does

Guides migration workflows from Solr, Elasticsearch, or self-managed OpenSearch to AOS/AOSS.
Assists with provisioning, lifecycle management, and monitoring of domains and collections.
Facilitates implementation of vector, semantic, and hybrid search, including RAG with Bedrock.
Supports log and trace analytics using PPL, OpenTelemetry, and anomaly detection.
Provides an Agentic AI Assistant for natural language querying and automated incident root cause analysis.
Enforces operational best practices for sharding, storage tiering, and security configurations.

When to use it

When planning a migration from legacy search engines to Amazon OpenSearch Service.
When setting up or optimizing search capabilities, including vector search and RAG.
When troubleshooting cluster health, performance, or shard allocation issues.
When performing log or trace analysis to investigate system incidents.
When you need to generate optimized PPL or DSL queries using natural language.

Advantages

Centralizes complex operational knowledge across six distinct OpenSearch capabilities.
Reduces manual query complexity by leveraging the Agentic AI Assistant.
Provides clear, actionable guidance on production failure modes and migration blockers.
Standardizes interactions through consistent CLI and data-plane patterns.
Promotes architectural best practices for scaling and cost-effective storage tiering.

Usage examples

“How do I migrate my existing Solr index to Amazon OpenSearch Service?”: The agent identifies the migration capability, explains that Solr to OpenSearch is a refactor migration, and guides you through using the Migration Assistant for data backfill.
“Why is my cluster status yellow and how can I fix it?”: The agent uses the AI Assistant capability to analyze cluster health, check shard allocation, and provide specific commands or steps to resolve unassigned shards.
“Help me set up a RAG pipeline using Bedrock and my OpenSearch domain.”: The agent routes to the search capability, providing instructions on configuring Bedrock connectors and vector index mappings for semantic retrieval.
“Find the root cause of the 503 errors in my application logs from the last hour.”: The agent utilizes the AI Assistant to query logs, perform time-series analysis, and correlate findings to identify the likely root cause of the errors.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/aws/amazon-opensearch-service/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip amazon-opensearch-service.zip -d .agents/skills

Documentation by AWS

Amazon OpenSearch Service — the unified skill

This skill answers anything about Amazon OpenSearch Service or Serverless across six capabilities. Step 0 below routes the question to ONE capability and points at that capability's entry-point reference. Everything else — when to dispatch, sub-references, capability-specific facts, cross-capability links — lives in the entry-point reference for that capability.

AWS MCP server is recommended, not required. Capability references show standard AWS CLI commands as the primary syntax (e.g., aws opensearch describe-domain, aws opensearchserverless create-collection). Where the AWS MCP server is available, its call_aws tool offers a streamlined alternative — but every operation in this skill MUST work via the AWS CLI alone. Data-plane HTTP calls against AOS / AOSS use awscurl for SigV4-signed requests; this works in both contexts.

Step 0: detect the capability — first thing you do

Pick one of the six capabilities below. State the detected capability in your first sentence (e.g., "Detected capability: SEARCH — semantic search setup with Bedrock embeddings."). Then load the entry-point reference; that file describes when to dispatch, indexes the rest of the capability's files, and routes you to the next step.

Capability	Entry-point reference
migration — Solr / Elasticsearch / self-managed OpenSearch into AOS or AOSS. Schema/query translation, sizing, cutover.	`references/assessment-workflow.md`
provisioning — Provisioning and managing AOS domains and AOSS collections. Lifecycle, upgrades, storage tiers, FGAC, monitoring.	`references/provisioning-reference.md`
search — Vector / semantic / hybrid / sparse / dense / RAG retrieval. Bedrock connectors, FAISS HNSW vs Lucene.	`references/search-semantic-search-guide.md`
log-analytics — Log search, observability, PPL, OSI ingestion, anomaly detection, OpenSearch Dashboards. Splunk/Datadog/ELK alternatives.	`references/log-analytics-guide.md`
trace-analytics — Distributed traces with OpenTelemetry. Span queries, service maps, Data Prepper.	`references/trace-analytics-trace-queries.md`
ai-assistant — Agentic AI Assistant: auto-discovers indices, generates optimized PPL/DSL queries, summarizes results, and investigates incidents end-to-end. No manual query crafting needed.	`references/ai-assistant.md`

If a prompt spans capabilities (e.g., "migrate from Solr AND set up RAG on the new domain"), pick the dominant capability for the response and close with a one-line handoff to the other capability's entry-point ref.

Universal rules (apply to ALL capabilities)

These rules apply to every response, regardless of capability. Capability-specific rules (sizing math, shape detection, Migration Assistant for Amazon OpenSearch Service capability matrix, k-NN engine selection) live in the entry-point references, not here.

Report header (every multi-section response). Begin every multi-section response with a single fenced metadata block: > Generated: <ISO 8601 timestamp> | Skill: amazon-opensearch-service v<N>. Get the time by calling the current_time tool (returns ISO 8601 in UTC). Read the skill version from this file's frontmatter version: field. For one-line answers (terse FOCUSED_OPERATIONAL replies, anti-pattern refusals) the header is optional; for any multi-section deliverable it is REQUIRED. Place it immediately after the report title and before the first ## heading.
No dollar estimates (HARD CONSTRAINT). Never produce $X/month, ~$1,500, or any dollar figure. Route every cost question to https://calculator.aws and stop. If a sub-reference contains dollar figures, treat them as informational context only and do NOT pass them through to the user.
No credential leakage (HARD CONSTRAINT). Never include master usernames, KMS key ARNs, VPC endpoint URLs, instance IPs, or account IDs in generated output.
Pick one for every A-vs-B decision. Name a primary recommendation in one line with a one-sentence reason. A "go with B if..." caveat is allowed AFTER the primary; never lead with conditional-only guidance.
Source restatement. The first 2–3 sentences must restate the source (engine + version + scale) when known, or restate the customer's question in concrete terms. The very first text the user sees must NOT be tool narration, meta-commentary, the report title, or simply restating the question verbatim.
No marketing tone. Do NOT use "seamless", "robust", "best-in-class", "production-hardened", "enterprise-grade", "world-class", "cleanly", "elegant". Do NOT stack 3+ vague hedges ("typically", "generally", "usually", "in most cases") in a single recommendation — be specific about when it does and does not apply.
Cross-capability handoff. When a user prompt spans capabilities (e.g., "migrate from Solr AND set up RAG on the new domain"), pick the dominant capability for the response, then close with a one-line handoff: "For <other capability>, see references/<other-capability>-<entry>.md."

Cross-cutting references (used across multiple capabilities)

These references are not capability-prefixed because they apply across capabilities. Capability entry-point references load them when relevant; SKILL.md never loads them directly.

references/sizing.md — sizing math, instance family details, OR1 trade-offs, watermarks, JVM heap rules.
references/vector-knn.md — k-NN engines, memory math, RAG ingestion patterns, ELSER alternatives.
references/observability.md — log analytics patterns, ISM, UltraWarm/Cold tiering, Splunk/Datadog migration playbooks.
references/security.md — FGAC, encryption, VPC patterns, audit logs, compliance posture.
references/personas.md — communication style per persona.
references/assessment-gotchas.md — production gotcha catalog (cite by number in Migration specifics or Risks/blockers tables; each gotcha carries a Category: tag that determines its lane).
references/assessment-knowledge-retrieval.md — topic → tool → URL recipe for batched verification.

Assets (assets/): report templates for FULL_ASSESSMENT renderings (Solr-source, ES-source, executive summary).

What this skill does NOT do

Estimate dollar costs. Pricing changes monthly and account-specific (RI, Savings Plan, EDP) discount math is outside this skill's reliable scope. Use https://calculator.aws.
Move data. Use Migration Assistant for Amazon OpenSearch Service (Historical Data Migration for backfill, Live Traffic Migration for live cutover).
Build embedding models. Use Amazon Bedrock or SageMaker.
Replace Splunk SPL or Datadog APM 1:1. Some queries / detectors / dashboards need rewriting.
Tune relevance for a specific catalog. Use OpenSearch Benchmark big5 workload + your own judgment list.

Guardrail — where this skill's own files live (MCP vs local install)

This skill can be loaded two ways, and they resolve the skill's own bundled files from different places. Determine how the skill was loaded before reading a reference or running a script:

Loaded through the AWS MCP server's retrieve_skill tool: The skill is not installed on the local filesystem. You MUST fetch each reference or script via retrieve_skill with the file parameter (e.g. file="references/architecture.md" or file="scripts/deploy.py"), and run the script from the returned content. Do NOT file_read these paths locally — they do not exist on disk.
Installed locally (e.g. .kiro/skills/your-skill/ or ~/.claude/skills/your-skill/): Read and run files from the local skill directory using relative paths.

This distinction applies only to the skill's own packaged files. User data and session artifacts are always read from and written to the user's working directory. Never fetch or write customer data through retrieve_skill.

Files

ATTRIBUTION.mdAdded by AutoGineer930 B
LICENSELicence file added by AutoGineer9.9 KiB
NOTICELicence file added by AutoGineer67 B
SKILL.md9.2 KiB
assets/elasticsearch-gap-register.md5.1 KiB
assets/elasticsearch-index-template-skeleton.md4.2 KiB
assets/elasticsearch-report-template.md7.8 KiB
assets/executive-summary-template.md3.0 KiB
assets/report-template.md12.9 KiB
assets/solr-gap-register.md3.8 KiB
assets/solr-index-template-skeleton.md3.7 KiB
assets/solr-report-template.md5.9 KiB
assets/tech-deepdive-template.md6.2 KiB
references/ai-assistant.md18.0 KiB
references/assessment-gotchas.md17.1 KiB
references/assessment-knowledge-retrieval.md9.3 KiB
references/assessment-shape-anti-pattern-pushback.md10.4 KiB
references/assessment-shape-comparative-decision.md12.9 KiB
references/assessment-shape-focused-operational.md9.9 KiB
references/assessment-shape-full-assessment.md20.7 KiB
references/assessment-shape-overview.md12.1 KiB
references/assessment-shape-schema-conversion.md8.9 KiB
references/assessment-shape-sizing-only.md9.4 KiB
references/assessment-shape-translation.md12.3 KiB
references/assessment-workflow.md48.4 KiB
references/compatibility-rubric.md5.9 KiB
references/log-analytics-guide.md18.0 KiB
references/log-analytics-osi-pipelines.md8.6 KiB
references/log-analytics-troubleshooting.md1.6 KiB
references/observability-ppl-reference.md15.1 KiB
references/observability.md14.5 KiB
references/personas.md10.4 KiB
references/provisioning-agentic-setup.md4.8 KiB
references/provisioning-domain-deploy-search.md2.2 KiB
references/provisioning-domain-provision.md3.4 KiB
references/provisioning-monitoring.md3.4 KiB
references/provisioning-reference.md6.8 KiB
references/provisioning-serverless-agentic-setup.md14.9 KiB
references/provisioning-serverless-deploy-search.md6.7 KiB
references/provisioning-serverless-deprovision.md4.0 KiB
references/provisioning-serverless-provision.md8.8 KiB
references/provisioning-storage-tiers.md2.1 KiB
references/provisioning-troubleshooting.md8.0 KiB
references/provisioning-upgrades.md5.4 KiB
references/readiness-rubric.md3.4 KiB
references/search-bedrock-connectors.md3.8 KiB
references/search-dense-vector-models.md5.5 KiB
references/search-document-processing-guide.md7.7 KiB
references/search-evaluation-guide.md5.3 KiB
references/search-index-config.md2.4 KiB
references/search-recipes.md12.9 KiB
references/search-semantic-search-guide.md18.2 KiB
references/search-sparse-vector-models.md4.9 KiB
references/search-troubleshooting.md1.9 KiB
references/security.md9.3 KiB
references/sizing.md12.2 KiB
references/source-elasticsearch.md6.7 KiB
references/source-opensearch.md4.8 KiB
references/trace-analytics-trace-ingestion.md3.3 KiB
references/trace-analytics-trace-queries.md13.9 KiB
references/trace-analytics-troubleshooting.md1.6 KiB
references/vector-knn.md12.9 KiB

references/search-document-processing-guide.md

Document Processing with Docling

This guide covers how to process PDF, DOCX, PPTX, XLSX, HTML, and other document formats for ingestion into OpenSearch using Docling.

The AWS MCP server is recommended for executing these commands but is not required — all steps use standard AWS CLI syntax.

Overview

Docling is an open-source Python library (MIT license) by IBM Research that converts unstructured documents into structured data. It detects page layout, reading order, table structure, code blocks, formulas, and images using AI models, and runs locally on commodity hardware.

Supported Input Formats

PDF, DOCX, PPTX, XLSX, HTML, Markdown, AsciiDoc, CSV, images (PNG, JPEG, TIFF, BMP, WEBP), audio (MP3, WAV).

Choosing a processing approach

Match the Docling pipeline to the document's dominant content, because a single default pipeline degrades on tables, scans, and figures.

Document character	Approach	What to enable	Why
Digital-text, prose-heavy (papers, reports, contracts)	Semantic	Hierarchy-aware `HybridChunker` respecting headings	Preserves section boundaries so chunks stay self-contained
Table-heavy (financial reports, spec sheets)	Tables	TableFormer table extraction; serialize tables to markdown	You MUST enable table extraction explicitly, because auto-detection does NOT reliably recognize table-heavy documents and untreated tables collapse into unusable text
Scanned / image-only pages	Scanned	OCR (`do_ocr=True`; RapidOCR / PP-OCRv4) before chunking	Digital-text pipelines produce empty chunks on scans because there is no extractable text layer
Figure / diagram-rich	Multimodal	VLM picture descriptions (e.g. SmolVLM-256M)	Figures carry meaning that is lost unless described, because embeddings index text only

You SHOULD default to the semantic approach for digital-text documents and disable OCR (do_ocr=False, since Docling enables it by default), because OCR adds significant latency and is unnecessary when a text layer exists.

Chunking for Search Ingestion

Docling provides two chunking strategies for breaking documents into search-ready pieces:

HierarchicalChunker (structure-based)

Splits at every section/heading boundary. Produces many small chunks that respect document structure.

HybridChunker (recommended for OpenSearch)

Combines structure-aware splitting with token limits. Preserves document hierarchy while ensuring chunks fit within embedding model constraints.

Parameters: max_tokens=512 (HybridChunker takes no overlap parameter — it merges adjacent peer chunks rather than applying a fixed overlap).

Processing Pipeline for Document Search

The recommended end-to-end flow:

Convert — Use Docling to parse the document into structured form.
Chunk — Use HybridChunker with token limits matching your embedding model.
Export — Write chunks as JSONL with text + metadata fields.
Index — Load into OpenSearch using the ingest pipeline.
Search — Query using your configured search pipeline.

JSONL chunk format

Each line of the exported .jsonl MUST be a standalone JSON object with at minimum a text field, because the downstream bulk-index and OSIS ingestion paths reject lines that lack indexable content.

{"text": "...", "headings": ["Section Title"], "source_file": "doc.pdf", "chunk_id": 0, "page_number": 1}

Field	Type	Description
`text`	string	Chunk content (required)
`headings`	array	Section headings this chunk belongs to
`source_file`	string	Original source filename
`chunk_id`	int	Sequential chunk index within the file
`page_number`	int	Source page number

You SHOULD validate that every line parses as JSON and carries a non-empty text field before uploading to S3, because a single malformed line can fail an entire bulk request.

Choosing Chunk Size

For BM25 (keyword search): larger chunks (1000+ tokens) work well since BM25 benefits from more context.
For dense vector / semantic search: 256–512 tokens is typical, matching embedding model input limits.
For hybrid search: 512 tokens is a good default. The HybridChunker above has no overlap parameter (it merges adjacent peer chunks); a fixed ~50-token overlap applies only if you use a custom, non-Docling chunker.

Adjusting chunking when retrieval quality is poor

Re-processing is expensive, so you SHOULD change chunking only when a quality check provides evidence, because blind re-ingestion wastes compute without a targeted fix.

Symptom	Likely cause	Fix
Results lack surrounding context	Chunks too small	Increase `max_tokens`
Results contain irrelevant noise	Chunks too large	Decrease `max_tokens`
Tables absent from results	Table extraction not run	Re-process with TableFormer enabled
Scanned pages return empty or garbled text	OCR not run	Re-process with OCR enabled
Figures / diagrams missing from results	Visual content not described	Re-process with VLM picture descriptions

You MUST NOT automatically re-ingest on a poor result, because re-ingestion is costly and may not address the true cause; confirm the symptom and the intended fix first.

Evaluating chunk quality

Before indexing at scale, you SHOULD sample chunks and judge them against the dimensions that matter for the document type, because indexing low-fidelity chunks propagates the defect into every downstream search result.

Document type	Dimensions to judge
Prose / semantic	text fidelity, reading order, chunk boundaries, heading structure
Tables	table structure, table-content fidelity, text fidelity, completeness
Scanned	OCR text fidelity, reading order, completeness
Multimodal	image descriptions, text fidelity, reading order, completeness

Rate each dimension good | fair | poor. When any dimension is poor, explain the cause and recommend a specific re-process; you MUST NOT re-ingest without user confirmation, because re-processing is expensive.

Performance Tips

Skip page images if not needed to save memory.
Use max_num_pages or page_range to limit processing for large documents.
Enable parallel processing for multi-core systems.
Docling enables OCR by default (do_ocr=True); disable it (do_ocr=False) for digital-text documents in the semantic profile, and keep it on only for scanned / image-only pages, since OCR adds significant latency.

Security Considerations

Exported chunks (JSONL) and the S3 objects that stage them carry the full text of the source documents, so treat them with the same sensitivity as the originals.

Encryption at rest. Enable server-side encryption on the target S3 bucket — SSE-S3 at minimum, and SSE-KMS with a customer-managed KMS key for compliance workloads (PCI-DSS, HIPAA) so you retain key control and an audit trail. Apply the same encryption posture (AWS-owned key by default, customer-managed KMS key for compliance) to the destination OpenSearch collection or domain that ingests the chunks.
Access control. Restrict the S3 bucket with a bucket policy (and, where applicable, aws:SourceVpce / IAM principal conditions) so only the ingestion role and authorized operators can read the staged JSONL. Block public access at the account and bucket level.
Sensitive / PII content. Source documents may contain PII or other regulated data, which propagates verbatim into chunks, the index, and any search result or agent response. Classify sources before ingesting, redact or exclude fields you do not intend to make searchable, and delete the intermediate JSONL from S3 once indexing is confirmed rather than leaving it as a long-lived copy of the corpus.

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Needs an MCP server

Security

What our scanner noticed· Low risk

Instructions that use the network

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Published by

AWS

Publisher

AWS (Official)

Source

aws/agent-toolkit-for-aws · plugins/aws-data-analytics/skills/amazon-opensearch-service · commit 0a218f8

Licence

Apache License 2.0 (redistribution allowed)

LICENSE (added by AutoGineer from LICENSE)
NOTICE (added by AutoGineer from NOTICE)
ATTRIBUTION.md (written by AutoGineer)

Synced

2026-11-08

Also distributed in

aws/agent-toolkit-for-aws

AutoGineer is not affiliated with or endorsed by AWS.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download amazon-opensearch-service.zip