# Elasticsearch Labs

Articles and tutorials from the Search team at Elastic

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Prescale the search tier ahead of traffic with Minimum Search VCUs for Elasticsearch Serverless

DevFeed: [Prescale the search tier ahead of traffic with Minimum Search VCUs for Elasticsearch Serverless](<https://devfeed.tech/articles/prescale-the-search-tier-ahead-of-traffic-with-minimum-search-vcus-for-elasticsearch-serverless-140645.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/minimum-search-vcus-elasticsearch-serverless>)

Author: Yuvraj Gupta

Published: 2026-11-17T16:32:27Z

Content type: tutorial

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [Elastic Cloud](<https://devfeed.tech/topics/elastic-cloud.md>), [Serverless](<https://devfeed.tech/topics/serverless.md>)

Tags: [api](<https://devfeed.tech/tags/api.md>), [autoscaling](<https://devfeed.tech/tags/autoscaling.md>), [billing](<https://devfeed.tech/tags/billing.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [elastic-cloud-serverless](<https://devfeed.tech/tags/elastic-cloud-serverless.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [operations](<https://devfeed.tech/tags/operations.md>), [search](<https://devfeed.tech/tags/search.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [traffic](<https://devfeed.tech/tags/traffic.md>)

### AI overview

Minimum Search VCUs lets you prescale an Elasticsearch Serverless search tier for predictable traffic while autoscaling remains active above the configured floor. Use the last 10 days of Search VCU usage to choose a whole-number minimum between observed baseline and peak, and set it about 30-60 minutes before demand arrives. The article cites sales and predictable workday or shift traffic as examples, and says to lower the minimum when the busy period ends. You are billed for at least the configured VCUs while enabled, even when idle or below that level; autoscaled capacity above the minimum is billed while provisioned. Configure it in the Elastic Cloud console by reviewing the VCUs chart, enabling the setting, entering a minimum, and confirming the change and billing implications. It can also be set or disabled through the API.

### Source excerpt

Set Minimum Search VCUs on Elasticsearch Serverless so the search tier stays provisioned at the level you choose. Autoscaling can still raise it as search load grows.

## Too much of a good thing: Why 20 candidates beat 200 when reranking search results

DevFeed: [Too much of a good thing: Why 20 candidates beat 200 when reranking search results](<https://devfeed.tech/articles/too-much-of-a-good-thing-why-20-candidates-beat-200-when-reranking-search-results-132453.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/jina-reranker-candidate-pool-size>)

Author: Scott Martens,Sofia Vasileva

Published: 2026-11-13T16:15:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [search-engines](<https://devfeed.tech/topics/search-engines.md>), [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [AI-driven information retrieval](<https://devfeed.tech/topics/ai-driven-information-retrieval.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [about](<https://devfeed.tech/tags/about.md>), [all](<https://devfeed.tech/tags/all.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [bm25](<https://devfeed.tech/tags/bm25.md>), [but](<https://devfeed.tech/tags/but.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [good](<https://devfeed.tech/tags/good.md>), [help](<https://devfeed.tech/tags/help.md>), [jina-ai](<https://devfeed.tech/tags/jina-ai.md>), [just](<https://devfeed.tech/tags/just.md>), [ml-research](<https://devfeed.tech/tags/ml-research.md>), [more](<https://devfeed.tech/tags/more.md>), [relevance](<https://devfeed.tech/tags/relevance.md>), [reranker](<https://devfeed.tech/tags/reranker.md>), [reranking](<https://devfeed.tech/tags/reranking.md>), [results](<https://devfeed.tech/tags/results.md>), [search](<https://devfeed.tech/tags/search.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [semantic-search](<https://devfeed.tech/tags/semantic-search.md>), [thing](<https://devfeed.tech/tags/thing.md>), [v3-5](<https://devfeed.tech/tags/v3-5.md>)

### AI overview

Across 25 retrieval benchmarks, jina-reranker-v3.5 performed best with different candidate pool sizes depending on first-stage retrieval: about 100 candidates for Elasticsearch BM25 and 20 for jina-embeddings-v5-text-small semantic search. With BM25, 50 candidates retained nearly all the accuracy of 100 while halving token use; with semantic search, 200 candidates scored worse on average than using no reranker. The article attributes this to listwise rerankers processing candidates together, making them vulnerable to noise and attention dilution. Better initial retrieval leaves less for reranking to improve. Results varied by benchmark, so the authors recommend measuring on the target data. For Elasticsearch, they outline adding a text_similarity_reranker and tuning rank_window_size: start at 100 for BM25 or 20 for semantic search, then compare performance.

### Source excerpt

We expected more candidates to help, but across 25 benchmarks in Elasticsearch jina-reranker-v3.5 peaked at 100 candidates for BM25 and just 20 for semantic search, where 200 scored worse than no reranker at all.

## No pipeline, no second copy: Querying Amazon S3 from Elasticsearch with ES|QL

DevFeed: [No pipeline, no second copy: Querying Amazon S3 from Elasticsearch with ES|QL](<https://devfeed.tech/articles/no-pipeline-no-second-copy-querying-amazon-s3-from-elasticsearch-with-es-ql-130587.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/query-amazon-s3-esql-data-federation>)

Author: Udayasimha Theepireddy,Srinivas Pendyala

Published: 2026-11-11T17:00:00Z

Content type: tutorial

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [Amazon S3](<https://devfeed.tech/topics/amazon-s3.md>), [kibana](<https://devfeed.tech/topics/kibana.md>), [Parquet files in Amazon S3](<https://devfeed.tech/topics/parquet-files-in-amazon-s3.md>), [etl](<https://devfeed.tech/topics/etl.md>), [Elastic Cloud](<https://devfeed.tech/topics/elastic-cloud.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [VPC Flow Logs](<https://devfeed.tech/topics/vpc-flow-logs.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [amazon](<https://devfeed.tech/tags/amazon.md>), [amazon-s3](<https://devfeed.tech/tags/amazon-s3.md>), [aws](<https://devfeed.tech/tags/aws.md>), [copy](<https://devfeed.tech/tags/copy.md>), [credentials](<https://devfeed.tech/tags/credentials.md>), [data](<https://devfeed.tech/tags/data.md>), [data-federation](<https://devfeed.tech/tags/data-federation.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [es-ql](<https://devfeed.tech/tags/es-ql.md>), [federated](<https://devfeed.tech/tags/federated.md>), [identity](<https://devfeed.tech/tags/identity.md>), [index](<https://devfeed.tech/tags/index.md>), [integrations](<https://devfeed.tech/tags/integrations.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [query](<https://devfeed.tech/tags/query.md>), [querying](<https://devfeed.tech/tags/querying.md>), [s3](<https://devfeed.tech/tags/s3.md>), [source](<https://devfeed.tech/tags/source.md>), [syntax](<https://devfeed.tech/tags/syntax.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

ES|QL Data Federation lets Elasticsearch query Parquet, CSV, TSV, and NDJSON files in Amazon S3 using the usual `FROM` syntax, without copying data into the cluster or defining mappings. It infers schemas from files and can push filters and column selection into the Parquet reader; datasets share a namespace with indices, allowing queries across hot indices and S3 data. Setup requires Elastic Cloud Hosted 9.5 or later with an Enterprise subscription; 9.5 requires enabling the feature and specifying the S3 region, while 9.6 enables it by default and detects the region. Federated identity uses temporary AWS credentials and is available only on Hosted; self-managed, ECE, and ECK deployments use static credentials. AWS access can be scoped to `s3:GetObject` and `s3:ListBucket`. Federation incurs S3 read and possible transfer costs, while indexing may be appropriate for workloads requiring kNN, joins, or time-series functions. The feature is described as experimental in 9.5 and 9.6, so the source recommends treating it as exploratory.

### Source excerpt

With ES|QL Data Federation, an S3 bucket is a data source you query with the same FROM syntax you use for an index, and federated identity keeps AWS credentials out of Elasticsearch.

## Fine-grained authorization for Elasticsearch RAG with SpiceDB

DevFeed: [Fine-grained authorization for Elasticsearch RAG with SpiceDB](<https://devfeed.tech/articles/fine-grained-authorization-for-elasticsearch-rag-with-spicedb-129234.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/fine-grained-authorization-elasticsearch-rag-spicedb>)

Author: Sohan Maheshwar,Carly Richmond

Published: 2026-11-11T16:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [Authorization](<https://devfeed.tech/topics/authorization.md>), [Retrieval Augmented Generation (RAG)](<https://devfeed.tech/topics/retrieval-augmented-generation-rag.md>), [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [access-management](<https://devfeed.tech/tags/access-management.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-rag](<https://devfeed.tech/tags/agentic-rag.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-applications](<https://devfeed.tech/tags/ai-applications.md>), [ai-chatbot](<https://devfeed.tech/tags/ai-chatbot.md>), [api](<https://devfeed.tech/tags/api.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [chatbot](<https://devfeed.tech/tags/chatbot.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [graph](<https://devfeed.tech/tags/graph.md>), [integrations](<https://devfeed.tech/tags/integrations.md>), [llm](<https://devfeed.tech/tags/llm.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [rag](<https://devfeed.tech/tags/rag.md>), [rag-pipeline](<https://devfeed.tech/tags/rag-pipeline.md>), [vector-database](<https://devfeed.tech/tags/vector-database.md>)

### AI overview

The excerpt describes a RAG authorization design in which Elasticsearch retrieves relevant chunks and SpiceDB checks each chunk's document ID and user before content reaches the LLM. A deterministic LangGraph authorization node filters unauthorized documents; missing IDs, denied checks, or errors fail closed, and the agent cannot bypass the check. SpiceDB models access as relationships, supporting department membership, document sharing, and individual exceptions without changing roles or reindexing. Compared with Elasticsearch document-level security, this keeps relationship rules in a shared graph rather than denormalizing them into indexed documents. The excerpt notes that post-filtering can reduce available context, recommends overfetching and optional metadata prefilters, and describes LookupResources as an alternative for some access patterns, with result-size limits. Production setup requires keeping relationships synchronized and using a persistent SpiceDB datastore.

### Source excerpt

Department filters break the first time Engineering shares one document with Sales, so SpiceDB checks each retrieved chunk against a relationship graph before your Elasticsearch RAG pipeline passes it to the LLM.

## AI connectors in Kibana: 90+ data sources your agent can call

DevFeed: [AI connectors in Kibana: 90+ data sources your agent can call](<https://devfeed.tech/articles/ai-connectors-in-kibana-90-data-sources-your-agent-can-call-129233.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/ai-connectors-kibana-agent-context>)

Author: Erik Currin

Published: 2026-11-10T15:30:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [kibana](<https://devfeed.tech/topics/kibana.md>), [Tool](<https://devfeed.tech/topics/tool.md>), [AI Context Management](<https://devfeed.tech/topics/ai-context-management.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [Documentation](<https://devfeed.tech/topics/documentation.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agent-builder](<https://devfeed.tech/tags/agent-builder.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-incident-response](<https://devfeed.tech/tags/ai-incident-response.md>), [alert](<https://devfeed.tech/tags/alert.md>), [all](<https://devfeed.tech/tags/all.md>), [analyst](<https://devfeed.tech/tags/analyst.md>), [api](<https://devfeed.tech/tags/api.md>), [api-keys](<https://devfeed.tech/tags/api-keys.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [bearer](<https://devfeed.tech/tags/bearer.md>), [confluence](<https://devfeed.tech/tags/confluence.md>), [connectors](<https://devfeed.tech/tags/connectors.md>), [consider](<https://devfeed.tech/tags/consider.md>), [github](<https://devfeed.tech/tags/github.md>), [integrations](<https://devfeed.tech/tags/integrations.md>), [kibana](<https://devfeed.tech/tags/kibana.md>), [pagerduty](<https://devfeed.tech/tags/pagerduty.md>), [slack](<https://devfeed.tech/tags/slack.md>)

### AI overview

Kibana's library of more than 90 AI connectors lets agents call services such as Slack, GitHub, Confluence, PagerDuty, Gmail, Drive, Jira, Salesforce, and Snowflake while reasoning over Elasticsearch data. Connector specs define reachable data, authentication, and agent-callable actions, so the text says no integration code or separate pipeline is needed. Kibana Connectors handle OAuth, API keys, and bearer tokens; Elastic's Auth Relay Service can let users authorize Google or Slack without sharing credentials with Kibana. For personal data such as inboxes or DMs, the source recommends each user authorize with personal OAuth; organizational data such as shared repos or incident timelines typically uses an admin-configured system account or API key. Setup is through an agent's tools configuration in Kibana. Examples describe combining indexed telemetry or metrics with external records to investigate incidents or explain engagement changes.

### Source excerpt

Your index has the alert, but the Slack thread, GitHub PR, Confluence doc and PagerDuty timeline that explain it sit elsewhere. AI connectors are how your agent reads all four while it reasons.

## How our agent skills research, build, and test a complete Elastic integration

DevFeed: [How our agent skills research, build, and test a complete Elastic integration](<https://devfeed.tech/articles/how-our-agent-skills-research-build-and-test-a-complete-elastic-integration-114234.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/agent-skills-elastic-integration>)

Author: Chris Berkhout

Published: 2026-11-02T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [Agent Skills](<https://devfeed.tech/topics/agent-skills.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [Developer skills](<https://devfeed.tech/topics/developer-skills.md>), [Integration testing](<https://devfeed.tech/topics/integration-testing.md>), [Test automation](<https://devfeed.tech/topics/test-automation.md>), [Programming Languages / Compilers](<https://devfeed.tech/topics/programming-languages-compilers.md>), [NVD](<https://devfeed.tech/topics/nvd.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>)

Tags: [agent-skills](<https://devfeed.tech/tags/agent-skills.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [api](<https://devfeed.tech/tags/api.md>), [cel](<https://devfeed.tech/tags/cel.md>), [developer-experience](<https://devfeed.tech/tags/developer-experience.md>), [elastic](<https://devfeed.tech/tags/elastic.md>), [integration](<https://devfeed.tech/tags/integration.md>), [integrations](<https://devfeed.tech/tags/integrations.md>), [research](<https://devfeed.tech/tags/research.md>), [test](<https://devfeed.tech/tags/test.md>)

### AI overview

Elastic's Integration Skills built a NIST NVD integration with two data streams: current CVE snapshots and individual change events. Research found that 78% of records in a day's collection were republications of older CVEs, often with security-relevant changes, leading to the two-stream design. Without an API key, backfill requests must be paced to stay within the shared rate limit. The integration selects one CVSS assessment for ECS fields using a stated version and source priority, while preserving all metric arrays in integration-specific fields. Pipeline and system tests passed, but the text cautions that passing tests cannot establish correct API interpretation, sensible defaults, useful real data, or adequate documentation. Manual inspection and review of the research brief, README, CEL programs, and test coverage remain part of deciding whether to trust the result. Suggested refinements include a latest transform, script tests for error paths, a real logo, and a dashboard.

### Source excerpt

We pointed our agent skills at the NIST NVD API and got two data streams and a validated CEL program, with pipeline and system tests passing. The research showed 78% of a day's CVE records are republications of older ones, which changed the design.

## We cut 40% of our MCP tools to get into Claude's connector directory

DevFeed: [We cut 40% of our MCP tools to get into Claude's connector directory](<https://devfeed.tech/articles/we-cut-40-of-our-mcp-tools-to-get-into-claude-s-connector-directory-108974.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/mcp-connector-claude-directory>)

Author: Sean Story

Published: 2026-10-30T00:00:00Z

Content type: news

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [MSP MCP](<https://devfeed.tech/topics/msp-mcp.md>), [AI connector platforms](<https://devfeed.tech/topics/ai-connector-platforms.md>), [Claude](<https://devfeed.tech/topics/claude.md>), [anthropic](<https://devfeed.tech/topics/anthropic.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [22](<https://devfeed.tech/tags/22.md>), [3](<https://devfeed.tech/tags/3.md>), [about](<https://devfeed.tech/tags/about.md>), [account](<https://devfeed.tech/tags/account.md>), [actions](<https://devfeed.tech/tags/actions.md>), [added](<https://devfeed.tech/tags/added.md>), [admin](<https://devfeed.tech/tags/admin.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [anthropic](<https://devfeed.tech/tags/anthropic.md>), [changes](<https://devfeed.tech/tags/changes.md>), [claude](<https://devfeed.tech/tags/claude.md>), [developer-experience](<https://devfeed.tech/tags/developer-experience.md>), [directory](<https://devfeed.tech/tags/directory.md>), [elastic](<https://devfeed.tech/tags/elastic.md>), [fewer](<https://devfeed.tech/tags/fewer.md>), [integrations](<https://devfeed.tech/tags/integrations.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [mcp-tools](<https://devfeed.tech/tags/mcp-tools.md>), [publishing](<https://devfeed.tech/tags/publishing.md>), [roadmap](<https://devfeed.tech/tags/roadmap.md>), [tool](<https://devfeed.tech/tags/tool.md>), [tools](<https://devfeed.tech/tags/tools.md>)

### AI overview

Anthropic's review of Elastic's MCP connector reduced the Elasticsearch project catalog from 36 tools to 22 and prompted changes across more than 100 tool definitions. The checklist required tool annotations and clear separation of read and write actions: Elastic split some combined tools, such as `cases.attachments`, and excluded `execute_connector_sub_action`. Reviewers also examined descriptions, response metadata, token efficiency, error messages, and tools unavailable under some licenses; the process caught bugs and led to a Verified badge. Publishers with customer-specific server URLs and OAuth settings should document configuration needs in advance. Claude Enterprise currently allows only one directory connector instance per account, so additional Elastic environments require custom connectors. Elastic says the Agent Builder server will remain available and plans a universal `mcp.elastic.co` endpoint, but gives no release date.

### Source excerpt

Anthropic's reviewers read every tool in our MCP connector and sent back fixes, wording changes and a case for publishing fewer of them, which is part of why mcp.elastic.co is now on the roadmap.

## Elastic Cloud Serverless audit logs: Ship the trail to Observability or Security

DevFeed: [Elastic Cloud Serverless audit logs: Ship the trail to Observability or Security](<https://devfeed.tech/articles/elastic-cloud-serverless-audit-logs-ship-the-trail-to-observability-or-security-106464.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/elastic-cloud-serverless-audit-logs>)

Author: Valentin Crettaz

Published: 2026-10-29T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [Elastic Cloud](<https://devfeed.tech/topics/elastic-cloud.md>), [Logging](<https://devfeed.tech/topics/logging.md>), [Threat monitoring](<https://devfeed.tech/topics/threat-monitoring.md>), [observability pipelines](<https://devfeed.tech/topics/observability-pipelines.md>), [Security](<https://devfeed.tech/topics/security.md>), [observability](<https://devfeed.tech/topics/observability.md>), [telemetry](<https://devfeed.tech/topics/telemetry.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [alerting](<https://devfeed.tech/tags/alerting.md>), [attacks](<https://devfeed.tech/tags/attacks.md>), [audit](<https://devfeed.tech/tags/audit.md>), [audit-logs](<https://devfeed.tech/tags/audit-logs.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [dashboards](<https://devfeed.tech/tags/dashboards.md>), [detection](<https://devfeed.tech/tags/detection.md>), [elastic](<https://devfeed.tech/tags/elastic.md>), [elastic-cloud](<https://devfeed.tech/tags/elastic-cloud.md>), [elastic-cloud-serverless](<https://devfeed.tech/tags/elastic-cloud-serverless.md>), [forensics](<https://devfeed.tech/tags/forensics.md>), [kibana](<https://devfeed.tech/tags/kibana.md>), [logs](<https://devfeed.tech/tags/logs.md>), [object](<https://devfeed.tech/tags/object.md>), [observability](<https://devfeed.tech/tags/observability.md>), [operations](<https://devfeed.tech/tags/operations.md>), [rules](<https://devfeed.tech/tags/rules.md>), [security](<https://devfeed.tech/tags/security.md>), [serverless](<https://devfeed.tech/tags/serverless.md>), [space](<https://devfeed.tech/tags/space.md>)

### AI overview

Elastic Cloud Serverless log delivery is generally available for shipping a source project's Elasticsearch and Kibana activity to an Observability or Security project in the same organization. Multiple sources can share a destination; ingest and retention are billed there, and retention is configurable. Five Ignore filters can drop matching events before delivery to reduce volume and cost. The stream covers saved-object changes, index access, sign-ins and access failures, with project and producer fields for attribution. Eight included alert rules watch for credential attack patterns; the first-seen identity rule needs at least a day of history. Seven dashboards include Kibana Object Forensics, Elasticsearch Data Access and Identity & Access. Enabling delivery requires Admin or Editor access on the source project. Organization administration logs are described as upcoming, not part of this release.

### Source excerpt

Eight alerting rules and seven dashboards arrive with the logs, including Kibana Object Forensics, where a deleted detection rule shows up with the person who deleted it and the space it was in.

## 56 scanned pages to a cited answer in 5 minutes: RAG pipeline for the documents search cannot read

DevFeed: [56 scanned pages to a cited answer in 5 minutes: RAG pipeline for the documents search cannot read](<https://devfeed.tech/articles/56-scanned-pages-to-a-cited-answer-in-5-minutes-rag-pipeline-for-the-documents-search-cannot-read-95936.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/rag-pipeline-scanned-pdf-jina-ocr>)

Author: Ben-Amin York Jr.

Published: 2026-10-22T00:00:00Z

Content type: tutorial

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [Retrieval Augmented Generation (RAG)](<https://devfeed.tech/topics/retrieval-augmented-generation-rag.md>), [hybrid-search](<https://devfeed.tech/topics/hybrid-search.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [document ai](<https://devfeed.tech/topics/document-ai.md>), [search-engines](<https://devfeed.tech/topics/search-engines.md>), [pdf](<https://devfeed.tech/topics/pdf.md>), [Elastic Cloud](<https://devfeed.tech/topics/elastic-cloud.md>)

Tags: [api](<https://devfeed.tech/tags/api.md>), [api-key](<https://devfeed.tech/tags/api-key.md>), [back-end](<https://devfeed.tech/tags/back-end.md>), [base64](<https://devfeed.tech/tags/base64.md>), [catalog](<https://devfeed.tech/tags/catalog.md>), [chat](<https://devfeed.tech/tags/chat.md>), [chatbot](<https://devfeed.tech/tags/chatbot.md>), [claude](<https://devfeed.tech/tags/claude.md>), [claude-haiku-4-5](<https://devfeed.tech/tags/claude-haiku-4-5.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [database](<https://devfeed.tech/tags/database.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [hybrid-search](<https://devfeed.tech/tags/hybrid-search.md>), [jina-ai](<https://devfeed.tech/tags/jina-ai.md>), [markdown](<https://devfeed.tech/tags/markdown.md>), [ocr](<https://devfeed.tech/tags/ocr.md>), [pdf](<https://devfeed.tech/tags/pdf.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [rag](<https://devfeed.tech/tags/rag.md>), [search](<https://devfeed.tech/tags/search.md>), [structured](<https://devfeed.tech/tags/structured.md>), [table](<https://devfeed.tech/tags/table.md>), [vector-database](<https://devfeed.tech/tags/vector-database.md>)

### AI overview

The described pipeline turns scanned PDFs into searchable, page-cited answers by sending each page image to jina-ocr-v1, which returns structured Markdown with tables, reading order, and inline handwriting. An Express server runs four pages in parallel through Elastic Inference Service and records page markers; Elasticsearch's semantic_text field chunks and embeds the Markdown at ingest. Hybrid retrieval combines lexical matching for exact terms with semantic matching for meaning, then Claude Haiku 4.5 generates answers grounded in retrieved sources and recommends only catalog products. The example uses a 56-page plant guide and a 48-plant catalog. Unlike the described legacy OCR chain of separate layout, recognition, table, and reading-order models, jina-ocr-v1 handles these in one pass. Each page must still be looped through individually; the article says legibility holds up to about 2048 x 2048 pixels. EIS is metered per token and billed as a Serverless add-on.

### Source excerpt

Ask a scanned, table-heavy PDF a question in plain language and get an answer stamped with the page it came from, because jina-ocr-v1 turns every page into structured Markdown that Elasticsearch embeds at ingest.

## Bringing Elastic into Claude: The 'why,' the 'how,' and the 'gotchas'

DevFeed: [Bringing Elastic into Claude: The 'why,' the 'how,' and the 'gotchas'](<https://devfeed.tech/articles/bringing-elastic-into-claude-the-why-the-how-and-the-gotchas-92967.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/publish-claude-plugin-directory>)

Author: Sean Story

Published: 2026-10-21T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [Claude](<https://devfeed.tech/topics/claude.md>), [project release](<https://devfeed.tech/topics/project-release.md>), [anthropic](<https://devfeed.tech/topics/anthropic.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [Documentation](<https://devfeed.tech/topics/documentation.md>), [Pull Request](<https://devfeed.tech/topics/pull-request.md>), [dogfooding](<https://devfeed.tech/topics/dogfooding.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai-tools](<https://devfeed.tech/tags/ai-tools.md>), [claude](<https://devfeed.tech/tags/claude.md>), [claude-plugin](<https://devfeed.tech/tags/claude-plugin.md>), [developer-experience](<https://devfeed.tech/tags/developer-experience.md>), [elastic](<https://devfeed.tech/tags/elastic.md>), [gotchas](<https://devfeed.tech/tags/gotchas.md>), [live](<https://devfeed.tech/tags/live.md>), [merge](<https://devfeed.tech/tags/merge.md>), [plugin](<https://devfeed.tech/tags/plugin.md>), [release](<https://devfeed.tech/tags/release.md>)

### AI overview

Elastic's Claude plugin took three weeks from first commit to publication, with the main hurdles being review and release behavior rather than implementation. The team created a separate `elastic/claude-plugin` repository because Claude plugins can include Claude-specific features beyond portable Agent Skills. They advise choosing a plugin path that excludes internal development files, since automated scans of the repository root can trigger verification warnings. Pointing the plugin at `main` caused a new `v0.0.1` version for every merge, so they switched to a `stable` branch and promote batches from `main` when ready to release. Review took about a week; automated warnings about credentials, a missing icon, and download-and-run commands were ultimately waived. Publication happened immediately after review, with no notification or staging step, and initially placed the plugin only in the unverified Community directory; access to other marketplaces requires separate verification. The authors recommend testing before submission and waiting for reviewer direction before reacting to warnings.

### Source excerpt

Point your Claude plugin at main and every merge ships as a release, and that's the least surprising thing that happened between submitting and going live.

## Your AI chat is not a screenshot: Live shared conversations in Elastic Agent Builder

DevFeed: [Your AI chat is not a screenshot: Live shared conversations in Elastic Agent Builder](<https://devfeed.tech/articles/your-ai-chat-is-not-a-screenshot-live-shared-conversations-in-elastic-agent-builder-85467.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/agent-builder-conversation-sharing>)

Author: Graham Hudgins,Bruno Farache

Published: 2026-10-12T00:00:00Z

Content type: release

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [OpenClaw](<https://devfeed.tech/topics/openclaw.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Single sign-on (SSO)](<https://devfeed.tech/topics/sso.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [agent-builder](<https://devfeed.tech/tags/agent-builder.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai](<https://devfeed.tech/tags/ai.md>), [incident-response](<https://devfeed.tech/tags/incident-response.md>), [kibana](<https://devfeed.tech/tags/kibana.md>)

### AI overview

Elastic Agent Builder now supports live shared conversations. Owners can share a thread with named teammates or make it available to users who can access its agent; collaborators can read its history and continue interacting in the same conversation. Sharing is limited by existing Kibana privileges and applies only to that conversation. The feature is generally available on Elastic Cloud Serverless, with Elastic Cloud Hosted and Elastic Self-Managed support planned for version 9.6.

### Source excerpt

Whoever picks up the problem next reads the whole thread and carries on with the same agent, with no access to anything beyond that one conversation.

## How we're building a context engine to cut agent token use by 71%

DevFeed: [How we're building a context engine to cut agent token use by 71%](<https://devfeed.tech/articles/how-we-re-building-a-context-engine-to-cut-agent-token-use-by-71-78744.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/context-engineering-for-agents>)

Author: Anish Mathur

Published: 2026-10-08T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [AI Context Management](<https://devfeed.tech/topics/ai-context-management.md>), [Qdrant](<https://devfeed.tech/topics/qdrant.md>), [Retrieval Augmented Generation (RAG)](<https://devfeed.tech/topics/retrieval-augmented-generation-rag.md>)

Tags: [agent-frameworks](<https://devfeed.tech/tags/agent-frameworks.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents-in-production](<https://devfeed.tech/tags/agents-in-production.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [connectors](<https://devfeed.tech/tags/connectors.md>), [context-engineering](<https://devfeed.tech/tags/context-engineering.md>), [context-layer](<https://devfeed.tech/tags/context-layer.md>), [database](<https://devfeed.tech/tags/database.md>), [hybrid-search](<https://devfeed.tech/tags/hybrid-search.md>), [operations](<https://devfeed.tech/tags/operations.md>)

### AI overview

Elastic describes a context engine built on Elasticsearch that prepares structured knowledge indicators from enterprise data and retrieves relevant context for AI agents. In a simulated account-analysis task, the same agent and model used 9 tool calls instead of 33, took 64 seconds instead of 213, and used 71% fewer input tokens; the article also says it surfaced a relationship the baseline missed.

### Source excerpt

With a context engine, the same agent and model answered in 64 seconds using 9 tool calls. The baseline took 213 seconds, 33 calls, and missed a key relationship entirely.

## The fastest work is the work you never do: 100x faster sorted queries in Elasticsearch

DevFeed: [The fastest work is the work you never do: 100x faster sorted queries in Elasticsearch](<https://devfeed.tech/articles/the-fastest-work-is-the-work-you-never-do-100x-faster-sorted-queries-in-elasticsearch-78803.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/elasticsearch-performance-tuning-esql-topn>)

Author: Attila Sahi

Published: 2026-10-05T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [search-engines](<https://devfeed.tech/topics/search-engines.md>), [Elastic Cloud](<https://devfeed.tech/topics/elastic-cloud.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [es-ql](<https://devfeed.tech/tags/es-ql.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [inside-elastic](<https://devfeed.tech/tags/inside-elastic.md>), [lucene](<https://devfeed.tech/tags/lucene.md>)

### AI overview

Elasticsearch 9.6 introduces min-competitive TopN for ES|QL queries sorted by @timestamp. As the TopN heap fills, ES|QL shares its competitive threshold with Lucene and across execution drivers, allowing work on candidates that cannot improve the result to be skipped. In a benchmark over 346 million Nginx log documents, the query fell from about 500 seconds to 5 seconds on one node and gained 37x on Elastic Cloud Serverless. Benefits vary with the query and data; the current optimization applies to @timestamp-sorted queries.

### Source excerpt

How Elasticsearch 9.6 lets ES|QL push its current TopN threshold into Lucene, and how that 100x shows up on some sorted queries and barely registers on others.

## Check 100 candidates, not 10 million documents: Faster kNN filters in Elasticsearch

DevFeed: [Check 100 candidates, not 10 million documents: Faster kNN filters in Elasticsearch](<https://devfeed.tech/articles/check-100-candidates-not-10-million-documents-faster-knn-filters-in-elasticsearch-78859.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/knn-filter-post-filtering>)

Author: Panagiotis Bailis

Published: 2026-10-02T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [hybrid-search](<https://devfeed.tech/topics/hybrid-search.md>), [Qdrant](<https://devfeed.tech/topics/qdrant.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [inside-elastic](<https://devfeed.tech/tags/inside-elastic.md>), [lucene](<https://devfeed.tech/tags/lucene.md>), [search](<https://devfeed.tech/tags/search.md>), [vector](<https://devfeed.tech/tags/vector.md>), [vector-database](<https://devfeed.tech/tags/vector-database.md>), [vector-search](<https://devfeed.tech/tags/vector-search.md>)

### AI overview

Elasticsearch introduces adaptive post-filtering for filtered kNN queries, choosing per query whether to filter before or after vector search. In benchmarks on a 10-million-vector corpus, post-filtering was faster in 104 of 120 filtered comparisons, with recall within 0.01 of pre-filtering. The article describes candidate sizing, retry and fallback behavior, and the feature's planned default availability in Elasticsearch 9.6.

### Source excerpt

Elasticsearch now decides for each query whether to run a kNN filter before or after the vector search. On a 10M-vector corpus post-filtering was faster in 104 of 120 benchmark pairs while still returning k results.

## Signed, sealed, delivered: ECK certificate management with Vault and cert-manager

DevFeed: [Signed, sealed, delivered: ECK certificate management with Vault and cert-manager](<https://devfeed.tech/articles/signed-sealed-delivered-eck-certificate-management-with-vault-and-cert-manager-78757.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/eck-certificate-management-enterprise-pki>)

Author: Anand Vyas

Published: 2026-10-01T00:00:00Z

Content type: tutorial

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [HashiCorp Vault](<https://devfeed.tech/topics/hashicorp-vault.md>), [Secrets Management](<https://devfeed.tech/topics/secrets-management.md>), [certificates](<https://devfeed.tech/topics/certificates.md>), [sealed secrets](<https://devfeed.tech/topics/sealed-secrets.md>)

Tags: [certificate](<https://devfeed.tech/tags/certificate.md>), [certificates](<https://devfeed.tech/tags/certificates.md>), [container](<https://devfeed.tech/tags/container.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [operations](<https://devfeed.tech/tags/operations.md>), [pki](<https://devfeed.tech/tags/pki.md>)

### AI overview

The tutorial shows how to replace ECK's self-signed Elasticsearch certificates with enterprise PKI certificates. HashiCorp Vault keeps and uses the intermediate CA key outside Kubernetes, while cert-manager requests, renews, and rotates certificates; its CSI driver delivers transport certificates directly to Elasticsearch pods. The article walks through Vault and Kubernetes authentication, issuer and certificate setup, trust configuration, deployment, and verification.

### Source excerpt

Replace ECK's self-signed certificates with your enterprise PKI without renewing a certificate by hand, as Vault signs with a CA key sealed outside Kubernetes and cert-manager delivers fresh certificates to every Elasticsearch pod.

## Using Jev as a search reranker: benchmarks and how to implement

DevFeed: [Using Jev as a search reranker: benchmarks and how to implement](<https://devfeed.tech/articles/using-jev-as-a-search-reranker-benchmarks-and-how-to-implement-78762.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/ecommerce-search-reranking-llm-alternative-jev>)

Author: Dustin Coates

Published: 2026-09-30T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [code search](<https://devfeed.tech/topics/code-search.md>), [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [hybrid-search](<https://devfeed.tech/topics/hybrid-search.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [sentence-transformers](<https://devfeed.tech/topics/sentence-transformers.md>), [Elastic Cloud](<https://devfeed.tech/topics/elastic-cloud.md>)

Tags: [0-1](<https://devfeed.tech/tags/0-1.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [bm25](<https://devfeed.tech/tags/bm25.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [hybrid-search](<https://devfeed.tech/tags/hybrid-search.md>), [implement](<https://devfeed.tech/tags/implement.md>), [integrations](<https://devfeed.tech/tags/integrations.md>), [jev](<https://devfeed.tech/tags/jev.md>), [python](<https://devfeed.tech/tags/python.md>), [relevance](<https://devfeed.tech/tags/relevance.md>), [reranking](<https://devfeed.tech/tags/reranking.md>), [search](<https://devfeed.tech/tags/search.md>), [shopping](<https://devfeed.tech/tags/shopping.md>)

### AI overview

The article describes a benchmark of Jev, TypeSafe AI's System One model, for reranking ecommerce search results retrieved by Elasticsearch. On a 250-query Amazon Shopping Queries test set, its best composite policy raised nDCG@10 from 0.9351 to 0.9565 and Exact MRR from 0.9201 to 0.9616. The benchmark compares several Jev scoring policies with BM25, hybrid retrieval, and a Jina reranker, and provides a Python implementation. The authors caution that this is one test on one dataset, with a high proportion of Exact judgments.

### Source excerpt

We had Jev score Elasticsearch hybrid search results and let a short Python policy do the reranking, taking nDCG@10 from 0.9351 to 0.9565 on 250 Amazon Shopping Queries, and the code is all here.

## How Elasticsearch Serverless hollow shards cut indexing-node shutdowns by 30%

DevFeed: [How Elasticsearch Serverless hollow shards cut indexing-node shutdowns by 30%](<https://devfeed.tech/articles/how-elasticsearch-serverless-hollow-shards-cut-indexing-node-shutdowns-by-30-78812.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/elasticsearch-serverless-hollow-shards>)

Author: Iraklis Psaroudakis,Eugene Cistiakovas

Published: 2026-09-29T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [Elastic Cloud](<https://devfeed.tech/topics/elastic-cloud.md>)

Tags: [autoscaling](<https://devfeed.tech/tags/autoscaling.md>), [cost](<https://devfeed.tech/tags/cost.md>), [elastic](<https://devfeed.tech/tags/elastic.md>), [elastic-cloud](<https://devfeed.tech/tags/elastic-cloud.md>), [elastic-cloud-serverless](<https://devfeed.tech/tags/elastic-cloud-serverless.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [inside-elastic](<https://devfeed.tech/tags/inside-elastic.md>), [operations](<https://devfeed.tech/tags/operations.md>)

### AI overview

Elasticsearch Serverless hollow shards release idle Lucene writers and segment readers from indexing-node memory, retaining shard metadata until another write requires the engine to reload. In its first month in production, the change reduced high-percentile fleet indexing-node shutdown times by up to 30% and lowered indexing-tier heap usage. The tradeoff is a median 218 ms delay on the first write to a hollow shard; search latency is unchanged. Autoscaling still counts hollow shards as full, and the article notes that handling bursts of simultaneous unhollowing remains a planned improvement.

### Source excerpt

Idle indexing shards in Elasticsearch Serverless now drop their Lucene writers and segment readers from memory until the next write, freeing indexing-tier heap at the cost of a median 218ms wait on that first write.

## The best LLM writes correct Elasticsearch ES|QL 59% of the time. Here's what breaks the other 41%.

DevFeed: [The best LLM writes correct Elasticsearch ES|QL 59% of the time. Here's what breaks the other 41%.](<https://devfeed.tech/articles/the-best-llm-writes-correct-elasticsearch-es-ql-59-of-the-time-here-s-what-breaks-the-other-41-78894.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/text-to-esql-benchmark>)

Author: Jeffrey Rengifo,Gustavo Llermaly

Published: 2026-09-28T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [text2sql](<https://devfeed.tech/topics/text2sql.md>), [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [aggregate](<https://devfeed.tech/tags/aggregate.md>), [ai](<https://devfeed.tech/tags/ai.md>), [cardinality](<https://devfeed.tech/tags/cardinality.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [es-ql](<https://devfeed.tech/tags/es-ql.md>), [llm](<https://devfeed.tech/tags/llm.md>), [model](<https://devfeed.tech/tags/model.md>), [queries](<https://devfeed.tech/tags/queries.md>), [query-languages](<https://devfeed.tech/tags/query-languages.md>), [sql](<https://devfeed.tech/tags/sql.md>)

### AI overview

The article benchmarks four language models on 6,000 ES|QL queries using BIRD's Mini-Dev questions and answer rows. The strongest model returned correct data on 59% of single-attempt queries. The analysis identifies mismatched join keys, invalid SQL syntax, counts inflated by one-to-many joins, and guessed data values as major failure sources. A compact ES|QL reference helped smaller models but reduced the strongest model's accuracy, while sample values can help address errors caused by assumptions about stored data.

### Source excerpt

We scored 6,000 ES|QL queries from four models against BIRD's answer key. Most misses come from mismatched join keys, SQL syntax the parser rejects, counting after a one-to-many join, or a value the model guessed.

## GPU-accelerated vector indexing in Elasticsearch with NVIDIA cuVS: 138M vectors in under 10 minutes

DevFeed: [GPU-accelerated vector indexing in Elasticsearch with NVIDIA cuVS: 138M vectors in under 10 minutes](<https://devfeed.tech/articles/gpu-accelerated-vector-indexing-in-elasticsearch-with-nvidia-cuvs-138m-vectors-in-under-10-minutes-78899.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/vector-indexing-gpu-elasticsearch-cuvs>)

Author: Bao Tong,Manas Singh,Corey Nolet

Published: 2026-09-25T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [CUDA](<https://devfeed.tech/topics/cuda.md>), [NVIDIA Blackwell](<https://devfeed.tech/topics/nvidia-blackwell.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-accelerated](<https://devfeed.tech/tags/gpu-accelerated.md>), [index-data](<https://devfeed.tech/tags/index-data.md>), [indexing](<https://devfeed.tech/tags/indexing.md>), [inside-elastic](<https://devfeed.tech/tags/inside-elastic.md>), [latency](<https://devfeed.tech/tags/latency.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [vector-database](<https://devfeed.tech/tags/vector-database.md>)

### AI overview

Elasticsearch and NVIDIA cuVS use GPUs to build vector indexes, with a benchmark indexing 138 million vectors in under 10 minutes. The article reports up to 7x higher indexing throughput, 6x lower p90 search latency during concurrent indexing, similar recall to CPU-built indexes, and shorter force-merge times.

### Source excerpt

Moving index builds to the GPU leaves the CPU free for queries, which is how vector indexing throughput went up 7x and p90 search latency fell 6x while indexing ran, with no change to recall.

## Ask Elastic Agent Builder why it's slow: Natural-language trace analysis

DevFeed: [Ask Elastic Agent Builder why it's slow: Natural-language trace analysis](<https://devfeed.tech/articles/ask-elastic-agent-builder-why-it-s-slow-natural-language-trace-analysis-78711.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/agent-performance-trace-analysis-agent-builder>)

Author: Meghan Murphy

Published: 2026-09-24T00:00:00Z

Content type: tutorial

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [LLM observability](<https://devfeed.tech/topics/llm-observability.md>), [API Logging](<https://devfeed.tech/topics/api-logging.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [elastic-agent](<https://devfeed.tech/tags/elastic-agent.md>), [es-ql](<https://devfeed.tech/tags/es-ql.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [operations](<https://devfeed.tech/tags/operations.md>), [performance](<https://devfeed.tech/tags/performance.md>)

### AI overview

The article explains how Elastic Agent Builder's built-in trace-analysis skill turns natural-language questions into ES|QL queries over Agent Builder OpenTelemetry traces. It provides examples for examining token usage by model, tool error rates, slow conversation turns, and recent prompts, and describes relevant privacy settings and query limits.

### Source excerpt

Four agent performance questions your Agent Builder traces can answer, covering token spend by model, tool error rates, slow conversation turns, and recent prompts. The ES|QL for each is here, including the type cast SUM() needs.

## Agentic workflows in Elasticsearch: pause an AI agent for human approval, resume 72 hours later

DevFeed: [Agentic workflows in Elasticsearch: pause an AI agent for human approval, resume 72 hours later](<https://devfeed.tech/articles/agentic-workflows-in-elasticsearch-pause-an-ai-agent-for-human-approval-resume-72-hours-later-78715.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/ai-agent-orchestration-human-approval-workflow>)

Author: Alex Salgado

Published: 2026-09-21T00:00:00Z

Content type: tutorial

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>), [human review](<https://devfeed.tech/topics/human-review.md>), [Loop Engineering](<https://devfeed.tech/topics/loop-engineering.md>)

Tags: [agent-orchestration](<https://devfeed.tech/tags/agent-orchestration.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [index-data](<https://devfeed.tech/tags/index-data.md>), [operations](<https://devfeed.tech/tags/operations.md>), [workflows](<https://devfeed.tech/tags/workflows.md>)

### AI overview

This tutorial demonstrates an Elasticsearch workflow for diagnosing failed data ingestion and remediating it with AI agents while keeping human approval at key steps. Workflow state and audit records persist in Elasticsearch, allowing an approval gate to wait for up to 72 hours before resuming. The example includes alert-triggered diagnosis, approval or revision, automated reindexing, and a final verification or escalation decision.

### Source excerpt

Build AI agent orchestration where the workflow waits for a human approval and then executes the fix on its own, with nothing extra to provision and the whole decision trail queryable in Elasticsearch.

## AI video search with Elasticsearch and Jina: Find the exact seconds of footage you need

DevFeed: [AI video search with Elasticsearch and Jina: Find the exact seconds of footage you need](<https://devfeed.tech/articles/ai-video-search-with-elasticsearch-and-jina-find-the-exact-seconds-of-footage-you-need-78721.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/ai-video-search-jina-embeddings>)

Author: JD Armada

Published: 2026-09-18T00:00:00Z

Content type: tutorial

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [elasticsearch](<https://devfeed.tech/topics/elasticsearch.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [cosine similarity search](<https://devfeed.tech/topics/cosine-similarity-search.md>), [Cohere API for embeddings](<https://devfeed.tech/topics/cohere-api-for-embeddings.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [elasticsearch](<https://devfeed.tech/tags/elasticsearch.md>), [jina-ai](<https://devfeed.tech/tags/jina-ai.md>), [query](<https://devfeed.tech/tags/query.md>), [search](<https://devfeed.tech/tags/search.md>), [vector](<https://devfeed.tech/tags/vector.md>), [vector-database](<https://devfeed.tech/tags/vector-database.md>)

### AI overview

This tutorial explains how to build an AI video search app that finds footage by text description and returns matching timestamps. It covers scene-based chunking, Jina multimodal embeddings, Elasticsearch vector indexing and nearest-neighbor search, and deduplicating results so editors can browse distinct clips. It also discusses proxy sizing and the index's memory costs as footage libraries grow.

### Source excerpt

Cut each clip at its shot boundaries and embed every scene as a vector, and a plain text query gives back the file plus the exact seconds to drop on a timeline.

## jina-ocr-v1: One OCR model for layout, tables, math and 100+ languages

DevFeed: [jina-ocr-v1: One OCR model for layout, tables, math and 100+ languages](<https://devfeed.tech/articles/jina-ocr-v1-one-ocr-model-for-layout-tables-math-and-100-languages-78870.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/ocr-model-jina-ocr-v1-document-parsing>)

Author: Scott Martens

Published: 2026-09-17T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [Amazon Textract](<https://devfeed.tech/topics/amazon-textract.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [LLM observability](<https://devfeed.tech/topics/llm-observability.md>), [Retrieval Augmented Generation (RAG)](<https://devfeed.tech/topics/retrieval-augmented-generation-rag.md>)

Tags: [ai-tools](<https://devfeed.tech/tags/ai-tools.md>), [api](<https://devfeed.tech/tags/api.md>), [jina-ai](<https://devfeed.tech/tags/jina-ai.md>), [layout](<https://devfeed.tech/tags/layout.md>), [model](<https://devfeed.tech/tags/model.md>), [ocr](<https://devfeed.tech/tags/ocr.md>)

### AI overview

The article describes jina-ocr-v1, an end-to-end document parser that converts scans, photographs, slides, and handwriting in more than 100 languages into structured text. It covers its layout-aware output, table and math extraction, handwriting recognition, performance comparisons, model architecture, input constraints, and API and local access options.

### Source excerpt

jina-ocr-v1 scores 83.4 on olmOCR-bench with 570 million active parameters, the highest of any OCR model under 600 million, and it outscores GPT-5.2 on OmniDocBench.

## One button, three places: How we rebuilt Kibana's page headers with stricter APIs

DevFeed: [One button, three places: How we rebuilt Kibana's page headers with stricter APIs](<https://devfeed.tech/articles/one-button-three-places-how-we-rebuilt-kibana-s-page-headers-with-stricter-apis-78752.md>)

Original publisher: [Read original article](<https://www.elastic.co/search-labs/blog/design-system-governance-kibana-page-headers>)

Author: Anton Dosov,Ryan Keairns,Krzysztof Kowalczyk,Alex Marhaba

Published: 2026-09-16T00:00:00Z

Content type: article

Language: en

Sources: [Elasticsearch Labs](<https://devfeed.tech/sources/elasticsearch-labs.md>)

Topics: [User interface design](<https://devfeed.tech/topics/ui-design.md>), [engineering-culture](<https://devfeed.tech/topics/engineering-culture.md>), [kibana](<https://devfeed.tech/topics/kibana.md>)

Tags: [breadcrumbs](<https://devfeed.tech/tags/breadcrumbs.md>), [button](<https://devfeed.tech/tags/button.md>), [design-system](<https://devfeed.tech/tags/design-system.md>), [developer-experience](<https://devfeed.tech/tags/developer-experience.md>), [governance](<https://devfeed.tech/tags/governance.md>), [inside-elastic](<https://devfeed.tech/tags/inside-elastic.md>), [kibana](<https://devfeed.tech/tags/kibana.md>)

### AI overview

Elastic describes replacing Kibana's open-ended page header APIs with typed contracts and a shared shell that controls presentation. The redesign aims to make controls and actions consistent across teams, removes displayed breadcrumbs because navigation supplies hierarchy, and limits visible actions before they move into an overflow menu. The redesigned chrome is available in Elastic Cloud Serverless and is scheduled for Elastic Cloud Hosted and self-managed users in 9.6.

### Source excerpt

We gave Kibana's shared shell typed contracts, which is how design system governance became the default, and why the new page headers have no breadcrumbs.

[Next page](<https://devfeed.tech/sources/elasticsearch-labs.md?cursor=WyIyMDI2LTA5LTE2VDAwOjAwOjAwKzAwOjAwIiwgIjQwYzVlMzA1LTNiYjktNGMwNS1iYWU3LWM1ODRkZWVjZDA5ZCJd>)