Versions (relevant - OpenSearch/Dashboard/Server OS/Browser): 3.7
Describe the issue : I have an ISM policy with min_index_age, min_shard_size & min_doc_count. Since it’s an OR condition, I am having indices with 0 documents rolling over after 7 days.
I saw this github link and wanted to check when this feature will be available.
https://github.com/opensearch-project/index-management/pull/1667
pablo
August 21, 2026, 6:38pm
2
@muraliv This pull request was merged on Jul 19 and released with OpenSearch 3.8.0 as per release notes.
# OpenSearch and OpenSearch Dashboards 3.8.0 Release Notes
## Release Highlights
OpenSearch 3.8 expands the platform's search, AI, and observability capabilities with enhanced vector performance, broader agent integrations, and new tools to help you close the distance between ingestion and investigation.
### New and Updated Features
#### Search Modernization
* **Ingest vectors up to 4x faster with base64 encoding:** Base64-encoded vector ingestion for knn_vector fields eliminates JSON array serialization overhead, cutting network payload by 74% and delivering up to 4.16x higher bulk ingestion throughput with 83% lower median latency.
* **Deliver 2x faster radial search with improved recall:** A redesigned graph traversal bounds exploration on radial queries, delivering up to 2.1x higher throughput with 45% lower median latency while improving mean recall from 0.85 to 0.97.
* **Extend MCP integration across agent types:** Model Context Protocol support now extends to flow and conversational flow agents, enabling all four agent architectures to connect to external MCP-compliant tool servers using a consistent connector configuration.
* **Discover and customize external MCP tools:** A new list tools API programmatically discovers all available tools on a connected MCP server, and connector-level tool description overrides let you customize how your LLM perceives each tool without modifying the external server.
* **Stream ML predictions faster with gRPC transport:** New PredictModelStream and ExecuteAgentStream methods stream output token by token from remote models using protocol buffers over HTTP/2, delivering lower latency and reduced CPU overhead compared to REST streaming.
* **Scale LLM-powered evaluation across providers:** LLM-as-a-Judge in Search Relevance Workbench now supports any ML Commons connector, with ready-to-use blueprints for OpenAI, Azure OpenAI, DeepSeek, Ollama, Google Gemini, Anthropic Claude on Amazon Bedrock, and the Bedrock Converse API.
* **Parameterize search experiments with Mustache templates:** Search Relevance Workbench now supports Mustache template variables in search configurations, letting you build multi-parameter experiments with custom fields like category, brand, or status filters directly in your query DSL.
#### Observability and Analytics
show original