OpenSearch vs OpenTelemetry

Expanding the OpenSearch Community: Welcoming our new OpenSearch Ambassadors

The OpenSearch Ambassador Program expands to over 30 global leaders, welcoming 11 new community advocates in our September 2026 cohort. Meet the new ambassadors driving open-source search and AI forward.

Online index migration and shard scaling in OpenSearch with the AOSC plugin

Learn how the open-source AOSC plugin migrates live OpenSearch indexes—changing mappings, settings, or shard counts—without losing writes or requiring downtime.

OpenTelemetry Go Logs API and SDK reach release candidate status

OpenTelemetry Go v1.47.0-rc.1 is here. This release promotes the Logs API and SDK to release candidate (RC), the final stage before we provide stable v1 compatibility guarantees. We believe the design is ready, and now we need the community to test …

LLMOps for Compound AI Systems: Observability & Cost

The article argues that most GenAI pilots fail in production due to insufficient system-level engineering rather than poor models. It presents an LLMOps playbook for compound AI systems (embedders, retrievers, vector stores, re-rankers, validators, tool calls, and multiple LLMs) centered on five controls: a model gateway for routing and budgeting, pipeline-level traces for end-to-end observability, semantic caching keyed by query embeddings, lightweight eval gates for safety and quality, and tiered scaling of heavy infrastructure. A concrete engineering example reports a 38% reduction in token spend and 25% lower median latency after implementing a gateway, semantic cache, and tracing. The post includes a short pseudocode example (using qdrant-style vector operations) and an operational checklist for iterating LLMOps as an operating model.

• The article defines five LLMOps controls: model gateway, pipeline-level traces, semantic caching, eval gates, and tiered scaling.
• Author recommends using OpenTelemetry-compatible spans to instrument embed, search, rerank, prompt build, LLM call, and tool call stages.
• A cited engineering example achieved a 38% reduction in token spend and 25% lower median latency after implementing three LLMOps controls.

agent-cost: Measure LLM Usage, Separate Task Attribution

The author describes agent-cost, a small tooling primitive that reads local logs from LLM CLIs (e.g., Claude Code and Codex) to produce auditable, machine-readable usage facts (model, token kind, timestamp, count) and an estimated price. The tool is designed to run with no network calls at runtime, carry a versioned price catalog (with SHA-256 digest), and keep session measurement distinct from task attribution. Unknown or unsupported pricing and ambiguous session-to-task bindings are surfaced (labels like "unpriced" or "lower_bound") rather than silently allocated. The author re-ran the published coding-agent-cost 0.1.0 package and notes a catalog version 2026-07-29 and workflows that validate the measure/v1 protocol and data quality.

• agent-cost reads local logs from LLM CLIs (examples: Claude Code and Codex) and normalizes usage events into facts containing model, token kind, timestamp, and count.
• At runtime agent-cost makes no network calls and declares no Python runtime dependencies; installation from PyPI still requires trust in the supply chain.
• agent-cost carries a versioned pricing catalog with a SHA-256 digest and marks unknown models/prices as 'unpriced' or 'lower_bound' instead of inventing values.