Core-API HTTP/2 servers and Apollo Federation entity resolvers are now traced. Test Optimization adds EFD retry count override, fixes Jest suite skipping, and addresses multiple stability issues. Also ships a dd-trace-electron package with reduced dependencies.
Core-API HTTP/2 servers and Apollo Federation entity resolvers are now traced. Test Optimization gains EFD retry count override, Jest suite skipping diagnosis, and multiple fixes for retry isolation, duration handling, and git upload.
Adds tracing for Model Context Protocol server requests and client operations, propagates LLM session_id to following spans and across services, and samples event loop delay per iteration on Node 26.5+. Fixes include Azure Functions span links for non-HTTP triggers, GraphQL errorExtensions on mercurius request spans, and OpenTelemetry span processor wiring.
Adds tracing for Model Context Protocol (MCP) server requests and client operations, and propagates LLM observability session_id to following spans and across services. Also fixes Azure Functions span links, GraphQL mercurius errorExtensions, and OpenTelemetry span processor wiring, plus AppSec improvements and test optimization fixes.
Added Mercurius support with a top-level graphql.request span. Fixed a hung exit by forwarding abort telemetry synchronously, and capped MsgpackChunk growth at the agent intake limit.
GraphQL requests now produce a top-level span for mercurius integrations. MsgpackChunk growth is capped at the agent intake limit to prevent unbounded memory, and AppSec reports request headers on Lambda spans.
Added tracing support for @anthropic-ai/claude-agent-sdk and Vercel AI SDK v7. GraphQL instrumentation now supports collapse, depth, variables, and error extensions via environment variables, and MongoDB Collection#bulkWrite operations now open a parent span. Also improved stability across Fastify, child_process, Crash Tracking, and Test Optimization features.
Added instrumentation support for Anthropic's Claude Agent SDK and expanded GraphQL observability with configurable collapse, depth, variables, and error extensions via environment variables. Also added Vercel AI SDK v7 support, MongoDB Collection#bulkWrite parent span, and multiple fixes across Fastify, OpenTelemetry bridging, crash tracking, and test optimization.
OpenTelemetry OTLP trace metrics support enables native ingestion, and AppSec in-application WAF now functions inside AWS Lambda. Plus durable operation tagging, vitest no-worker init mode, and fixes for MongoDB nosql-analysis scoping, AWS durable execution checkpoints, and data streams flush behavior.
Drops support for Node.js 18 and 20, removes deprecated whitelist/blacklist plugin types, gates legacy addLink overload, and raises minimum versions for Cypress (12.0.0), Jest (28.0.0), Mocha (8.0.0), and Playwright (1.38.0). LLM Observability span resource names are now consistent between openai v3 and v4. Includes vitest no-worker init mode, OTLP trace metrics support, and in-App WAF support for Lambda.
Opt-in OpenTelemetry HTTP semantic conventions (DD_TRACE_OTEL_SEMANTICS_ENABLED) and improved OTel trace compatibility. Restored loopback middleware phase ordering and fixed late routes superseding endpoint fallback. Includes AI Guard standalone mode support and sampling decisions for LLM Observability.
Fixed a redis db.name placeholder that poisoned inferred service naming, and route agentless spans to the regional data-center intake. New SQS and Kinesis consumer spans for promise-based receives, and Next.js app-route errors are now captured via onRequestError.
Added allocations profiling support and fixed minor mark-sweep GC events for profiling. Also improved AppSec performance on object length and added --auto-discovery flag to datadog-ci junit upload.
Fixed an issue where the Datadog API key was exposed in cleartext for agentless intake, and preserved caller trace context for pooled MySQL2 queries. Added toolDefinitions support in LLMObs.annotate() and span tags to AI Guard spans for anomaly detection.
Added support for AWS Bedrock Converse and ConverseStream APIs in LLM Observability. Fixed OTLP spans to include full 128-bit trace IDs instead of truncated IDs. Also surfaced prompt cache tokens for Vercel AI SDK across all supported providers, fixed Cucumber v13 parallel mode support, and pinned Vitest 3 to Node 18.
Reverted the application of http.endpoint and queryStringObfuscation to client spans. Added OpenFeature flag evaluation enrichment for APM traces, experimental NATS support, Cypress TIA line coverage reporting, and Hono single-handler route resource naming. Performance improvements to msgpack encoding and tag formatting; MongoDB query obfuscation and tagging speedups; fixes to Jest coverage metric reporting, debugger reference desugaring, and OracleDB tracing suppression.
Fixed an unfinished CONNECT span in undici instrumentation and delta temporality not being applied to OTLP counter exports. Also reduced Redis and ioredis instrumentation overhead, skipped stacktrace symbolication on GraphQL validation errors, and fixed opentracing type imports.
Added initial support for Electron applications. Fixed config name validation to check both index.d.ts and index.d.v5.ts, kept v5 IAST programmatic aliases working, stopped accumulating DBM comments on reused PostgreSQL query objects, and fixed kafkajs to read clusterId from existing metadata.
Fixed span event handling to tolerate non-string names and whitespace-suffixed step functions JSON. Enhanced LLM Observability span processor to track span kind, embedding, and retrieval input and output.
f121c4281f] - (SEMVER-PATCH) fix(scripts): harden mocha-parallel-files against crashes and silent … (Ruben Bridgewater) #8243
*…
