releases.shpreview

AI-powered root cause analysis joins the observability workflow

May 11–17, 2026

Sentry Seer can now send root causes to GitHub Copilot for fixes, while Datadog releases the open-weights Toto 2.0 time series foundation model. Across both platforms, AI-assisted debugging moves from analysis to automatic remediation.

Sentry Seer sends root causes to Copilot and Claude

Sentry's AI-powered root cause engine Seer now feeds directly into code generation. In open beta, Seer can send identified root causes to GitHub Copilot to produce fixes, closing the loop from anomaly detection to pull request. Meanwhile, the Claude integration for Seer — which lets users trigger Claude to fix bugs from Seer's analysis — reached general availability this week. Together, these releases make Sentry one of the first observability platforms where automated root cause analysis automatically hands off to AI code generation.

Datadog Toto 2.0: time series foundation models open in five sizes

Datadog released Toto 2.0, a time series foundation model family available under open weights in five sizes from 4 million to 2.5 billion parameters. All sizes are trained from a single recipe, and the larger models show reliably better performance — a first for time series foundation models. Separately, Cloud Cost Management now attributes AI spend across providers, giving teams unified visibility into GPU and LLM API costs. And a new guide shows how RUM handles micro-frontend observability with automatic service attribution and build-time instrumentation.

GenAI spans get special treatment in Sentry and Datadog SDKs

Both Sentry and Datadog shipped SDK improvements for AI spans. Sentry Python 2.60.0 adds a stream_gen_ai_spans option that extracts GenAI spans from transactions and sends them as separate v2 envelope items, preventing them from being dropped when payloads exceed size limits. The same feature landed in Sentry JavaScript 10.53.0 as streamGenAiSpans, along with console instrumentation filtering and Node 26 support. The SDK also fixed a bug where user data was incorrectly gated for streamed spans.

Datadog's APM v5.103.0 added initial Electron support and fixed accumulated DBM comments on reused PostgreSQL query objects. The Python APM library v4.8.4 stopped a malformed dist-info entry from flooding stderr, and v4.8.3 migrated Azure CosmosDB spans from http.status_code tags to the db.response.status_code metric.

Security and platform updates

The Datadog Agent 7.78.4 patches CVE-2026-35469 in the SPDY/3 frame parser, which could let attackers exhaust process memory via crafted frames. On the Sentry side, Sentry CLI 3.4.2 fixes an argument injection vulnerability in the JavaScript API's serializeOptions function and stops sending auth tokens to object storage during snapshot operations. The RUM SDK v7.1.0 promoted trackResourceHeaders to GA and fixed a bug that could postpone batches indefinitely when upserting a view. Dash0 shipped several Operator releases with bump collector components and fixes for resource uniqueness verification and CRD status attributes.

Sentry dashboards now maintain a revision history of recent changes, giving teams better auditability for their monitoring views.

Releases covered