← Back to changelog

May 20, 2026

Enriched observations is the default for new Cloud projects

Niklas Semmler

New Langfuse Cloud projects use enriched observations for blob storage, PostHog, and Mixpanel exports. The legacy traces/observations sources stay available on existing projects and self-hosted deployments.

Blob storage, PostHog, and Mixpanel integrations on Cloud projects created on or after 2026-05-20 use the Enriched observations (recommended) source. The Traces and observations (legacy) and Traces and observations (legacy) and enriched observations sources are hidden from the UI for these projects.

For blob storage, the REST API also rejects legacy values for post-cutoff projects with 400 BAD_REQUEST. PostHog and Mixpanel have no public REST surface.

What this means in practice:

• One row per observation, with trace context inlined — user_id, session_id, tags, release, trace_name, and environment ride along on the same row. No warehouse-side join needed.
• Per-column control (blob storage) — pair the source with field groups to keep just the columns your pipeline needs.
• Fewer files per window (blob storage) — only observations_v2/ and scores/, instead of traces/, observations/, and scores/.

Nothing changes for existing projects (Cloud or self-hosted) or self-hosted deployments: all three sources remain available. The legacy sources are still supported there for backwards compatibility, but the recommendation is the same — switch to Enriched observations when you're ready.

If a downstream consumer expects the legacy traces/ and observations/ file shapes, see the Export Field Reference for the column map between the two layouts.