🚀 Features
Section titled “🚀 Features”OpenTelemetry tracing for the Platform frontend
Section titled “OpenTelemetry tracing for the Platform frontend”Jun 11, 2026 · @avaq, @jachris
The Platform frontend can now export OpenTelemetry traces over OTLP, giving operators end-to-end visibility into the requests its backend handles.
Enable tracing and point it at your OTLP collector:
PUBLIC_ENABLE_TRACING=trueOTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318Traces are sent to ${OTEL_EXPORTER_OTLP_ENDPOINT}/v1/traces by default. Set OTEL_EXPORTER_OTLP_TRACES_ENDPOINT to override the traces endpoint directly.
When exporting to a collector that requires authentication, supply headers as comma-separated key=value pairs:
OTEL_EXPORTER_OTLP_HEADERS="authorization=Bearer <token>,x-tenant=acme"Common fields now appear as columns
Section titled “Common fields now appear as columns”Jun 3, 2026 · @gitryder
We improved the Stream view by displaying fields that are shared across all schemas in dedicated columns. Timestamps are the first supported field, making it easier to scan and compare at a glance while still allowing you to expand rows to view the full event. When no shared timestamp field is available, Stream view continues to use the existing layout.
🔧 Changes
Section titled “🔧 Changes”Security and performance improvements
Section titled “Security and performance improvements”Jun 4, 2026 · @avaq
We overhauled the frontend’s login and session handling. Compared to the previous release:
- OIDC provider metadata is now discovered once at application startup instead of on every session refresh, shaving the round-trip cost off many requests.
- User keys (access tokens) are now always cached server-side in the user’s session, alleviating the cost of obtaining new access tokens for many requests.
- Users of Platform deployments using header-based auth (like Google IAP) are now also granted a login session, allowing them to benefit from session based caches. This means that access tokens and user metadata is no longer fetched externally on every request, potentially saving multiple round trips worth of time with each request.
- Session IDs no longer leak to JavaScript, and are solely kept inside secure HTTP cookies.
🐞 Bug Fixes
Section titled “🐞 Bug Fixes”Library tab no longer hangs after interrupted load
Section titled “Library tab no longer hangs after interrupted load”Jun 16, 2026 · @Zedoraps, @claude
The Library tab now loads reliably after a previous attempt was interrupted by a page reload or navigation. Previously the tab could get stuck at “Loading…” indefinitely until the browser session was cleared.
Examples drawer loads on nodes with sparse package metadata
Section titled “Examples drawer loads on nodes with sparse package metadata”Jun 16, 2026 · @Zedoraps, @claude
The Explorer’s Examples drawer no longer hangs at “Loading…” on Tenzir
Nodes that have installed packages whose examples are missing a name
or description (for instance several mappers under microsoft/,
fortinet/, otel/, and zscaler/ in the Tenzir Library).
Fix docs not loading
Section titled “Fix docs not loading”Jun 16, 2026 · @avaq
The documentation panel should now show the documentation website again, instead of a white page.
Delete dashboards when removing a static workspace
Section titled “Delete dashboards when removing a static workspace”Jun 4, 2026 · @gitryder
Previously, when a workspace was removed from the static configuration file, dashboards that users had created within that workspace were left behind as orphaned records in the database. These could resurface if a new workspace later reused the same workspace ID. Removing a static workspace now also deletes these user-created dashboards.
Explorer downloads on nodes using the legacy execution engine
Section titled “Explorer downloads on nodes using the legacy execution engine”Jun 3, 2026 · @tobim, @claude · #184
Downloading results from the Explorer now works on Tenzir Nodes v6+ that are
configured to run pipelines on the legacy execution engine (tenzir.neo: false). Previously, preparing the download failed on such nodes because the
pipeline that uploads the results relied on functionality that is only
available in the new execution engine. The download pipeline now always runs
on the new execution engine, regardless of the node’s configuration.