Skip to content

v2.1.0

Latest

Choose a tag to compare

@mpangrazzi mpangrazzi released this 13 Oct 10:49
· 2 commits to main since this release
39446c3

Hayhooks 2.1.0 introduces a unified terminal dashboard for deployment health, executions, and tracing, replacing the browser tracing dashboard. #275

Highlights

  • Browse deployments and server health, recent ordinary executions, and paginated durable execution history across deployments.
  • Inspect traces and component spans with filters, search, timing, errors, expandable trees, and a terminal waterfall. Switch views with d (deployments), e (executions), and v (traces).
  • Use live refresh, pause, manual refresh, themes, and keyboard or mouse navigation. The dashboard retains the last good snapshot when a data source is unavailable and performs GET requests only.
  • Export a noninteractive status snapshot with hayhooks dashboard --once --json; add --pipeline NAME for durable executions or --traces for local traces.
  • Add bounded, owner-scoped execution indexes for Memory and Redis stores, with pagination, retention, and resilient cleanup.

Get started

pip install --upgrade "hayhooks[tui]==2.1.0"
hayhooks run
# In another terminal, under the same OS account:
hayhooks dashboard

The terminal UI is optional and requires no Node.js or browser. Local trace inspection does not require an external exporter; optional OTLP export remains supported.

Migration and behavior changes

  • The React dashboard, /dashboard, --with-tracing-dashboard and its aliases, and --dashboard-path are removed. Use hayhooks dashboard --traces instead.
  • The local tracing API is now authenticated GET /traces. The old tracing config, entrypoint, SSE stream, and server-clear APIs and incremental cursor parameters are removed. API clients must poll bounded snapshots.
  • Local tracing capture and durable monitoring indexes are enabled by default. Disable them with --without-tracing and --without-monitoring, or the corresponding HAYHOOKS_TRACING_ENABLED=false and HAYHOOKS_MONITORING_ENABLED=false settings.
  • Rename server capture settings from HAYHOOKS_DASHBOARD_TRACE_* to HAYHOOKS_TRACING_*. The old slow-component setting becomes HAYHOOKS_TRACING_SLOW_COMPONENT_MS; obsolete frontend and trace-stream settings are removed.
  • Same-user POSIX clients discover a private operator credential for loopback servers. Remote servers and Windows require explicit operator authentication; set HAYHOOKS_TRACING_TOKEN on the server and HAYHOOKS_DASHBOARD_TOKEN on the client, or integrate verified host authentication.

Ordinary execution history and traces are bounded and process-local; durable history follows the execution store's retention. Standalone hayhooks mcp run and hayhooks a2a run servers do not expose the dashboard trace API in this release.

See the dashboard CLI guide and tracing migration guide.

Full changelog