Observability

Org-wide platform operations and LLM Gateway cost and usage in one place

Because the platform is natively traceable, every request already carries a correlation ID through an event-driven backbone, and Observability is where that raw event stream becomes org-wide insight. Open it from Observability in the Governance sidebar. It has two tabs: Platform for organization-wide operations, and LLM for LLM Gateway cost and usage. Pick a time range (1h, 24h, 7d, 30d) from the top of the page; it applies to both tabs.

Observability here is scoped to your organization as a whole. The per-workspace and per-automation deep-dive (metrics, latency, error detail, usage, dependencies, and traces) now lives in the Builder, per workspace, under Builder > workspace > Observability. See Builder Observability. Rows in the Platform tab link straight through to the matching workspace’s Observability in Builder.

Platform tab

The Platform tab is an operations view across every workspace in your organization. It focuses on health and errors, not adoption; usage and adoption metrics live in AI Insights, and a link points there instead of duplicating them.

• Quick stats: Unique Users, Total Interactions, Error Rate, and Average RPM across the org for the selected period. The error rate turns red once it crosses the alert threshold.
• Top workspaces by errors: workspaces ranked by error count, each row showing interactions, users, error count, and error rate. Click a row to open that workspace’s Observability in Builder.
• Recent errors: the latest error events with their code, message, timestamp, and originating workspace. Each links through to the source workspace in Builder.

Some blocks render progressively: they stay empty until the org has activity in the selected period.

LLM tab

The LLM tab is the LLM Gateway cost and usage view, sourced from the analytics.llm.completion events emitted by the llm-gateway workspace.

• Headline cards: Total Cost, Input Tokens, Output Tokens, LLM Calls, and Carbon (kg CO₂eq).
• Cost by Model: per-model breakdown with cost share, token count, and call count.
• Cost by Provider: call count and average latency per provider.
• Cost Timeline: per-bucket cost trend across the selected period.

This is the right place to track LLM spend org-wide. For a single workspace’s cost and performance, use that workspace’s Observability in Builder.

Where per-workspace detail lives

The deep-dive that used to live here, health scores, latency percentiles, error tables with activity-feed deep links, usage timelines, dependency graphs, and the real-time event feed, now lives in the Builder, scoped to one workspace at a time.

To investigate a specific workspace or automation:

1. From the Platform tab, click the workspace row (or open the workspace directly in Builder).
2. Open Observability on that workspace in Builder.
3. Use its per-workspace tabs to drill into metrics, latency, errors, usage, dependencies, and traces.

See Builder Observability for the full per-workspace reference.

Access to raw events stays bound to each workspace’s ACLs. Governance surfaces aggregated, org-wide metrics (error counts, cost, RPM) suitable for a platform, security, or FinOps owner. Drilling from a metric down to the underlying events in Builder requires access to that workspace, either through an explicit share or platform-wide superadmin rights. The metrics travel up into the Governance scope; the underlying data stays compartmentalized below it.

Next steps

Model Governance

Control model access and costs

Builder Observability

Per-workspace metrics, latency, errors, and traces