Documentation

Product guide for platform teams.

These docs explain what the agent connects to, what metadata it uses, what customer data it never touches, and how governed remediation works.

What Omnitrace connects to

Databricks workspace APIs, system metadata, cloud cost metadata, infrastructure metadata, and Jira workflow context.

What Omnitrace reads

Operational metadata, configuration, usage, billing, performance symptoms, ownership, and verification signals.

What Omnitrace never reads

Customer table contents, business records, files, query result sets, and application payloads are outside the product boundary.

How remediation is governed

Every strategy has autonomy settings, approval paths, blast-radius controls, and read-back verification.

Connections

Omnitrace connects to operational systems, not business data stores.

A typical onboarding flow starts with read-only discovery. The agent uses Databricks and cloud APIs to understand spend, configuration, health, ownership, and remediation opportunities. Jira is used when teams want findings routed into an existing approval workflow.

Databricks

Workspace metadata, compute configuration, usage, jobs, SQL warehouse, table health, and reliability signals.

Cloud provider

Cost Explorer-style cost metadata, infrastructure metadata, and model endpoints hosted by the customer's provider when configured.

Workflow

Jira ownership, approval, status, and audit context for findings that require human review.

Metadata only

The agent reasons over metadata and telemetry.

Omnitrace uses operational evidence to decide whether a finding is real, who owns it, what it costs, what action is safe, and how to verify the final state.

Cluster and SQL warehouse configuration

Databricks usage and billing metadata

Job, query, and table health signals

Spark reliability symptoms such as OOM, spill, GC, skew, and runtime drift

Cloud cost and infrastructure metadata

Jira ticket, owner, and approval workflow context

Data boundary

Omnitrace does not access customer table contents or business records.

Customer data stays outside the Omnitrace product boundary. The agent does not need query result sets, table rows, data files, application payloads, PII, PHI, or business records to detect waste and reliability drift.

When AI reasoning is enabled for enterprise deployments, Omnitrace can use model endpoints hosted by the customer's cloud provider. That keeps operational metadata and prompt context within the customer's selected cloud boundary.

Read security overview

Governed remediation

The workflow is detect, explain, approve, apply, verify.

Step 1

Detect

Step 2

Explain

Step 3

Approve

Step 4

Apply

Step 5

Verify

Teams control autonomy per strategy: manual review, approval-required actions, or low-risk automatic remediation. Omnitrace records the policy, the action, the tool response, and the verification outcome.

Deployment readiness

Gather the context that makes rollout safe and measurable.

Workspace count and rough monthly Databricks spend

One or two painful cost or reliability examples

Primary rollout focus: FinOps, reliability, remediation, or governance

Preferred deployment boundary: hosted service, customer cloud, or private deployment