Skip to main content
This page covers field definitions only. For the order to read the response in and how to interpret it, see reading analysis results.

Endpoint

Request

Path parameters

Response structure

The response fields fall into two groups.
Which fields land in extractedData is decided by the workflow’s output schema. Omni does not force any common keys. What is fixed is that every leaf value is wrapped in the same envelope (value · reasoning · sourceStep).

Top-level fields


confidence Analysis Score

The score shown at the top of the analysis detail screen in the dashboard.

confidence.components

When a part is null, the weights of the remaining parts are normalized. When all three are null, score and level are null too.

details score deductions

Lists each item that lowered the confidence score. Empty array when nothing was deducted. reasonCode is an extensible enum. For the current values and how to read them, see reading analysis results.

riskAssessment

How riskScore is calculated

It is computed from the per-action verdicts alone. Document content and extraction results play no part.
The base is 25, so riskScore can never fall below 25. When every action passes, it is exactly 25 (low). The 0–25 range in the table above is the level boundary; a value of 0–24 does not occur in practice.
riskFactors carries one {actionName} verification failed line per failed action, plus a single {count} action(s) need manual review line when any action needs review.

systemMetadata

systemMetadata.outputFormatting

Branch on this status before you read extractedData. How to branch is covered in reading analysis results.

systemMetadata.workflowHistory

Holds only the work that actually ran. The full step definitions are in workflowActions from GET /workflows/:workflowId.

systemMetadata.tokenUsage

tokenUsagePerWorkId uses work IDs as keys and the structure above as values.

rawActionResults

The raw, pre-assembly results the actions returned. Keys are action names, and which actions exist follows the workflow’s workflowActions definition.

rawActionResults mcpResults

The same calls as agentAuditLog.steps[].mcpcalls[], organized by action. It uses camelCase, like the top level.

agentAuditLog

An object with two keys, steps and summary.
agentAuditLog is an object, not an array, and its inner fields use snake_case (work_id · executed_at · duration_ms). That differs from the camelCase of the top-level fields.

agentAuditLog.steps

agentAuditLog.steps mcpcalls

agentAuditLog.summary


findings

result is agentAuditLog.steps[].status mapped for the UI (needs_review becomes warning).

recommendations

targetItems


Example response

A real response is much longer. The example below carries every top-level key but shows only one representative entry for each array and nested object.

Enum summary

Error codes

Common errors (400 validation · 401 authentication · 403 permission · 500 server) are listed in the API overview.