> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sndbox.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Logs and evidence

> Inspect the values, timing, diagnostics, and artifacts attached to the exact run revision.

The execution inspector separates a run summary from the timeline of node attempts. Select a node to switch between input, output, logs, error/skip detail, and browser diagnostics.

## Recorded node evidence

```ts title="NodeExecution" theme={"system"}
interface NodeExecution {
  nodeId: string;
  status: "idle" | "waiting" | "running" | "successful" | "failed" | "skipped" | "cancelled";
  startedAt?: string;
  completedAt?: string;
  durationMs?: number;
  input: unknown;
  output: unknown;
  logs: string[];
  retryCount: number;
  error?: { code: string; message: string; detail?: string; suggestion?: string };
  skipReason?: string;
  branchFollowed?: string;
  browserDiagnostics?: BrowserDiagnostics;
  inputItems: WorkflowItem[];
  outputItems: WorkflowItem[];
  collection?: CollectionEvidence;
}
```

Collection evidence records authoritative input/output/rejected counts, named branch counts, iteration and batch counts, ordering policy, stop reason, warnings, and bounded item samples. The inspector can search and filter samples and trace an item through its origin, parent, source node, positions, branch history, loop iteration, attempt, and correlation IDs.

The history preview is separate from runtime data. When a collection exceeds the configured preview count, `previewTruncated` is true while the authoritative count remains visible; downstream nodes still receive the complete runtime result.

## Redaction

Authorization, cookie, token, password, credential, and other protected fields are removed or redacted before ordinary logs and execution output. Webhook URLs and OAuth tokens remain inside host-owned connection handling.

## Remote checkpoints

The control plane appends a checkpoint after each completed node with execution/revision, node/version, attempt, result, input hash, output reference, side-effect class, idempotency key, completion time, and runner identity. `(execution, node, attempt)` is unique. Temporary browser session IDs are not resumable output references.

## Retention

Local evidence stays in the desktop database until the user removes it. Removing an execution also removes its loop-iteration checkpoints. Workspace execution detail follows the configured privacy and retention policy; outcome summary and audit evidence have separate handling rules.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.