What happens to an uploaded log

File journey

1. Transfer. The browser sends the selection to the Unstacked process running in this local-only environment.
2. Validation. The server rejects unsupported executable content and applies upload, archive-entry, decompressed-size, and compression-ratio limits.
3. Parsing and masking. The parser masks detected credentials and personal data before building the normal analysis view. Pattern detection is best effort and cannot guarantee that every sensitive value is found.
4. Raw cleanup. Uploaded raw files and intermediate merged files are deleted when parsing finishes or fails.
5. Session cleanup. The analysis session expires 2 hours after its last use. Choosing Clear Logs or Finish & Clear deletes the current server-side session immediately.

When data leaves the parser

Standard log analysis

Does not call an AI provider or IP-intelligence provider. Uploaded log content stays within the running Unstacked deployment.

Focused AI analysis

Off by default. Only the text placed in the AI selection is sent, and only after a provider is connected and Analyze with AI is chosen.

IP diagnostics

A separate toolkit feature can send the observed IP address to the intelligence provider shown in its results. Turning enrichment off avoids that lookup.

Browser-local toolkit analyzers

Tools marked Browser-local—including Markdown Reader, support-bundle preflight, secret scanning, subnet calculation, trace-context, metrics, Kubernetes, runtime, plan, configuration, and payload analysis—process their pasted input or selected file samples in the browser. They do not send that input to Unstacked.

Local sanitizer learning

When you mark a file-sanitizer finding as not sensitive, the browser stores only a one-way hash of that matched value on this device. Recent scan history stores counts only. If you explicitly opt in to shared feedback, Unstacked stores only the curated rule ID and your Sensitive / Not sensitive decision in the shared rule catalog. File contents, filenames, matched values, IP addresses, and user identifiers are never sent as learning data.

Storage and operational records

Analysis session

A temporary parsed-log database, scoped to its signed-in owner in hosted mode. It expires after 2 hours of inactivity.

Job progress

Generated job identifiers, progress, counts, and timings remain in process memory for approximately 15 minutes after activity.

Aggregate usage counter

For a successful job, the application stores a generated job identifier, processed-file count, and creation time. It does not store filenames or log contents in that counter.

Infrastructure metadata

The hosting or network provider may process ordinary request metadata such as source IP, path, status, and timing. Region, backups, and infrastructure-log retention are deployment-operator settings, not fixed by this application.

Keep sensitive analysis local

Run the desktop or self-hosted application in local-only mode when a bundle cannot be sent to a hosted service. Local-only mode removes the hosted sign-in path and keeps processing within that deployment.

In an approved local-only deployment with unredacted retention explicitly enabled, the log viewer can offer a temporary Reveal original values control. It is off by default, unavailable on the hosted deployment, and does not change the default sanitized export.