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

# Missing conversation turns after lightweight de‑id migration

> How to diagnose missing conversation turns after Gusto's de-identification pipeline migration and request historical recovery.

> [!NOTE]
>
> **Applies to:**
>
> * Plan - 
> * Deployment - Braintrust-hosted
> *
> * Use case - Investigate missing conversation turns following Gusto's lightweight de-identification pipeline migration and request historical recovery

## Summary

This guidance describes missing turns following Gusto's lightweight de-identification pipeline migration. Conversation turns appear in the Raw Full trace but are missing from a saved conversation View. Confirm the data in Raw, compare the saved View definition, and request historical recovery from the team that owns your de-identification pipeline if needed.

## What is happening

Some traces may not have been fully processed during the cutover, which could explain the missing content. The raw span data still contains the original inputs.

The cutover accounted for most reported issues. The customer reported no missing turns in recent logs.

## Fix or suggestion

### Option 1: Diagnose and request historical recovery

1. Open the trace in [** Logs**](https://www.braintrust.dev/app/~/logs). Select **Spans** → **Raw** → **Full trace**. Confirm the missing turn appears there.
2. Use the same time range and the same saved View in both organizations. Note the View name.
3. Open the saved View editor and copy the View code from the **Edit** pane. Inspect filters and any span-data fetch logic that could exclude the turn.
4. If the turn exists in Raw but is absent from the View and the View code does not filter it, request historical recovery from the team that owns your de-identification pipeline. Include the trace IDs, the saved View name, the copied View code, and an example missing turn. Mention your concern about affected scorer data.

### Option 2: Immediate workaround

Use the Raw Full trace view or export the raw spans for the affected time range. The Raw view shows the original inputs even when turns are missing from the saved View.

## How to confirm it worked

* After historical recovery, open the same trace and select the same time range and saved View. Check whether the missing turn appears in the rendered View.

## Notes

Gusto created TIGER-1137 to track a backfill of missing safe logs.


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