Update/remove specific field from historical traces?

[justine-arjo]

Describe your question

Hi Langfuse team,

I have a question regarding the immutability constraints of the observability data model.
In my current project, some traces contain a user.name field that I would now like to remove for privacy/PII-minimization reasons. I understand from the FAQ that the tracing data model is mostly immutable, and that updates sent with the same trace ID can overwrite certain fields, especially if the full event (with the original timestamp) is re-ingested.

Before I proceed, could you clarify the following:

1. Is it possible to remove or nullify a specific field across all existing traces?
2. If not supported, is there any recommended approach to so? Ask Langfuse on Discord suggested to retrieve traces via the API, delete and re-create them with the desired metadata, but it is not very practical for production data.

Thank you!

Langfuse Cloud or Self-Hosted?

Langfuse Cloud

If self-hosted, what version are you running?

No response

SDK and integration versions

No response

Pre-Submission Checklist

• I have checked for existing issues/discussions and consulted Langfuse AI.

[inkeep-ai-assistant[bot]]

Based on the available documentation, Langfuse does not support removing or nullifying specific fields from existing traces(1). The data model is designed around trace immutability, and the deletion capabilities are focused on removing entire traces rather than modifying individual fields(1).

Available Options

1. Delete and Recreate Traces

The approach suggested by Langfuse on Discord—retrieving traces via the API, deleting them, and recreating them with the desired metadata—is the only documented method for this use case(1). While not ideal for production data at scale, this is the supported workflow.

You can delete traces through:

• Single trace deletion: DELETE /api/public/traces/{traceId}(1)
• Batch deletion: DELETE /api/public/traces(1)
• Query-based deletion: Filter traces in the UI and delete all matching results(1)

2. Data Retention Policies

If your goal is to minimize PII exposure over time, you can configure Data Retention policies to automatically remove older traces(1). This won't address existing data immediately but can prevent long-term storage of sensitive fields.

3. Preventive Measures

For future traces, consider implementing data masking at ingestion time to redact PII before it reaches Langfuse(2). This prevents the issue from occurring in new traces.

Important Limitations

Trace deletions in Langfuse are rate-limited and typically complete within 15 minutes, with no confirmation provided(1). You'll need to query the data again to verify deletion(1).

For large-scale field removal from historical traces, the delete-and-recreate approach remains the only documented solution, though it requires careful orchestration for production environments.

📚 Sources:

1. Data Deletion
2. LLM Security & Guardrails

---

Have another question? Just tag @inkeep.

[Lotte-Verheyden]

Hi @justine-arjo, Inkeep's answer is correct. How many traces do you need to remove? If it's a lot, we might be able to help you on that. For future traces, best to look into data masking.