“Body exceeded 1mb limit” error on dataset upload

[kourosh-samia]

We’re encountering the following error when attempting to add a large dataset in Langfuse Enterprise:

Error: Unexpected token 'B' — "Body exceeded 1mb limit"

We are MedStar and use the paid Enterprise version. Is there a configuration setting we can adjust to increase the request/body size limit to resolve this (e.g., an environment variable or server setting)? Guidance for the appropriate value and where to set it (Docker Compose, app config, etc.) would be appreciated.

We added MAX_EVENT_PAYLOAD_SIZE=5242880 in the docker-compose.yml file under langfuse-worker and langfuse-web.

As a feature request, it would be very helpful if limits like this could be managed in the Admin UI rather than only via YAML/environment configuration.

Thank you for your help.

Server = Red Hat 9.6
Python 3.12.7
langfuse 3.117.2

[inkeep-ai-assistant[bot]]

The "Body exceeded 1mb limit" error you're encountering is a known limitation in Langfuse. Based on the available information, there are several important points to address:

Root Cause: The 1MB limit is enforced by a hard-coded body size limit in the Langfuse ingestion API, regardless of environment variables you set (1). The API route has a configuration of "bodyParser: { sizeLimit: '4.5mb' }" for the ingestion endpoint, but dataset item creation has a separate 1MB limit that hasn't been overridden (1)(2).

Environment Variables: The correct environment variables for self-hosted deployments are LANGFUSE_MAX_EVENT_SIZE_BYTES and LANGFUSE_MAX_BATCH_SIZE_BYTES, not MAX_EVENT_PAYLOAD_SIZE (1)(3). However, even setting these variables won't resolve the dataset creation limit issue because the server-side API route enforces its own restrictions (1).

Limitations for Dataset Items: For dataset item creation specifically, the 1MB limit is a default Next.js restriction that hasn't been overridden in the codebase, even though other endpoints support larger payloads (2). This affects the ability to create dataset items with large payloads directly through the API or SDK (3)(2).

Workarounds: For large files, the recommended approach is to use S3 or S3-compatible storage and reference files using LangfuseMedia links rather than embedding raw data directly in dataset items (3). The format is:
@@@langfuseMedia:type={MIME_TYPE}|id={LANGFUSE_MEDIA_ID}|source={SOURCE_TYPE}@@@ (3).

Self-Hosted Solutions: If you're self-hosting, you could potentially modify the API code to increase the limit and redeploy, but this would require code changes rather than just configuration adjustments (1)(2).

For your Enterprise deployment, you may need to contact Langfuse support directly for assistance with increasing these limits, as the current architecture has hard-coded restrictions that can't be resolved through environment variables alone.

📚 Sources:

1. Event size limit error despite increasing LANGFUSE_MAX_EVENT_SIZE_BYTES and LANGFUSE_MAX_BATCH_SIZE_BYTES
2. bug: 413 error on creating a dataset item -> Body exceeded 1mb limit
3. bug: ApiError: status_code: 413, body: Body exceeded 1mb limit

---

Have another question? Just tag @inkeep.

[Steffen911]

@kourosh-samia Could you share which API calls you're making and also corresponding logs from the langfuse-web container? If you're making a UI driven flow, a short recording would be useful too.