Export to blob storage

Where is this feature available?

Hobby
Not Available
Core
Not Available
Pro
Teams Add-on required
Enterprise
Available
Self Hosted
Available

Overview

Schedule exports to Amazon S3, S3-compatible storage, Google Cloud Storage, or Azure Blob Storage every 20 minutes, hourly, daily, or weekly.

Configure the export

Create the integration

Open Project Settings > Integrations > Blob Storage.
Select your provider and enter its bucket, path, and credentials.
Choose the file format, schedule, history, and field groups.
Enable the integration and save. The first run starts shortly afterward.

Setting	Choices and behavior
Provider and access	Amazon S3, S3-compatible storage, Google Cloud Storage, or Azure Blob Storage.
Format	Parquet (default), CSV, JSON, or JSONL. Text formats can be gzip-compressed; Parquet uses its own encoding and compression.
Schedule	Every 20 minutes, hourly, daily, or weekly. Each run exports one time window.
History	Full history starts at the project's earliest data; From setup date starts when the integration is enabled; From custom date starts at a selected date. Changing this setting resets the sync position and re-scans from that point.

Parquet notes

Observation files omit input_price, output_price, and total_price. Use cost_details and total_cost; trace and score fields are unaffected.
On self-hosted ClickHouse older than 25.11, upgrade ClickHouse or use CSV, JSON, or JSONL because incomplete Parquet output may not be reported.

Use Validate before saving to check that Langfuse can access the target with the supplied credentials. The integration then writes all output beneath the configured prefix without changing other objects in the bucket.

Langfuse writes exports to your bucket using this structure:

{prefix}{project-id}/
├── observations_v2/
│   └── {timestamp}.{parquet|json|jsonl|csv}[.gz]
├── scores/
│   └── {timestamp}.{parquet|json|jsonl|csv}[.gz]
└── manifests/
    └── {timestamp}.json

Field selection

Observation columns are configurable; score columns are fixed.

New integrations select all eleven field groups. core is required; disable other groups to reduce file size or omit sensitive data.

Group	Fields
`core`	`end_time`, `id`, `parent_observation_id`, `project_id`, `start_time`, `trace_id`, `type`
`basic`	`bookmarked`, `environment`, `is_root_observation`, `level`, `name`, `public`, `session_id`, `status_message`, `user_id`, `version`
`time`	`completion_start_time`, `created_at`, `updated_at`
`io`	`input`, `output`
`metadata`	`metadata`
`model`	`input_price`, `model_id`, `model_parameters`, `output_price`, `provided_model_name`, `total_price`
`usage`	`cost_details`, `total_cost`, `usage_details`, `usage_pricing_tier_id`, `usage_pricing_tier_name`
`prompt`	`prompt_id`, `prompt_name`, `prompt_version`
`metrics`	`latency`, `time_to_first_token`
`trace_context`	`release`, `tags`, `trace_name`
`tools`	`tool_call_names`, `tool_calls`, `tool_definitions`

Changes apply only to future exports. See the observation field reference for types and descriptions.

Scores are always exported and their fields cannot be selected.

Category	Fields
Identity	`id`, `project_id`, `timestamp`, `created_at`, `updated_at`
Context and links	`trace_id`, `observation_id`, `session_id`, `dataset_run_id`, `environment`
Score	`name`, `data_type`, `value`, `string_value`, `source`, `comment`

See the score field reference for types and descriptions.

Configure through the API

Use the public endpoint to read or update the integration:

GET /api/public/integrations/blob-storage
PUT /api/public/integrations/blob-storage

On PUT, exportFieldGroups must include core; omit the property to keep the current selection. compressed applies only to CSV, JSON, and JSONL.

See the REST API reference for authentication, provider settings, and the complete schema.

Process exports

Use each run's manifest as the unit of work.

Know when an export is complete

A run is complete only when its manifest exists at {prefix}{project-id}/manifests/{timestamp}.json. Langfuse writes it after every data file uploads successfully.

If an upload fails, no manifest is written. The run is retried and any data objects under the same keys are overwritten.

When a manifest appears:

Read files[].
Fetch every full object key in files[].key. Do not derive filenames or list table directories.
Deduplicate records by ID because adjacent export windows share an inclusive boundary.

Field	Purpose
`window`	Inclusive `minTimestamp` and `maxTimestamp` for the run
`exportSource`, `tables`	Export source and tables included
`files[]`	Object key, table, format, compression, uploaded size, and row count for each data file

For Parquet, files[].rowCount is null; inspect the file metadata when you need a count. Ignore unknown manifest fields so additive changes do not break your parser.

Trigger your pipeline

Subscribe to object-created events filtered to {prefix}{project-id}/manifests/, then fetch the manifest and process each files[].key.

Amazon S3: S3 Event Notifications or EventBridge
Google Cloud Storage: Pub/Sub notifications for OBJECT_FINALIZE
Azure Blob Storage: Event Grid for Microsoft.Storage.BlobCreated
MinIO: bucket notifications
Backblaze B2: event notifications

If your S3-compatible provider has no object-created events, poll the manifest prefix and checkpoint the latest processed key. Manifest names sort by export timestamp.

Make manifest processing idempotent. Provider events can be delivered more than once, and a catch-up can create several manifests close together.

Troubleshooting

Export timing, backfills, and retries — Understand delays, catch-up bursts, failed runs, and retry behavior.
Empty export files — Learn why files can be empty and how pipelines should handle them.
Configuration changes, errors, and re-exporting — Validate changes, clear sticky errors, or export historic data again.

Upgrade a legacy export

Complete this upgrade if your integration uses Traces and observations (legacy) or your bucket receives traces/ and observations/ files. On Langfuse Cloud, remaining legacy exports switch at the v4 cutover. On self-hosted v4, the legacy source stops producing data after the server switches to events_only.

Switch to enriched observations

Most integrations can switch directly once their consumer supports the enriched layout:

On self-hosted Langfuse v4, first switch the server to dual write mode with the Langfuse v4 preview enabled and complete the historic data migration.

Enriched export sources are unavailable in legacy mode or when the Langfuse v4 preview is disabled. Switch the integration before the server's events_only cutover.

Prepare the new consumer. Process observations_v2/ and scores/ from each manifest's files[].key. Use the enriched field reference.
Switch the export. Select Enriched observations (recommended).
Verify and retire legacy. Process the next completed export with the new consumer. After it succeeds, retire the consumer that reads traces/ and observations/.

Optional: keep both layouts during the cutover.

If your pipeline cannot tolerate a cutover gap, temporarily select Traces and observations (legacy) and enriched observations after step 1.

This writes observations/ and observations_v2/ for subsequent windows only; it does not backfill earlier exports. Run both consumers until the new one is validated, then continue with step 2.

Do not load both observation layouts into the same production dataset. Check the export compatibility matrix if this option is unavailable.

Cutover checks

All deployments: traces/ and observations/ stop receiving files after the switch, including empty files. Use manifests/ as the run-completion signal.
Langfuse Cloud: Ignore DEPRECATION_NOTICE.txt when processing exports; it is not part of a run.
Self-hosted Langfuse v4: Optional dual-layout validation is available only during dual write mode with the Langfuse v4 preview enabled.

Was this page helpful?

On this page