Export via Blob Storage Integration

Where is this feature available?

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

You can create schedule exports to a Blob Storage, e.g. S3, GCS, or Azure Blob Storage, for traces, observations, enriched observations, and scores.

Those exports can run on an hourly, daily, or weekly schedule. Navigate to your project settings and select Integrations > Blob Storage to set up a new export. Select whether you want to use S3, a S3 compatible storage, Google Cloud Storage, or Azure Blob Storage.

Start exporting via Blob Storage

To set up the export navigate to Your Project > Settings > Integrations > Blob Storage.

Fill in the settings to authenticate with your vendor, enable the integration, and press save. Within an hour an initial export should start and continue based on the schedule you have selected. The export supports CSV, JSON, and JSONL file formats. Read our blob storage documentation for more information on how to get credentials for your specific vendor.

Export source (Fast Preview)

Blob Storage integrations now include an Export Source selector. New integrations default to Enriched observations (recommended) (trace attributes are directly set on observations).

This source uses enriched observations with trace attributes and provides significantly better export performance. Scores are always included, regardless of the selected source.

Available options:

Traces and observations (legacy)
Traces and observations (legacy) and enriched observations
Enriched observations (recommended)

Traces and observations (legacy) sources may be deprecated in the future. All new export jobs should use Enriched observations (recommended), and existing legacy jobs are strongly recommended to upgrade.

Cloud projects created on or after 2026-05-20 will not see the Export Source selector — new Cloud projects export as Enriched observations (recommended) automatically. The REST API rejects legacy values for these projects with 400 BAD_REQUEST. Existing projects and all self-hosted deployments are unaffected.

Upgrade path for existing configurations

This migration path applies to pre-cutoff Cloud projects and self-hosted deployments. Post-cutoff Cloud projects already use Enriched observations (recommended) and cannot select legacy sources.

Existing integrations continue to use Traces and observations (legacy) until changed.

To migrate safely:

Switch to Traces and observations (legacy) and enriched observations.
Validate downstream jobs and data consumers while both sources are exported (this mode creates duplicate records by design).
Switch to Enriched observations (recommended) once validation is complete.

For rollout details, see the Simplify for Scale changelog.

Exported fields

For a complete reference of all fields included in each export file (traces, observations, enriched observations, and scores), see the Export Field Reference.

Choose which columns are exported

For the Enriched observations (recommended) and Traces and observations (legacy) and enriched observations sources, you can choose which column groups appear in each row of the enriched observations export. Eleven groups cover the full row, ten of which are toggleable; toggle them in Project Settings → Integrations → Blob Storage under Export Field Groups.

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

Per-unit pricing fields (input_price, output_price, total_price) live in the model group — they come from the matched model definition. Deselecting model skips the worker-side model pricing lookup entirely. The usage_pricing_tier_id and usage_pricing_tier_name fields stay in the usage group.

New integrations default to all eleven groups, so behavior matches earlier exports unless you narrow the selection. The Traces and observations (legacy) source uses a fixed column set and ignores field groups.

Configure via REST API

GET and PUT /api/public/integrations/blob-storage accept and return:

exportSource — LEGACY_TRACES_OBSERVATIONS, OBSERVATIONS_V2, or LEGACY_TRACES_AND_ENRICHED_OBSERVATIONS.
exportFieldGroups — a list of group names. Must include core when provided. Must be omitted or null for LEGACY_TRACES_OBSERVATIONS (the REST contract returns null on read and rejects non-null on write for that source). When omitted on update, the existing value is preserved.
compressed — boolean; defaults to true for new integrations. When true, files are written as .csv.gz, .json.gz, or .jsonl.gz.

See the API reference for the full schema.

Alternatives

You can also export data via:

UI - Manual batch-exports from the Langfuse UI
SDKs/API - Programmatic access using Langfuse SDKs or API

Was this page helpful?

On this page