About Inventory Reporting for CoreWeave AI Object Storage

Understanding inventory reports for CoreWeave AI Object Storage buckets and objects

Inventory Reporting provides a scheduled, point-in-time listing of all objects in a bucket. The output includes lists of objects with metadata for each object. You specify the source bucket for which objects will be inventoried and a destination bucket for the output. You can schedule the inventory listing to run daily or weekly, and you can configure the output to meet your analytics needs.

This feature is useful for storage, data platform, security, and compliance teams who need authoritative object-level inventory for governance, analytics, lifecycle, and cost optimization.

Configuration options

For your inventory reporting configuration, you can specify the following options:

• Schedule: How often you want the inventory report to run, either daily or weekly.
• Source Bucket: Choose the bucket that you want to inventory. You can choose to list all of the objects in the bucket or only objects with a specified prefix. If you don't specify a prefix, all objects will be inventoried.
• Destination Bucket: Choose the bucket where you want to write the inventory reports, and optionally specify a prefix (subdirectory) to organize the reports.
• Output formats:
  • JSON, CSV, or TSV files compressed with GZIP.
  • Parquet files compressed with Snappy.
  • Optimized row columnar (ORC) files compressed with ZLIB.
• Optional fields: Choose the optional fields that you want to include in the inventory report. See the detailed field reference for all available configuration options.

To configure and enable inventory reports, see Configure Inventory Reporting for CoreWeave AI Object Storage.

Output fields

When you configure your inventory report, you can choose whether to include or exclude certain fields from the output.

Some fields are always included in the report, and the optional fields are only included if you specify them in the OptionalFields configuration. The following fields are available:

Field Name	Description	When included in report
BucketName	Name of the bucket that contains the object	Always
ObjectKey	Unique key (path/name) identifying the object within the bucket	Always
VersionID	Version identifier for the object, if versioning is enabled	Always included if IncludedObjectVersions is set to All
IsLatest	Indicates if this is the latest version of the object	Always included if IncludedObjectVersions is set to All
IsDeleteMarker	Shows whether this entry is a delete marker	Always included if IncludedObjectVersions is set to All
Size	Size of the object in bytes	Optional
LastModifiedDate	Date and time when the object was last modified	Optional
IsMultipartUploaded	Indicates if the object was uploaded as a multipart upload	Optional
EncryptionStatus	Encryption status or method used on the object	Optional
StorageClass	Storage class assigned to the object (see Usage-Based Billing)	Optional
ChecksumAlgorithm	Algorithm used to compute the checksum of the object	Optional
LastAccessedDate	Timestamp of the last access (read) of the object	Optional

Notes and limitations

• No notifications: You won't receive notifications when the inventory listing has completed, so you must check the destination bucket to see if the inventory reports have been written.
• Pricing: Using Inventory Reporting incurs charges for storage of the inventory reports in your destination bucket. Use LastAccessedDate to derive the access tier (Hot, Warm, or Cold) of the objects.
• Consistency: Inventory reports provide eventual consistency. Object changes during the scan may not be reflected in that specific report, but changes will converge across successive runs. Reports are not transactional snapshots.
• Not for billing: Inventory reports are for governance and analytics, not billing reconciliation. Inventory reports will not match exactly the audit logs or the billing data.
• Configuration limit: You can set up to 1,000 inventory configurations per bucket.
• Encryption: SSE-C for report files is not supported.

Understanding the inventory report

Inventory reports consist of:

• A manifest file (manifest.json) that describes the report and lists all inventory data files
• One or more inventory data files containing the actual object listings

Manifest file

The inventory report contains a manifest file (manifest.json) that contains information about the inventory report, and it is written to the root of the subdirectory specified in the inventory configuration. The manifest file is a JSON object with the following fields:

Field	Description
fileSchema	Specifies which fields are included in the report
sourceBucket	Name of the source bucket
destinationBucket	Name of the destination bucket
version	Version of the inventory report
creationTimestamp	Creation date (start of inventory report generation)
files	List of the inventory files
files.key	Key identifying the file path, always included in the report
files.size	Size of the inventory file in bytes
files.md5checksum	MD5 checksum of the inventory file

The manifest file looks similar to this:

manifest.json

{
  "sourceBucket": "my-bucket",
  "destinationBucket": "arn:aws:s3:::my-destination-bucket",
  "version": "2016-11-30",
  "creationTimestamp": "1761777847000",
  "fileFormat": "JSON",
  "fileSchema": "Bucket,Key,VersionId,IsLatest,IsDeleteMarker,Size,StorageClass,LastAccessedDate",
  "files": [
    {
      "key": "inventory-reports/my-bucket/my-inventory-config/data/2025-10-29_2ead449d-ef16-5bb2-beb9-f3421d5c8af1.json.gz",
      "size": 183,
      "md5checksum": "6f7cbc70f083910058ac7d57615a304b"
    }
    ...
  ]
}

Support

If you need help with Inventory Reporting, contact CoreWeave support or email [email protected].