Versioned, content-addressed storage for everything the platform needs.
Models, prompts, datasets, configurations, evidence: stored with cryptographic integrity, versioned by content hash, retained per customer policy, deployed inside the customer perimeter.
// storage · for everything the platform needs · inside the perimeter
Storage that the audit can trust.
An AI platform that cannot identify exactly which model produced which output, against which data, under which configuration, is not auditable in any useful sense. Storage is the foundation that makes reproducibility possible.
Limit Store stores models, prompts, datasets, configurations, and evidence with content-addressed integrity. Every artifact has a cryptographic identity. Every version is preserved per policy. Every read and write is logged. Recall and reconstruction are routine operations, not forensic exercises.
Content-addressed identity. Reproducible retrieval.
- 01
Content-addressed artifact storage
Models, prompts, datasets, and configurations are identified by content hash. Tampering is detectable; versions cannot be silently replaced.
- 02
Versioning, lineage, and provenance for every artifact
Every artifact carries its full version history and provenance. Reproducing a past inference means recalling the exact model, prompt, configuration, and (where reproducible) the exact input that produced it.
- 03
Customer-controlled encryption at rest
Storage encryption uses keys issued by customer infrastructure. Limit Systems does not hold the keys to decrypt customer data.
- 04
Retention and lifecycle aligned to customer policy
Retention rules are configured per artifact class to match legal, regulatory, and operational horizons. Lifecycle actions are themselves audited.
Storage lives where the data lives.
Storage deploys inside the customer perimeter on customer-procured infrastructure. Object backends are pluggable: customer S3-compatible storage, on-premise object stores, or filesystems. Encryption keys originate in customer key management; the platform consumes them.
Cross-site replication, where used, is customer-controlled and customer-bounded. Backup and recovery flows are integrated with the customer's existing backup infrastructure rather than parallel to it.
- Backends
- S3-compatible object stores, on-premise object storage, filesystems. Customer choice.
- Encryption
- Keys from customer KMS or HSM. Limit Systems does not hold decrypt keys.
- Replication
- Customer-controlled cross-site replication for multi-site deployments.
- Backup
- Integrates with customer backup infrastructure rather than parallel to it.
This capability is one layer of the operating system. Every application uses it. Every customer perimeter that runs Limit Platform gets it. See the full architecture →
See it in your environment.
Walk us through the systems already in place. We'll show you how this layer fits.