LoginSign Up
← Back to all sparks
T

Tigris

DEVOPS
Velocity6.3

Tigris bends S3-compatible storage toward AI dataloaders and agents.

object-storageai-workloadsdataloadersagent-toolingdata-protection
◆Current state
Tigris is positioning S3-compatible object storage specifically for AI workloads. The recent window mixes genuine product releases — a bulk-read bundle API, soft delete, prefix-filtered lifecycle rules — with engineering blog posts showcasing agent tooling built on top of Tigris.
◆Where it's heading
The direction is to become the default storage substrate for AI agents and training pipelines: bundle reads for dataloaders, copy-on-write bucket forks for agent sandboxes, durable streams for reasoning traces, and a provider-agnostic SDK to pull users in from other clouds. Product and developer-marketing reinforce the same AI-storage thesis.
◆Prediction
Expect more AI-dataloader and agent-workflow primitives, plus continued SDK and ecosystem plays to broaden reach beyond raw S3 parity.

◆Recent moves

  1. 4d ago

    I taught a bucket to speak git

    An engineering blog post on objgit, a single-binary git server that stores repos directly in Tigris with no disk or database. A capability demo built on Tigris rather than a product release, but it underscores the 'storage as a primitive' thesis.

    View source ↗
  2. 16d ago

    Tar saved Unix backups in 1979. Now it saves your dataloader.

    ⚡ SPARK

    The bundle API lets clients pull thousands of objects in a single HTTP request as a streaming tar archive, eliminating the one-GET-per-object pattern. This is the clearest product expression of Tigris tuning object storage for AI dataloader workloads.

    View source ↗
  3. 18d ago

    Introducing Soft Delete for Tigris Buckets and Objects

    Soft delete arrives for buckets and objects: enable it once and every delete becomes recoverable for up to 90 days before permanent removal. A standard but meaningful data-protection feature that closes a gap against established object stores.

    View source ↗
  4. 25d ago

    Introducing storagesdk.dev

    storagesdk.dev is a provider-agnostic Node.js storage API spanning S3, R2, Azure, GCS, and Tigris, with snapshots and forks built in. An ecosystem play to lower switching costs into Tigris, even as it abstracts the provider underneath.

    View source ↗
  5. 1mo ago

    Give your agents disposable environments in Go

    A blog post on Kefka, a Go userspace shell sandbox that gives each AI agent a copy-on-write Tigris bucket fork plus tooling via WebAssembly. Ecosystem/demo content rather than a Tigris release, but it reinforces the agent-storage direction.

    View source ↗
  6. 1mo ago

    You wanted more lifecycle rules. They're here.

    Lifecycle rules now support multiple rules per bucket with prefix filters, so transitions and expirations can be mixed across different prefixes. A practical maturation of bucket data management toward parity with larger object stores.

    View source ↗