DOCSEARCHbyAlgolia
Submit Docs

Tigris Object Storage

website·Indexed 37 days ago

https://www.tigrisdata.com/docs/

An S3-compatible object storage service featuring zero egress fees, bucket forks, and snapshots. It includes the Tigris Acceleration Gateway (TAG) for local disk caching and specialized tools for AI agents, such as Agent Kit for storage workflows, Agent Shell for persistent filesystems, and integrations with LangGraph, CrewAI, and MCP.

Tokens
165.7K
Snippets
873
Records
1.2K
Agent score
70%

What's inside Tigris

  1. Overview of Tigris Object Storage

    Tigris is a globally distributed, S3-compatible object storage service. It allows users to store and retrieve objects using the S3 API via a single global endpoint: https://t3.storage.dev.

    Key features include:

    • S3 API Compatibility: Supports over 90% of the S3 API and works with any AWS SDK or S3-compatible tool.
    • Zero Egress Fees: No charges for data transfer out, across regions, or to the internet.
    • Automatic Global Distribution: Data is replicated to regions where it is accessed.
    • Bucket Forks: Instant, zero-copy clones of buckets for versioning or isolation.
    • Shadow Buckets: Enables zero-downtime migration from AWS S3, GCS, R2, or other S3 providers.
    • Strong Consistency: Provides strong consistency for all operations; reads after writes are consistent.

  2. Overview of Tigris Acceleration Gateway (TAG)

    Tigris Acceleration Gateway (TAG) is a high-performance, open-source (Apache License 2.0) S3-compatible caching proxy for Tigris object storage. It is deployed between an application and Tigris to transparently cache objects on local NVMe storage, providing sub-millisecond read latency. TAG requires no application code changes; users simply point their S3 client to the TAG endpoint instead of the Tigris endpoint. It is particularly useful for ML model training (caching datasets), inference data caching (model weights), and high-throughput data pipelines.

  3. Overview of Tigris Cloud Disk

    Tigris Cloud Disk presents a Tigris bucket as a Linux block device. It allows users to format the volume with any filesystem (such as ext4, XFS, or btrfs) and run any application, including databases. The volume inherits the durability, replication, and access controls of the underlying Tigris bucket. Capacity is effectively unlimited and sparse, meaning you only pay for the data actually stored. Because the disk's geometry, settings, and contents reside in the bucket, a disk can be unmounted from one machine and mounted on another across different regions without data replication.

  4. Overview of Tigris Presigned URLs

    A presigned URL grants temporary access to a private object without requiring credentials by including a cryptographic signature. They are used to share private files, enable direct browser uploads, or provide time-limited download links.

    Key constraints:

    • Maximum expiration: 90 days.
    • Default expiration: 1 hour (3600 seconds).
    • Supported operations: GET (download) and PUT (upload).

  5. Overview of Tigris Agent Kit

    Agent Kit is a TypeScript library designed to package storage workflows for AI agents on Tigris. It simplifies the use of @tigrisdata/storage and @tigrisdata/iam by providing four primary primitives:

    • Forks: Copy-on-write clones of a bucket for isolated, writable copies of shared datasets.
    • Workspaces: Per-agent buckets for intermediate outputs and session state, featuring optional TTLs and scoped credentials.
    • Checkpoints: Snapshots of a bucket's state that can be restored into a fresh fork.
    • Coordination: Webhooks on bucket events to trigger multi-agent pipelines without polling.

  6. Overview of DuckLake on Tigris

    DuckLake is an open lakehouse format that separates table metadata (stored in a SQL database catalog) from table data (stored as immutable Parquet files in object storage). When paired with Tigris, it provides a globally distributed, S3-compatible storage layer with no egress fees. DuckLake handles concurrent writes via the catalog database (Postgres, MySQL, SQLite, or DuckDB), eliminating the need for coordination services like DynamoDB.

  7. Overview of TigrisFS

    TigrisFS is a FUSE-based filesystem (forked from geesefs) that mounts S3-compatible storage as a local filesystem. It allows AI workloads to access data without being limited by local storage or memory capacity. Unlike some other solutions, TigrisFS does not obfuscate objects in the bucket and does not require a separate metadata service, meaning data remains accessible via both the TigrisFS filesystem and standard S3-compatible APIs/SDKs.

  8. Overview of Tigris Agent Plugin skills

    The tigris-storage plugin provides five specialized skill sets for AI agents:

    • tigris-authentication: CLI installation, OAuth/credential login, and configuration.
    • tigris-buckets: Creation, deletion, CORS, migrations, TTL, snapshots, and forks.
    • tigris-objects: Upload, download, list, move, delete, and presigning.
    • tigris-access-keys: Creation, listing, role assignment, and deletion.
    • tigris-iam: Management of IAM policies, users, invitations, and permissions.

    Additionally, the tigris-storage-agent subagent can handle complex workflows such as S3 migrations, sandbox environment creation, and security audits.

  9. Overview of Tigris MCP Server

    The Tigris MCP Server implements the Model Context Protocol (MCP) on top of Tigris S3-compatible object storage. It enables AI agents and editors to perform bucket management (create, list, delete), object operations (upload, list, download, move, delete), and generate pre-signed URLs for secure temporary access.

  10. Overview of Cognee memory engine for AI agents

    Cognee is a memory engine that builds knowledge graphs from text, files, or URLs, providing entities, relationships, and embeddings for AI agents to search against. When configured with Tigris S3 storage, the entire memory state (raw data, vector indexes, and knowledge graphs) is stored as objects, allowing for snapshots, forks, and rollbacks of an agent's memory state.

  11. Overview of Tigris Object Storage

    Tigris is an S3-compatible, multi-cloud object storage service designed for data-intensive workloads. It features zero egress fees (no charges for regional, cross-region, or internet data transfer) and automatic global distribution via a single global endpoint. Key features include bucket forks (instant zero-copy clones for isolated environments) and shadow buckets (lazy migration from AWS S3, GCS, or R2 with zero downtime).

  12. Manage Tigris partner integrations via Partner Dashboard

    The Partner Dashboard at partners.storage.dev allows partners to manage customer organizations, storage, and access. Key capabilities include:

    • Organizations: Create and manage tenant organizations. Deactivating an organization immediately invalidates all associated access keys.
    • Team Members: Manage provider-level administrators. All invited members have full access to manage organizations, buckets, keys, and billing.
    • Buckets: Create and configure S3-compatible buckets within specific organizations, including region and storage class settings.
    • Access Keys: Generate, rotate, and revoke S3-compatible credentials scoped to an organization or specific buckets.
    • Billing: Track usage (storage, requests, bandwidth) per organization for cost attribution.