AI workloads, data platforms, and infrastructure notes, written from the engineering edge between benchmarks and production.

RSS feed
/ /

A New Front Door to the VAST AI Operating System

The redesigned docs.vastdata.com mirrors the VAST AI OS—role-based paths, a version selector, product-area navigation, and documentation built for people and AI.

I

Itzik — VP Mission Alignment, VAST Data

·

·

8 min read


Introducing the redesigned docs.vastdata.com: new menus, a version selector, role-based paths, and documentation built for both people and AI

Figure 1: the new docs.vastdata.com homepage

Figure 1. The new docs.vastdata.com homepage.

VAST is no longer just a storage system. It is a database, storage, and compute system on a single platform, governed by a single control plane: the VAST AI Operating System (VAST AI OS documentation). Just as a traditional OS has a file system, kernel, scheduler, and runtime, the components of the VAST AI OS map to those same roles for AI and data workloads.

As the platform grew, our documentation had to grow with it. Admins, SREs, data engineers, ML engineers, and application developers all come to the docs with very different goals, and a single, storage-centric table of contents no longer served them well. So we rebuilt the experience from the front door in.

Today we are excited to introduce the new docs.vastdata.com. Here is what’s new.

Documentation that mirrors the AI OS

The new VAST AI OS Documentation start page presents the platform the way it is actually built: as one operating system with distinct, cooperating services. It gives you a single map of the stack, from DataStore and DataBase at the foundation, through DataSpace across sites and clouds, up to DataEngine and its SyncEngine, InsightEngine, and AgentEngine capabilities, with a link to the VAST Glossary for any term you’re unsure of.

Figure 2: VAST AI OS Documentation start page mapping platform components

Figure 2. The VAST AI OS Documentation start page maps every component of the platform.

On the homepage, the Product Areas grid carries the same structure. Each card names a part of the AI OS, tags what it covers, and explains in one line what it’s for:

VAST Platform: cluster, hardware, network, and VMS. Install, run, and support a VAST cluster.

• VAST DataStore: file, object, and block. NFS, SMB, S3, and block on one namespace.

• VAST DataSpace: global namespace and multi-cluster management across sites and clouds.

• VAST DataBase: tabular, transactional, and vector data.

• VAST DataEngine: functions, triggers, and pipelines. Serverless compute next to the data.

• VAST Event Broker: Kafka-compatible event streaming for applications and data pipelines.

• VAST Catalog: queryable metadata through Trino, Python, or Jupyter.

• VAST SyncEngine: onboard and migrate data, with its own release cadence.

Figure 3: Product Areas view of the full VAST AI OS on one screen

Figure 3. Product Areas: the whole AI OS on one screen.

The database, and the engine that runs on it

Two parts of that map deserve a closer look, because they are where most new work on the AI OS actually happens: the VAST DataBase and the VAST DataEngine.

The VAST DataBase is a transactional and analytical database on the same data. Newly written rows land instantly in persistent memory, then convert into optimized columnar chunks for analytics, so dashboards and applications read the current data rather than a copy that was exported overnight. It is fully ACID compliant, and it maintains itself, vacuuming, or compaction.

It is also the vector store. Embeddings live next to the structured data they describe, so an AI pipeline does not need a separate vector database, a second copy of the data, or a second set of access controls. The VAST Native Query Engine runs vector similarity search, SQL filters, joins, and aggregations in a single execution path, in place, on the cluster. A hybrid query can ask for nearest neighbors and constrain them with ordinary SQL in one request, without moving the dataset to an external engine.

Figure 4: VAST DataBase page with Native SQL and Vector Query Engine

Figure 4. The VAST DataBase page, open on the Native SQL + Vector Query Engine.

Around that database sits the VAST DataEngine, the programmable compute fabric. It runs lightweight Python functions directly on the cluster’s compute nodes. The VAST Event Broker, which is Kafka-compatible, detects file and object events as they happen and triggers those functions, so enrichment, transformation, and routing happen next to the data instead of in an external ETL job.

The same fabric runs Trino and Spark as managed services on the cluster, with direct access to the data, so teams keep the query engines they already use without standing up a separate cluster or copying data into it. Pipelines can be built in Python or in the low-code DataEngine interface, and triggers, functions, and job metrics are managed from one control plane. On the docs page, DataEngine is shown alongside the capabilities it orchestrates: SyncEngine for data movement, InsightEngine for real-time RAG and embeddings, and AgentEngine for agents, tools, and observability.

Figure 5: VAST DataEngine with serverless functions, triggers, and managed Trino and Spark

Figure 5. The VAST DataEngine: serverless functions, event triggers, and managed Trino and Spark.

The new docs treat these as first-class product areas rather than chapters buried inside a storage guide. The VAST DataBase and The VAST DataEngine each have their own page, and the administration, user, and Runtime SDK guides sit beside them in the sidebar.

A version selector, right where you need it

When you run infrastructure, the right answer depends on your release. Which version am I reading, and how do I switch? should never be a hard question. The new header shows a version selector on every page. It shows your current context (for example, 5.5 Docs) and lets you jump in one click to:

• Release documentation: 5.5 Docs, 5.4 Docs, and 5.3 Docs.
• API documentation: the 5.5 API Documentation.
• Cross-release content: KBs & Tutorials, Develop & Integrate, and Reference.

If you’re running 5.4 in production while evaluating 5.5, you no longer have to guess whether a procedure applies to your cluster. Pick your release and stay in it.

Figure 6: docs header version selector for 5.5, 5.4, 5.3, and API documentation

Figure 6. The version selector in the header switches between 5.5, 5.4, and 5.3 docs, the API docs, and cross-release content.

New menus that follow your workflow

The top navigation is now organized around what you’re trying to do: Getting Started, Guides, KBs/Tutorials, Develop, Connect & Integrate, and References. The Guides menu, for example, puts every supported release’s documentation and release notes side by side, alongside SyncEngine, VAST NFS, and a clearly labeled Legacy Documentation area for older content.

Figure 7: Guides menu listing documentation and release notes by release

Figure 7. The Guides menu lists docs and release notes for each supported release.

New to VAST? The Getting Started menu opens with the VAST AI OS Documentation, a Topic Hub, and the VAST Customer Quick Start Guide, with links out to VAST University and VAST Cosmos.

Inside each guide, a filterable sidebar, breadcrumbs, an in-page table of contents, and an article summary help you find the right section quickly. And from any page, Ctrl + K (or CMD +K on Mac) opens search across the whole site, which is also the fastest way to find an article you bookmarked on the old site. Old kb.vastdata.com links redirect automatically to docs.vastdata.com, where you can search for an article by its previous name.

Figure 8: VAST 5.5 guide page with new header, sidebar, and quick links

Figure 8. A 5.5 guide page with the new header, sidebar, and quick links.

Start from your role

The AI OS serves very different people, so the homepage now asks a simple question: “I am a…” Choose your role and land on a curated persona page:

• Storage administrator: runs the cluster day to day.
• Platform engineer / SRE: automates, integrates, and monitors.
• Data engineer: builds pipelines on DataBase, Event Broker, and DataEngine.
• Data scientist / ML engineer: queries, trains, and serves against VAST data.
• Application developer: talks to VAST through APIs, SDKs, and protocols.
• Field / solutions engineer: designs and deploys for customers.
Figure 9: Role-based entry points on the docs homepage

Figure 9. Role-based entry points on the homepage.

Each persona page explains how that role uses the VAST AI OS and links straight to its most common tasks. The Data Scientist / ML Engineer page, for instance, covers NFS performance features for training, S3 client setup, querying the VAST Catalog, vector search and vector indexes in the VAST DataBase, and recovering datasets or checkpoints from snapshots.

Figure 10: Data Scientist and ML Engineer persona page with common tasks

Figure 10. The Data Scientist / ML Engineer persona page with common tasks.

Quick links to what you use most

The pages people visit most are now one click from the homepage: the Latest Admin Guide, Latest APIs, CSP Onboarding Guide, Developer Tools, Solutions & Integration, DataEngine, SyncEngine, Virtualization, Drivers, and a dedicated, sign-in-protected VAST Partners area.

A home for developers

Developers now get a section of their own. The Develop menu brings together the Developer’s Guide and dedicated pages for the DataBase, Vector Search, Event Broker, DataEngine, DataStore, VAST Catalog, and SyncEngine, plus a link to VAST Orbit on GitHub. Content is organized by product, role, or software release, with expanded resources for building applications, integrations, and data pipelines on VAST.

Next to it, Connect & Integrate covers how VAST plugs into the rest of your stack: AI Inference, VAST Polaris, VAST on Cloud, Solutions and Integrations, the VAST Admin MCP, Virtualization, VAST Drivers, and VAST UPlink.

Built for people and for AI

An AI operating system deserves AI-ready documentation. More and more often, the first reader of a docs page is an AI assistant working for an engineer. The new site publishes an llms.txt index, a machine-readable map of every page. It lets LLMs, coding assistants, and agents find current, authoritative VAST content instead of guessing.

Tell us what works, right on the page

Every article now has feedback buttons at the top. If a procedure worked, or if something is missing or unclear, you can tell the docs team in one click without leaving the page.

What’s next

This redesign is a foundation, not a finish line. more persona tutorials, expanded API examples, deeper integration guides, etc, are coming

Explore the new site at docs.vastdata.com and tell us what you think. If something is hard to find, we want to know.

Related on Lots of Data: VAST AI OS API Documentation Hub · Agent Sandboxes and Storage · Anthropic Scion and AI Infrastructure