About PhoenixAI Anywhere

PhoenixAI Anywhere is the self-managed platform for running PhoenixAI entirely inside your own environment: your own cloud account, your data center, or a fully air-gapped network with no outbound connectivity.

Agentic and AI analytics have to run inside your own boundary, in highly secure and regulated environments, and they require full performance. Anywhere is built for that.

The control plane runs inside your environment too, next to the PhoenixAI database clusters it manages, rather than outside it.

What Anywhere provides

An enterprise core. More capability, more performance and more resilience than the open-source engine: Kerberos, row and column access policies, transparent data encryption, multi-AZ, cache high availability, multi-warehouse, autoscaling, vector and text search, Unity Catalog and Iceberg V3. The full list is below.

The tools to operate it. Self-managing an analytical database means operating it as well as running it. Clusters are declared as Kubernetes resources, so Helm, Argo CD, Flux and Terraform keep working the way your platform team already runs them. The Anywhere Console observes and reports on what those declarations produce: cluster inventory and health, monitoring and query insight, licensing, usage metering, and support bundles.

Where it runs

Two things vary independently.

Where the infrastructure sits. Your own cloud account, or your own data center.

How much outbound connectivity it has. Full, restricted to an allowlist, or none at all.

Nothing in Anywhere requires outbound connectivity to work. Licenses validate offline, usage metering is recorded and kept locally, and a support bundle is a file you generate and transfer deliberately. Most commercial data platforms assume a call home somewhere in that chain, which is what makes them hard to run in a network that has no route out.

Architecture

Component	What it does
Anywhere (control plane)	The Anywhere Console observes and reports: cluster inventory and health, node and pod topology, monitoring, query insight, audit-log search, licensing, usage metering, and support bundles. Underneath it, the Anywhere operator reconciles the PhoenixAI database clusters you declare, and keeps them running as nodes come and go
PhoenixAI database cluster	The database. A planning layer handles SQL, metadata and optimization; an elastic pool of compute nodes does the work
Your lakehouse	Queried in place: Iceberg, Delta Lake, object storage and Kafka

Anywhere is delivered on Kubernetes and installed with Helm. One Anywhere installation manages as many PhoenixAI database clusters as you give it, each with its own warehouses and its own elastic compute.

Table data lives in your object storage, so compute scales without moving it, and a cluster can be rebuilt from a snapshot without restoring the data itself.

What Anywhere adds to open-source StarRocks

The baseline here is open-source StarRocks that you run yourself. Anywhere runs the same analytical engine, so what you already rely on behaves the same way, and adds the enterprise capabilities and the platform above it.

Category	Capability	StarRocks Community	PhoenixAI Anywhere
Core analytical engine	Vectorized engine	✅	✅
	CBO planner	✅	✅
	External table query	✅	✅
	Upsert support	✅	✅
	Iceberg V2	✅	✅
	Multi-table materialized view	✅	✅
	TLS	✅	✅
Security and governance	Kerberos authentication	❌	✅
	Row access and column masking policies	❌	✅
	Transparent data encryption	❌	✅
	Graphic query profiling	❌	✅
Reliability and availability	Multi-AZ availability	❌	✅
	Cache high availability, multi-replica with automatic warmup	❌	✅
Elasticity and performance	Multi-warehouse	❌	✅
	Warehouse autoscaling	❌	✅
	Adaptive data distribution	❌	✅
	Incremental materialized view	❌	✅
	ARM hardware support	❌	✅
Lakehouse	Unity Catalog (Databricks)	❌	✅
	Iceberg V3	❌	✅
AI-native	Vector search	❌	✅
	Text search (BM25)	❌	✅
	AI Semantic SQL (Public Preview)	❌	✅
	AI ContextBase (Public Preview)	❌	✅
	AI Observer (Public Preview)	❌	✅

The Overview lists the entry points into the rest of the documentation.