• Phison
  • TrueNAS
  • Wren AI

Phison × TrueNAS × Wren AI

Storage, AI compute, and GenBI in one private stack.

TrueNAS stores and protects the data. The Phison AI Data Platform runs the models next to it. Wren AI turns both into governed answers for your people and your AI agents.

  • Fully air-gapped
  • No cloud LLM
  • No data migration
Private stack online
Your network
  1. People and AI agents ask

    Your team
    • Slack
    • Microsoft Teams
    • MCP agents
  2. Governed GenBI agent and context layer

    Wren AI
    • Context model (MDL)
    • RLS and CLS
    • Validated SQL
    • Audit logs
  3. Phison AI Data Platform

    Phison
    • Dedicated local models
    • Pascari aiDAPTIV™
    • aiDAPTIV Cache Memory
  4. Enterprise storage and data protection

    TrueNAS
    • OpenZFS
    • Snapshots
    • Replication
Air-gapped
No metered APIs
Zero data egress
Bytes sent outside your network
0
Per-token API fees
$0
Data sources, queried in place
20+
Network boundary for all three layers
1

Who brings what

Three layers. Each one owned by a specialist.

Buying storage, AI servers, and analytics software separately leaves you to integrate them. Here each layer has one owner, and the three are built to work together.

TrueNAS

Storage layer

Enterprise storage and data protection.

The data your databases read and everything the agent writes live on OpenZFS storage you operate.

  • End-to-end checksums
  • Snapshots and rollback
  • Replication to a second system
Phison

AI compute layer

Local models, without a GPU cluster.

The Phison AI Data Platform serves dedicated AI models inside your network. Pascari aiDAPTIV™ extends GPU memory into aiDAPTIV Cache Memory, so smaller systems run larger models.

  • Dedicated models, pre-deployed
  • No external API to call
  • Fixed cost instead of metered tokens
Wren AI

GenBI layer

Governed answers for people and AI agents.

Wren AI reads your schema, builds the context model, and answers plain-language questions with SQL you can inspect.

  • One context layer for team and agents
  • Row- and column-level security
  • Validated SQL and a full audit trail

One boundary

Every byte has a home inside your walls.

Follow what a single question touches. Each piece is stored, served, or computed by a layer you own, and none of it crosses the network edge.

01Source data
What it holdsThe databases and files Wren AI queries in place.
Where it livesOn TrueNAS storage, checksummed and snapshotted.
02The model
What it holdsDedicated AI models that write and check the SQL.
Where it livesServed locally on the Phison AI Data Platform. No prompt goes to a cloud LLM.
03Context model
What it holdsMetric definitions, joins, and access rules.
Where it livesManaged by Wren AI, stored on TrueNAS, and rolled back from a snapshot if an edit goes wrong.
04Memory, skills, and logs
What it holdsWhat the agent learns, plus user, timestamp, and SQL for every question.
Where it livesWritten to TrueNAS and replicated to a second system for recovery.

Snapshot, replication, and retention schedules are scoped with your storage team during rollout.

From the SC26 announcement

Why this collaboration matters

Phison built its business on NAND flash controllers and storage. The Phison AI Data Platform is our flagship for the next step: bringing that expertise into complete AI systems that enterprises can deploy on their own infrastructure. With Wren AI providing governed analytics and TrueNAS providing enterprise storage and data protection, customers can run models where their data lives and control how people and AI agents use it.
PhisonMichael WuGM & President, Phison US
Organizations are looking for practical ways to bring AI closer to their data while maintaining control, security, and resilience. Our collaboration with Wren AI and Phison combines trusted business intelligence, accelerated AI infrastructure, and enterprise-grade data protection into a solution that helps customers deploy private AI with confidence.
TrueNASJamila GunawardenaDirector of Solutions and Alliances, TrueNAS
AI-native security starts with giving enterprises control over how AI accesses and uses their data. Our vision is one context layer that serves as a single source of truth, from foundation models to privately hosted models. Phison gives customers one platform for running AI on their own infrastructure, and Wren AI gives them one place to define what their data means and who can use it. That lets enterprises change where AI runs while keeping their business definitions and data policies consistent.
Wren AIHoward ChiCEO, Wren AI

Hardware partners

Start from the hardware you have.

Each solution keeps the model, the data, and the answers inside your network. Pick the one that matches what is already in your racks.

Reference architecture

Take the full-stack blueprint into your next architecture review.

A six-page reference architecture for the teams sizing storage, local AI compute, and governed GenBI as one stack.

  • Sovereignty and the open core
  • Full-stack architecture
  • TrueNAS storage foundation
  • aiDAPTIV compute layer
  • 14-day deployment plan

Free PDF, 6 pages, built for infrastructure, data, and security teams

Cover of the Phison, TrueNAS, and Wren AI sovereign AI reference architecture
Full-stack architecture and TrueNAS storage foundation page
Sovereignty and open core page

Full-stack FAQ

The three-layer stack, answered.

TrueNAS provides enterprise storage and data protection. Phison provides the Phison AI Data Platform, which serves dedicated AI models locally. Wren AI provides the governed GenBI agent and context layer that people and AI agents use to ask questions.

No. Storage, model inference, and the GenBI agent all run inside your network. There are no calls to a cloud LLM, and the stack runs fully air-gapped.

Yes. Wren AI queries data in place, so existing TrueNAS storage stays as it is. You can start with Wren AI on TrueNAS and a model endpoint you run, or add the Phison AI Data Platform for the model runtime.

Wren AI connects to 20+ data sources, including Postgres, MySQL, ClickHouse, and DuckDB, and queries them in place with no ETL and no migration.

Request a demo or write to aidp@getwren.ai. We scope storage, AI compute, and GenBI together with your data, infrastructure, and security teams.

Architecture review

Own the storage, the model, and the answers.

We'll scope the full stack with your data, infrastructure, and security teams, and run a proof of concept on your own schema.