Product beta · local-first AI operations

Sovereign AI Platform

A local AI workspace for teams that cannot send sensitive files, prompts, or workflows to cloud AI APIs. It starts with a lab baseline and moves toward an operable Alpha 1 layer: private knowledge, workflow automation, runtime controls, diagnostics, offline updates, and benchmark proof.

Demo proof Talk deployment
Current proof
Lab baseline and documented single-node Docker Compose deployment.
Alpha 1 proof
Sanitized walkthrough of knowledge, workflows, controls, and diagnostics.
Illustrative diagram of a local AI workspace inside a customer-controlled boundary
Illustrative product architecture, not a product screenshot.

What runs locally

The product layer above local models.

This is not local chat in nicer packaging. The work is turning local models into a workspace that a business team can use and an operator can inspect.

01

Local AI workspace

Chat, documents, repeatable workflows, model state, and proof artifacts stay inside the customer boundary.

02

Private knowledge

Ask approved documents without sending sensitive material to cloud AI APIs.

03

Workflow automation

Classify, extract, route, summarize, and generate outputs through inspectable local flows.

04

Model/runtime controls

Import, pin, inspect, and roll back local model/runtime state before teams rely on it.

05

Operator diagnostics

Health, logs, offline update status, and exportable diagnostic bundles for restricted environments.

Workflow path

From approved files to controlled output.

The Alpha 1 path is deliberately narrow: prove the useful workflow, prove the operator controls, then widen only after the evidence is clean.

Illustrative workflow path from approved documents through retrieval, automation, and review
  1. 01 Load approved knowledge

    Use sample, sanitized, or owner-approved documents. No private browser sessions or customer data.

  2. 02 Retrieve and run

    Ask questions, summarize files, extract fields, route cases, and generate outputs inside the boundary.

  3. 03 Review and measure

    Inspect output quality, throughput, model/runtime state, and diagnostics before expanding usage.

Illustrative operator controls for model state, diagnostics, RBAC, audit logs, and offline updates

Operator controls

Built for the people who have to say yes.

Security and IT do not approve a mystery box. Alpha 1 is aimed at the operational surface they need: role-aware access, auditability, runtime state, rollback, diagnostics, and a path for offline updates.

RBAC + audit logsWho can access what, and what happened.
Offline updatesUpdate packages and diagnostics that work in restricted networks.
Benchmark harnessFixed non-sensitive inputs before wider rollout.
Air-gapped deploymentDesigned for customer-controlled infrastructure.

Proof boundary

What is real now, what Alpha 1 must prove, and what waits.

Current proof

Lab baseline

Ollama + OpenWebUI baseline and documented Docker Compose deployment. Useful, but not yet the product proof.

Alpha 1 layer

Product proof

Local AI workspace, Private knowledge, Workflow automation, model/runtime controls, RBAC, audit logs, Operator diagnostics, offline updates, and benchmark harness.

Roadmap

Later surfaces

Agents, voice, observability, extension packaging, richer media workflows, and hardware-aware profiles. Not current proof.

Demo proof pack

See the evidence boundary before you book time.

The demo page explains what can be shown now, what must be captured from a sanitized Alpha 1 flow, and what stays unpublished until there is proof.

Open demo proof Contact