Architecture
============

OneByZero Neo is designed as a three-layer architecture that cleanly separates capability foundations, agent construction, and governance & control. This separation allows teams to scale AI adoption safely—from experimentation to enterprise-wide deployment—without coupling innovation to risk.

The layers are progressive and composable: each higher layer builds on the guarantees and assets of the layer beneath it.

High-Level Architecture
-----------------------

.. image:: ../images/neo-architecture.jpg
   :alt: OneByZero Neo Architecture
   :align: center
   :width: 70%

Foundry
^^^^^^^

The Foundry is the foundational layer of OneByZero Neo. It provides all shared, reusable assets required to build reliable and compliant AI agents.

Key responsibilities of the Foundry include:

* Centralized datasets and knowledge bases
* Tooling and integrations exposed to agents as skills
* Guardrails for safety, compliance, and policy enforcement
* Evaluation frameworks for offline and simulated testing
* Shared configuration and secrets management

The Foundry is intentionally agent-agnostic. Assets created here can be reused across multiple agents, teams, and environments, ensuring consistency and reducing duplication.

This layer is typically owned by platform teams and AI governance stakeholders.

Studio
^^^^^^

The Studio is where agents are designed, configured, tested, and prepared for deployment. It is the primary workspace for developers, AI engineers, and solution teams.

Key responsibilities of the Studio include:

* Agent creation and lifecycle management
* Defining personas, backstories, and intents
* Attaching knowledge bases, tools, and memory
* Configuring behavior controls and decision logic
* Performing evaluations and comparisons across agent versions

The Studio consumes assets from the Foundry and produces deployable agent artifacts. It supports rapid iteration while maintaining alignment with organizational standards enforced at the Foundry layer.

This layer is optimized for speed, experimentation, and collaboration.

Command Center
^^^^^^^^^^^^^^

The Command Center provides centralized oversight, governance, and operational control across all deployed agents.

Key responsibilities of the Command Center include:

* Monitoring live agent behavior and performance
* Enforcing policies, budgets, and quotas
* Observability across conversations, costs, and latency
* Managing promotions across environments (Dev → Test → Prod)
* Approving or rejecting agent-driven introspection changes
* Auditing usage for compliance and risk management

The Command Center ensures that scale does not come at the cost of control. It enables organizations to treat AI agents as managed digital coworkers, with clear reporting lines, accountability, and governance.

This layer is primarily used by administrators, platform owners, and business leaders.

Layered Design Principles
-------------------------

The three-layer architecture is guided by the following principles:

* **Separation of concerns**: Build, govern, and operate independently
* **Reuse by default**: Foundry assets are shared, not rebuilt
* **Safe velocity**: Studio enables speed without bypassing controls
* **Enterprise readiness**: Command Center ensures visibility and trust

Together, these layers allow OneByZero Neo to support everything from single-agent pilots to large-scale, multi-team AI deployments.