AVICENNA

AVICENNA is the Governance Architect built into VB-OS Cloud. It helps you configure governance environments through conversation: describe what you need to govern, review the proposed configuration, and authorize creation. AVICENNA proposes; you authorize; VB-OS evaluates deterministically.

Open AVICENNA from any project using the Architect tab in the project sub-navigation, or toggle the panel with Cmd+Shift+.

How AVICENNA Works

AVICENNA operates as a proposer, not a decision authority. The relationship is:

You describe what you want to govern
AVICENNA proposes a governance configuration
You review and authorize
VB-OS deterministically evaluates using that configuration

AVICENNA cannot create, modify, or delete any governance artifact without your explicit authorization. Every change goes through a confirmation gate before execution.

Getting Started

When you first open AVICENNA, the welcome screen offers three entry points:

Start with a problem: describe what you want to govern in your own words (e.g., “I need to make sure every vendor payment is properly authorized before it goes through”)
Start with something you have: bring an existing policy, checklist, procedure, or system (e.g., “We have a compliance checklist that requires three approvals for any transaction over $10,000”)
Build directly: specify exact governance artifacts such as boundaries, connectors, flows, and evidence mappings

Below the entry points, AVICENNA lists governance categories to help you get started:

Category	Example
A decision	“May this candidate proceed to the next stage?”
An action	“Can this agent deploy to production?”
An AI / agent	“What is this autonomous agent allowed to do?”
A workflow	“What gates must this process pass through?”
A professional activity	“Govern how AI assists in my practice.”
A system / API	“Govern what this service can do without approval.”
A document / record	“Do not release until these conditions hold.”

Select a category or type your own description, then press Cmd+Enter to send.

Conversation Flow

After you send your first message, AVICENNA begins a structured conversation:

Understanding: AVICENNA asks clarifying questions about your governance requirements
Policy Intent: once it understands what you need, AVICENNA presents a Policy Intent summarizing the rules and evidence requirements it has identified. Each intent includes a hash for traceability. Review it and click Confirm Intent if it matches your requirements.
Governance Plan: AVICENNA produces a detailed plan listing the specific steps it will take (create boundaries, configure connectors, set up flows, map evidence sources). Each step shows a description and status. The plan includes a plan hash and lists all referenced artifacts.
Authorization: you review the plan and click Create This Setup at the confirmation gate to authorize execution. The confirmation gate shows the step count and artifact count.
Execution: AVICENNA creates the governance artifacts through the platform’s standard APIs. Each step’s status updates in real time (pending, success, failed, skipped).

You can send follow-up messages at any point to refine requirements or ask questions.

What You See During a Conversation

Mode Indicator

The panel header shows the current conversation mode alongside the project name.

Policy Intent

The Policy Intent view displays:

Rules: the governance rules AVICENNA has identified from your description
Evidence Requirements: the evidence fields that the boundary will require
Intent Hash: a truncated hash for traceability

Review these carefully. Click Confirm Intent to proceed, or send a message to adjust.

Governance Plan

The Governance Plan view displays:

Title and description of the overall plan
Steps table: numbered steps with descriptions, status (pending/success/failed/skipped), and expandable details
DSL source: expand any step to see the VBL source code for boundary definitions
Evidence sources: expand to see where each evidence field comes from (connector, manual, API, derived) and which provider supplies it
Referenced artifacts: boundaries, connectors, and flows that the plan will create or modify, identified by type, name, and reference

Plan Execution

When you authorize a plan, each step executes in sequence. The execution progress view shows each step’s current status. If a step fails, the error message is displayed and subsequent dependent steps may be skipped.

Test Scenarios

AVICENNA may simulate your governance setup against test scenarios before asking you to authorize. The simulation view shows:

Scenario name: what the test represents
Evidence: the test evidence payload
Expected result: what outcome the boundary should produce
Actual result: what outcome it did produce
Status: pass or fail

Impact Analysis

For changes that affect existing governance artifacts, AVICENNA shows an impact analysis with:

Blast radius: none, low, medium, high, or critical
Affected boundaries: existing boundaries that the plan modifies
Affected flows: flows that reference affected boundaries
Affected deployments: active deployments of affected boundaries

Artifact Preview

When a plan includes boundary definitions, AVICENNA shows a preview of the boundary DSL source code so you can inspect the exact rules before authorizing.

Tool Calls

AVICENNA uses internal tools to read your project’s current state, look up existing boundaries, and validate configurations. Tool calls appear as expandable blocks in the conversation showing the tool name, input, and output. These are informational and do not require your action.

Contextual Actions

On domain pages throughout the console (boundary detail, flow builder, connector detail, evaluation detail, certification screens), you may see contextual action buttons that open AVICENNA with relevant context:

Explain: ask AVICENNA to explain the current artifact
Modify: ask AVICENNA to help modify the current artifact
Generate Tests: ask AVICENNA to generate test scenarios for the current artifact

These actions open the AVICENNA panel pre-loaded with the artifact context, so the conversation starts with the right information.

Connector Authorization

During a conversation, AVICENNA may determine that a connector needs to be authorized before it can proceed. When this happens, the conversation pauses and presents a connector authorization prompt. Complete the authorization, and the conversation resumes automatically.

Credential Safety

If you type something that looks like a credential (API key, password, token) directly into the chat, AVICENNA displays a warning:

Credential pattern detected Manage credentials in your project settings under the Integrate tab.

Credentials should be configured through the Integrate tab, not pasted into conversation.

Conversation History

Click the clock icon in the panel header to view your conversation history for the current project. You can:

Resume a previous conversation
Start a new conversation
See the topic and timestamp of each past conversation

What AVICENNA Cannot Do

Override VB-OS evaluation: AVICENNA configures governance artifacts, but evaluation remains deterministic. AVICENNA cannot change how VB-OS evaluates a workload against a boundary.
Act without authorization: every creation, modification, or deletion requires your explicit confirmation through the confirmation gate.
Access resources outside your project: AVICENNA operates within the project and organization context of your current session.
Execute external actions: AVICENNA configures governance within VB-OS. It does not execute actions in external systems.

What’s Next

Boundaries : manage the boundaries AVICENNA creates
Flows : configure post-evaluation behavior
Connectors : manage evidence source connections
Evaluations : run evaluations against your governance setup