Skip to main content

PLAYGROUND / INTERACTION LAB

Make workflow state visible.

Drag a workflow layer, run a request, and observe how models, knowledge, and tools advance continuously within a workspace.

Use it in Aivory View product model

INTERACTION LAB

Use your hands to understand the workspace.

First drag a stacked layer, then run a request path to see models, knowledge, and tools move through the workspace.

STACK / WORKSPACE FLOW

Split a request into observable layers.

Drag the top layer, or select any layer to bring it forward.

Current layer: Retrieve relevant context

WORKSPACE / REQUEST PATH

Run a workspace request path.

Ready to run
  1. 01ModelWaiting for the prior step
  2. 02ContextWaiting for the prior step
  3. 03ToolsWaiting for the prior step
  4. 04AnswerWaiting for the prior step
›Run it to observe Model → Context → Tools → Answer.
Request example
// Configure channels and tools in the Admin Console first.
const request = {
  model: 'gpt-5',
  knowledge_base_ids: ['kb_product_notes'],
  selected_tool_ids: ['builtin:aivory_web_search'],
}

// Send this request through an authenticated Aivory session.

WORKSPACE POLICY

Make models and tools follow workspace policy.

Models, knowledge bases, and tools are configured by administrators and follow explicit selection and permission boundaries on every request.

View channel and model configuration

Multi-channel models

Model records belong to a channel, so the same model ID can be configured independently across different channels.

RAG knowledge bases

Retrieved results and source citations remain in the workspace to serve long-running work.

Tools and MCP

Provider-hosted tools are never added by default. Built-in tools and MCP both require administrator approval.

Move from this interaction to a real deployment.

Open personal edition deployment