Skip to content

Capability Labs

The labs are deterministic environments for exercising the toolkit chain end to end.

  • Lab L1 runs a local Python fixture and exercises the basic Seam -> meshmapper -> optional Assay flow.
  • Lab L2 runs a Docker mini-mesh with separate support, planner, and billing services.
  • Lab L3 runs a deterministic framework-style matrix for LangGraph, CrewAI, AutoGen, OpenAI Agents, and Microsoft Agent Framework shapes with Assay framed trials.
  • Lab L4 runs a real LangGraph runtime mesh with deterministic graph nodes, Assay robustness labels, and rendered validation reports.
  • Lab L5 runs a content-decision mesh where the target only succeeds when a Seam proxy rule rewrites message content in path.
  • Lab L6 runs a Dockerized full agent mesh with A2A, MCP WebSocket, memory, policy gates, internal Seam intercepts, meshmapper hypotheses, optional Assay validation, robustness, and reports.
  • Lab L7/L8 run CrewAI- and AutoGen-shaped deterministic decision meshes using the L5 content-decision chain and framework/runtime metadata.
  • Lab L9/L10 run OpenAI Agents-style and Microsoft Agent Framework-style deterministic decision meshes with the same artifact contract and no live model calls by default.
  • Range R2 provides compact, standard, and full-split professional range profiles with Ludus configs, Ansible inventories, and helper scripts.

The labs are non-LLM and local by default. They test tool capability and artifact quality before optional live LLM or external service targets enter the matrix.

Use Professional Range Deployment when you want to move from local Docker demos into a reusable cyber range with VM boundaries, routed traffic, and shared provisioning.