MMESHSYSTEMS

MESH AI PLATFORM / DOCUMENTATION

Model Catalog & AI Playground

Choose approved model profiles and test response and Tool-call behavior separately.

Configuration is not verification

The Model Catalog distinguishes configuration, credential presence, inference enablement, Project approval and prior verification. Merely opening the catalog does not contact a model. A prior successful check is not a current health guarantee.

Test a model

  1. Ask an administrator to configure the approved provider and model on the server.
  2. Approve an immutable model profile for the Project where authorized.
  3. Select that profile in an Agent version.
  4. Explicitly authorize any inference test and possible provider charges.
  5. Test response behavior and actual Tool-call behavior separately; inspect latency, safe errors and trace metadata.

Playground scope

The Agent Playground runs against a selected immutable Agent version and current grants. Model testing must not enable a paid provider silently. An unavailable cost measurement does not mean a call is free.