gazebo-operator

What problem does it solve?

Turning natural-language requests into safe, correct Gazebo actions without manually figuring out the right gz topic/service endpoints, message types, or payload structure.

Core Features & Use Cases

• Runtime inspection & capability discovery: Inspects active world(s), services, topics, entities, and available interfaces before acting.
• Target and interface resolution: Resolves ambiguous natural-language targets via ranked candidate matching and selects the best transport interface card.
• Schema-aware payload generation: Generates gz.msgs.* payloads using the local references/gz_msg_registry.json registry, with fallbacks to field-path matching.
• Safe execution with verification: Executes actions through bundled scripts and verifies outcomes (pose, velocity effect heuristics, pause/step state, and transport acknowledgements).
• Optional extension via adapters: Supports custom actions through Python adapters under adapters/ for world/topic patterns not covered by the core set.

Quick Start

Ask the AI to move a model in the running simulation by describing which entity and motion fields you want, for example:

• "Move vehicle_blue by setting linear x to 0.5 and angular z to 0.05."