Skip to main content
This command connects your chatbot to a single, already-created simulation and processes its scenarios until the simulation reaches a terminal state.
As of client 1.0.0, connect replaces the previous start command and the --simulation-id option is required. Each connect process is dedicated to exactly one simulation. To create a simulation and connect to it in a single step, use snowglobe simulate or snowglobe replay instead.

​
Prerequisites

  • Installed and authenticated snowglobe
  • Wrapper created and tested
  • .snowglobe/agents.json and .snowglobe/config.rc exist (created by init)
  • A simulation already created in Snowglobe, and its simulation ID

​
What this does

  • Binds your wrapper so it can call Snowglobe during the session
  • Processes scenarios from the specified simulation and sends them to your chatbot
  • Sends the chatbot’s response back to Snowglobe
  • Streams basic status to your terminal while active
  • Exits once the simulation reaches a terminal state

​
Getting a simulation ID

Create a simulation in the Snowglobe dashboard, then copy its simulation ID from the simulation’s page (or its URL). Pass that value to --simulation-id.

​
Usage

snowglobe connect --simulation-id <simulation-id>

​
Running with multiple agents

If your project has more than one agent registered, snowglobe connect will display a numbered list of your agents and ask you to choose which ones to include in the session. You can select individual agents, ranges, or enter * to run all of them. Only the agents you select will be active during the session. To skip the prompt (e.g. in CI or automated workflows), pass agent IDs directly: 
snowglobe connect --simulation-id <simulation-id> --agent-ids <id1>,<id2>
snowglobe connect --simulation-id <simulation-id> --agent-ids '*'
Agent IDs can be found in the selection table or in your .snowglobe/agents.json file.
Running fewer agents at a time means faster polling and quicker test results — only select the agents you need for your current session.

​
During the run

  • Keep this terminal session open while the simulation runs
  • The process exits automatically once the simulation reaches a terminal state
  • Use Ctrl+C to stop early