Agent runtime

The runtime is what an agent server builds from the account's configuration and runs every agent against. This section opens with a summary of every control, then says what the runtime holds and how a change reaches every agent server, then covers the four parts that limit what an expression may do, grouped in the UI under Guardrails: the allowlist, the system variables, the compile rules and the runtime limits. The last page covers the connection pools every HTTP call goes through.

• Controlled executionThe agent runtime is the environment the agent server builds from the account's configuration and runs every agent in. It limits every agent in time and cost…
• Runtime syncAn agent server holds no state of its own, but it does hold a runtime: an in-memory copy of everything an agent needs to run. The runtime is built from the a…
• AllowlistThe allowlist is the complete set of names that an agent's Python expressions may use. Expressions do not get Python's builtins; the allowlist takes their pl…
• System variablesThe sys namespace carries what the runtime knows about a run. Four names are always there. Four more reach an agent only when a developer enables them here, …
• Compile rulesThe compile rules decide what an expression may contain: for each of the ten binary operators, whether an expression may use it as Python does, may not use i…
• Runtime limitsThe runtime limits cap a run: how long an agent may run, how many times a loop may iterate, how many rows a database call may return, how long an agent defin…
• Connection poolsEvery HTTP connection an agent server opens goes through a connection pool: the calls to Search2o Cloud, the LLM calls, and the calls made by api commands th…