Get what a simulation episode observed
Projects one episode’s session transcript into what its checks graded: the agent’s replies and every tool call across the session tree, each marked whether it was recorded instead of executed. Read on demand from the session transcript through the caller’s session access; an episode that started no session observes nothing.
Authorizations
A Recursion API key, created in the console under API keys.
Headers
Optional tenant consistency check. When present, it must name the tenant bound to the API key, including when x-organization-id is default.
1Organization in which to act. Required for a tenant-scoped API key. Optional for an organization-scoped key, where it must name that same organization. The value may be an organization id or default.
1Path Parameters
Simulation run id.
Simulation episode id.
Response
OK
The agent's replies, at most 100: the simulated Slack thread posts for a Slack trigger, otherwise the episode session's assistant messages, in order. The last is the final message.
Tool calls across the episode's session tree, in transcript order: at most 500.
True when replies past the first 100, tool calls past the first 500, or a call's arguments over 16 KiB were left out. The session itself holds everything.
How long the episode's turn ran.
The episode's session. Empty, with nothing observed, when no session started.
The session status checks read; empty while the session is still working.
completed, failed, cancelled, idle, awaiting_human