RECORDED EXECUTION

Your search, made visible.

Waiting

Evaluator score—No evaluated answer yet

All-in calls—Preparation + application

Tokens / costUnknownCost not reported

Follow the work as it happens.

Connect LiveInspector() to your runtime.
Nodes appear here as actual attempts begin.

with LiveInspector() as inspector:
    sdk = DreamRSI(..., inspector=inspector)
    result = await sdk.run(task)

Try dreamrsi watch --demo without a model.

RECORDED REPLAY

Compare only outcomes present in a frozen tree. A missing continuation is unknown, not a failed answer.

No comparison recorded for this tree.

await inspector.compare(sdk, world, [policy_a, policy_b])

Uses the configured replay engine. Never promotes a policy.