AxContextPolicyConfig
Defined in: https://github.com/ax-llm/ax/blob/5b28f9093bb70863b59459bb6df5062d005bce41/src/ax/agent/rlm.ts#L448
Public context policy for the Actor loop.
Users choose replay style via preset and overall prompt budget via budget.
Properties
| Property | Type | Description |
|---|---|---|
budget? | AxContextPolicyBudget | Overall prompt budget and compression aggressiveness. |
preset? | AxContextPolicyPreset | Opinionated preset for how the agent should replay and compress context. - full: prefer raw replay of earlier actions - adaptive: balance replay detail with checkpoint compression while keeping more recent evidence visible - lean: prefer live state + compact summaries over raw replay detail - checkpointed: keep full replay until the rendered actor prompt grows beyond the selected budget, then replace older successful turns with a checkpoint summary |