A self-improving RLM agent for coding workflows and long-running autonomous tasks.
Rust
Mark the rlm.harness calls as synchronous in the prompt (#3435)
Sebastian Müllerc5bf5ae4740f
diff --git a/crates/pa-core/src/prompts/layers/core.md b/crates/pa-core/src/prompts/layers/core.mdindex 89069808be..5b90cb5352 100644--- a/crates/pa-core/src/prompts/layers/core.md+++ b/crates/pa-core/src/prompts/layers/core.md@@ -87,7 +87,7 @@ Memories are created by two mechanisms:- Effect:- The refinement event is always saved in the harness state's refinement history- A harness message is sent to the agent with the refinement result-- Active memory management by the agent+- Active memory management by the agent (the `rlm.harness.*` and `rlm.get_harness_state` calls are synchronous: call them without `await`)- `refine.run(instructions: str | None = None, global_: bool = False) -> dict`: agent-triggered refinement (see above); returns immediately and runs when the current turn ends- `refine.status() -> dict`: whether a refinement is already pending for this turn or currently in flight- `rlm.harness.create_memory(title: str, content: str, *, id: str | None = None, path: str = "general", metadata: dict | None = None, global_: bool = False) -> HarnessEntry`: creates a memory; use `global_=True` for cross-session entries (Python reserves `global`, so the parameter is spelled `global_`)