Recall profiles

Pick the recall path that matches the task risk.

Recall profiles let operators trade latency, evidence, and depth for different agent workflows.

Install pathPyPI package
RuntimeHosted Cloud
InterfacePython + CLI + MCP
Releasev2.1.3
profiles

Start balanced, increase depth when the decision is risky.

fast

Latency-first recall for lightweight context refreshes.

balanced

Default profile for most coding and operator workflows.

verified

Adds stronger evidence posture when the memory will affect a decision.

deep

Use for richer recall paths where quality matters more than speed.

example

Pass the profile with the recall request.

client.recall(
    "billing rollout risks",
    profile="verified",
    limit=8,
)
boundary

Profiles do not make generated answers automatically true.

Recall improves the evidence available to an agent. The final answer still needs normal review, tool verification, and product judgment.