rlm-mem

Persistent agent memory enables cross-session continuity by storing chunks in a layered JSONL memory store with deterministic linking. This runtime applies policy-scoped write/read separation to ensure safe recall with provenance and safeguards.

Policy-scoped memory management enforces read and write separation with optional user-global controls to securely persist agent memories. It uses mandatory guard scripts and deterministic chunking to maintain provenance-rich recall across long-running workloads.

To initialize a secure memory runtime, set up the canonical package path at your project root and execute the mandatory guard checks. This verifies memory store integrity and enforces the required security policies before runtime operation.

Deterministic memory chunking supports long-running agent workloads by consistently dividing and linking memory chunks for provenance-rich recall. The runtime applies safe recall policies to manage memory-driven decision support across sessions.

Standard memory lacks policy-scoped write/read separation and mandatory guard scripts, risking unsafe recall and lost provenance across sessions. A canonical runtime with a layered JSONL store is required to enforce security and deterministic chunking.