Memory index
Filesystem-native indexing with BM25, embeddings, reranking, and compiled operational facts.
Platform
Built for operational memory, not a demo chat box.
Remanentia is structured around durable recall, measurable retrieval, controlled answer generation, and a clear boundary between internal and public-facing data.
Agents and applications
DIRECTOR-CLASS-AIRoute, verify, guard
REMANENTIARecall, vectors, compiled memory
Allowed corporaPublic docs, selected project facts, internal-only memory
Scheduled worker
Supervised refresh keeps vector state current and skips unchanged corpora instead of rebuilding unnecessarily.
Factual guard
DIRECTOR-CLASS-AI checks retrieved evidence before an answer is accepted for public or operational use.
Public safety
Public APIs use corpus allowlists and result redaction. Private logs stay internal unless deliberately released.
Local-first backends
Local model servers are preferred where available, with explicit hosted and no-generation fallback modes.
Agent memory backend
Agents can use REMANENTIA as a memory service while DIRECTOR-CLASS-AI manages verification and policy.