Reference
Memory Layer
Seren builds persistent memory of past investigations, service relationships, and known failure patterns to make each investigation faster.
R
Redis
Working memory. Active investigation context with TTL-based expiry.
Q
Qdrant
Long-term semantic memory. Past investigations retrieved by similarity.
N
Neo4j
Service dependency graph. Entities, relationships, known failure patterns.
S
Supabase
Audit log, user data, encrypted integration tokens, memory versioning.
Contradiction Resolution
When Seren learns something that conflicts with existing memory — a service that moved teams, an escalation policy that changed — it classifies the conflict and updates the belief state rather than appending noise.